DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 07/08/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 1 recites the limitation “the performance” in line 12. There is insufficient antecedent basis for this limitation in the claim. For examination purposes, “the performance” has been interpreted as “a performance”.
Claim 14 recites the limitation “the performance” in line 12. There is insufficient antecedent basis for this limitation in the claim. For examination purposes, “the performance” has been interpreted as “a performance”.
Claim 19 recites the limitation “the performance” in line 12. There is insufficient antecedent basis for this limitation in the claim. For examination purposes, “the performance” has been interpreted as “a performance”.
Dependent claims 2-13 are rejected based on being directly or indirectly dependent on rejected claim 1.
Dependent claims 15-18 are rejected based on being directly or indirectly dependent on rejected claim 14.
Dependent claim 20 is rejected based on being directly or indirectly dependent on rejected claim 19.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1,
Claim 1 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 1 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“generating, … using a plurality of source tables for a prediction domain, a plurality of subdomain-specific graphs for the prediction domain, each comprising a respective plurality of graph nodes and a respective plurality of weighted edges between the respective plurality of graph nodes”
“generating … a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs”
“generating … a composite graph embedding based on the plurality of subdomain-specific embeddings and a designated predictive task”
“initiating … the performance of the designated predictive task based on the composite graph embedding”
As drafted, under their broadest reasonable interpretations, cover mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitations in the context of this claim encompass generating a plurality of subdomain-specific graphs for the prediction domain using a plurality of source tables, each subdomain-specific graph comprising nodes and weighted edges (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use source tables for a prediction domain to generate a plurality of subdomain-specific graphs comprising respective nodes and weighted edges for the prediction domain); generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs); generating a composite graph embedding based on the subdomain-specific embeddings and a designated predictive task (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the subdomain-specific embeddings and a designated predictive task to generate a composite graph embedding); and initiating the performance of the designated predictive task based on the composite graph embedding (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the composite graph embedding to initiate the performance of the designated predictive task).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitations:
“computer”
“by one or more processors”
“by the one or more processors”
“using a graph-based machine learning model”
“using the graph-based machine learning model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 2,
Claim 2 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 2 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding”
As drafted, is part of the abstract idea of claim 1 of generating a plurality of subdomain-specific embeddings. The limitation of claim 2 further limits the limitation of claim 1 by further defining the subdomain-specific embeddings. The above limitation in the context of this claim encompasses generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph, the subdomain-specific embedding being based on attention weights assigned to graph nodes and weighted edges of a corresponding subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs based on attention weights assigned to the graph nodes and weighted edges of the subdomain-specific graph).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 1 of a generic computer, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 3,
Claim 3 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 3 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein the plurality of attention weights comprises a plurality of node-level weights”
As drafted, is part of the abstract idea of claim 2 of generating subdomain-specific embeddings based on attention weights. The limitation of claim 3 further limits the limitation of claim 2 by further defining what the attention weights comprise. The above limitation in the context of this claim encompasses generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph, the subdomain-specific embedding being based on attention weights comprising node-level weights assigned to graph nodes and weighted edges of a corresponding subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs based on attention weights comprising node-level weights assigned to the graph nodes and weighted edges of the subdomain-specific graph). The limitation:
“generating … the plurality of node-level weights for the plurality of graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the plurality of graph nodes”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses generating the node-level weights for the graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the graph nodes (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate the node-level weights based on a plurality of node attributes corresponding to the graph nodes).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitations:
“computer”
“using the graph-based machine learning model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 2 of a generic computer, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 4,
Claim 4 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 4 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein the plurality of node attributes comprises one or more node labels for the designated predictive task”
As drafted, is part of the abstract idea of claim 3 of generating the plurality of node-level weights based on a plurality of node attributes. The limitation of claim of claim 4 further limits the limitation of claim 3 by further defining what the node attributes comprise. The above limitation in the context of this claim encompasses generating the node-level weights for the graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the graph nodes, the node attributes comprising labels for the designated predictive task (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate the node-level weights based on a plurality of node attributes comprising labels for the designated predictive task corresponding to the graph nodes). The limitations:
“generating, using a semi-supervised loss function, a model loss for the graph-based machine learning model based on the composite graph embedding”
“updating, using a machine learning training technique, the composite graph embedding based on the model loss”
As drafted, under their broadest reasonable interpretations, cover mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitations in the context of this claim encompass generating a model loss for the graph-based machine learning model based on the composite graph embedding using a semi-supervised loss function (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use a semi-supervised loss function to generate a model loss for the graph-based machine learning model based on the composite graph embedding); and updating the composite graph embedding based on the model loss using a machine learning training technique (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the model loss to update the composite graph embedding).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 3 of a generic computer, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 5,
Claim 5 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 5 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein the plurality of attention weights comprises a plurality of semantic-level weights”
As drafted, is part of the abstract idea of claim 2 of generating subdomain-specific embeddings based on attention weights. The limitation of claim 5 further limits the limitation of claim 2 by further defining what the attention weights comprise. The above limitation in the context of this claim encompasses generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph, the subdomain-specific embedding being based on attention weights comprising semantic-level weights assigned to graph nodes and weighted edges of a corresponding subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs based on attention weights comprising semantic-level weights assigned to the graph nodes and weighted edges of the subdomain-specific graph). The limitation:
“generating … the plurality of semantic-level weights for the plurality of weighted edges of the subdomain-specific graph based on one or more metapaths within the subdomain-specific graph”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses generating the semantic-level weights for the weighted edges of the subdomain-specific graph based on metapaths within the subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate the semantic-level weights based on metapaths within the subdomain-specific graph).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitations:
“computer”
“using the graph-based machine learning model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 2 of a generic computer, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 6,
Claim 6 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 6 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein the designated predictive task is a machine learning classification task”
As drafted, is part of the abstract idea of claim 1 of generating a composite graph embedding based on the subdomain-specific embeddings and a designated predictive task. The limitation of claim 6 further limits the limitation of claim 1 by further defining what the designated predictive task comprises. The above limitation in the context of this claim encompasses generating a composite graph embedding based on the subdomain-specific embeddings and a designated predictive task that is a machine learning classification task (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the subdomain-specific embeddings and a designated predictive task that is a machine learning classification task to generate a composite graph embedding). The limitation:
“generating … a predictive classification for an unlabeled graph node associated with the plurality of subdomain-specific graphs”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses generating a predictive classification for an unlabeled graph node associated with the subdomain-specific graphs (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate a predictive classification for an unlabeled graph node associated with the subdomain-specific graphs).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitations:
“computer”
“using a machine learning classification model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 1 of a generic computer, processors, and machine learning models, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 7,
Claim 7 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 7 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“a plurality of graph nodes for a subdomain-specific graph of the plurality of subdomain-specific graphs comprises a set of common graph nodes that are within each of the plurality of subdomain-specific graphs and a set of subdomain-specific graph nodes specific to the subdomain-specific graph, and the set of common graph nodes comprises the unlabeled graph node”
As drafted, is part of the abstract idea of claim 6 of generating a predictive classification for an unlabeled graph node. The limitation of claim 7 further limits the limitation of claim 6 by further defining the unlabeled graph node. The above limitation in the context of this claim encompasses generating a predictive classification for an unlabeled graph node associated with the subdomain-specific graphs including graph nodes comprising a set of common graph nodes that are within each of the plurality of subdomain-specific graphs and a set of subdomain-specific graph nodes specific to the subdomain-specific graph, the unlabeled graph node being within the common graph nodes (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate a predictive classification for an unlabeled graph node associated with the subdomain-specific graphs including graph nodes comprising a set of common graph nodes that are within each of the plurality of subdomain-specific graphs and a set of subdomain-specific graph nodes specific to the subdomain-specific graph, the unlabeled graph node being within the common graph nodes).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 6 of a generic computer, processors, and machine learning models, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning models for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 8,
Claim 8 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 8 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“modifying each of the plurality of subdomain-specific graphs by assigning the predictive classification to the unlabeled graph node”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses modifying the plurality of subdomain-specific graphs by assigning the predictive classification to the unlabeled graph node (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can assign the predictive classification to the unlabeled graph node to modify the subdomain-specific graphs).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 7 of a generic computer, processors, and machine learning models, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning models for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 9,
Claim 9 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 9 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein each of the plurality of subdomain-specific graphs comprises a separate heterogeneous and undirected graph data structure”
As drafted, is part of the abstract idea of claim 1 of generating a plurality of subdomain-specific graphs. The limitation of claim 9 further limits the limitation of claim 1 by further defining what the subdomain-specific graphs comprise. The above limitation in the context of this claim encompasses generating a plurality of subdomain-specific graphs for the prediction domain using a plurality of source tables, each subdomain-specific graph comprising nodes and weighted edges and a separate heterogeneous and undirected graph data structure (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use source tables for a prediction domain to generate a plurality of subdomain-specific graphs comprising respective nodes and weighted edges and a separate heterogeneous and undirected graph data structure for the prediction domain).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 1 of a generic computer, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 10,
Claim 10 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 10 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“wherein each of the plurality of source tables comprise respective subdomain data for a subdomain of the prediction domain”
As drafted, is part of the abstract idea of claim 1 of generating a plurality of subdomain-specific graphs using a plurality of source tables for a prediction domain. The limitation of claim 10 further limits the limitation of claim 1 by further defining what the plurality of source tables comprises. The above limitation in the context of this claim encompasses generating a plurality of subdomain-specific graphs for the prediction domain using a plurality of source tables comprising respective subdomain data for a subdomain of the prediction domain, each subdomain-specific graph comprising nodes and weighted edges (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use source tables comprising respective subdomain data for a subdomain of the prediction domain for a prediction domain to generate a plurality of subdomain-specific graphs comprising respective nodes and weighted edges for the prediction domain). The limitation:
“wherein … a subdomain-specific graph of the plurality of subdomain-specific graphs is generated based on subdomain data from a corresponding source table of the plurality of source tables”
As drafted, under its broadest reasonable interpretations, cover mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses generating a subdomain-specific graph based on subdomain data from a corresponding source table (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use subdomain data from a corresponding source table to generate a subdomain-specific graph).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 1 of a generic computer, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 11,
Claim 11 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 11 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“in response to the one or more modification data objects, regenerating the subdomain-specific graph”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)) and insignificant extra-solution activity (See MPEP 2106.05(g)). The above limitation in the context of this claim encompasses regenerating the subdomain-specific graph in response to the one or more modification data objects (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the one or more modification data objects to regenerate the subdomain-specific graph).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) or insignificant extra-solution activity (See MPEP 2106.05(g)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). The limitation:
“receiving one or more modification data objects associated with the corresponding source table”
As drafted, is an additional element that corresponds to insignificant extra-solution activity. In particular, the additional elements are merely directed towards mere data gathering. See MPEP 2106.05(g). In addition, the recitation of additional elements in claim 10 of a generic computer, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) or insignificant extra-solution activity (i.e. receiving data). Furthermore, the “receiving …” limitation is insignificant extra-solution activity that is well-understood, routine, and conventional according to MPEP 2106.05(d) (“The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity… i. Receiving or transmitting data over a network). Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 12,
Claim 12 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 12 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: Please see the analysis of claim 11. The limitations of claim 12 are only additional elements to the abstract ideas of claim 11.
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) or insignificant extra-solution activity (See MPEP 2106.05(g)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). The limitation:
“wherein the one or more modification data objects are received at a defined time interval”
As drafted, is an additional element that corresponds to insignificant extra-solution activity. In particular, the additional elements are merely directed towards mere data gathering. See MPEP 2106.05(g). In addition, the recitation of additional elements in claim 11 of a generic computer, processors, and machine learning model are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Furthermore, the “receiving …” limitation of claim 11 is an additional element that corresponds to insignificant extra-solution activity as mere data gathering. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) or insignificant extra-solution activity (i.e. receiving data). Furthermore, the “receiving …” and “… are received …” limitations are insignificant extra-solution activity that is well-understood, routine, and conventional according to MPEP 2106.05(d) (“The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity… i. Receiving or transmitting data over a network). Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 13,
Claim 13 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 13 is directed to a method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“wherein the prediction domain comprises a clinical domain”
As drafted, is part of the abstract idea of claim 12 of generating a plurality of subdomain-specific graphs using a plurality of source tables for a prediction domain. The limitation of claim 13 further limits the limitation of claim 12 by further defining what the prediction domain comprises. The above limitation in the context of this claim encompasses generating a plurality of subdomain-specific graphs for the prediction domain comprising a clinical domain using a plurality of source tables comprising respective subdomain data for a subdomain of the prediction domain, each subdomain-specific graph comprising nodes and weighted edges (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use source tables comprising respective subdomain data for a subdomain of the prediction domain for a prediction domain comprising a clinical domain to generate a plurality of subdomain-specific graphs comprising respective nodes and weighted edges for the prediction domain).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) or insignificant extra-solution activity (See MPEP 2106.05(g)). The limitation:
“computer”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). The limitation:
“wherein … the defined time interval is associated with a claim aggregation frequency”
As drafted, is an additional element that is part of the insignificant extra-solution activity of claim 12 of receiving modification data objects at a defined time interval. In addition, the recitation of additional elements in claim 12 of a generic computer, processors, and machine learning model are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Furthermore, the “receiving …” and “… are received …” limitations of claim 12 are additional elements that corresponds to insignificant extra-solution activity as mere data gathering. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic computer, processors, and machine learning model for applying the abstract ideas) or insignificant extra-solution activity (i.e. receiving data). Furthermore, the “receiving …” and “… are received …” limitations are insignificant extra-solution activity that is well-understood, routine, and conventional according to MPEP 2106.05(d) (“The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity… i. Receiving or transmitting data over a network). Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 14,
Claim 14 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 14 is directed to a system, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“generate, using a plurality of source tables for a prediction domain, a plurality of subdomain-specific graphs for the prediction domain, each comprising a respective plurality of graph nodes and a respective plurality of weighted edges between the respective plurality of graph nodes”
“generate … a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs”
“generate … a composite graph embedding based on the plurality of subdomain-specific embeddings and a designated predictive task”
“initiate the performance of the designated predictive task based on the composite graph embedding”
As drafted, under their broadest reasonable interpretations, cover mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitations in the context of this claim encompass generating a plurality of subdomain-specific graphs for the prediction domain using a plurality of source tables, each subdomain-specific graph comprising nodes and weighted edges (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use source tables for a prediction domain to generate a plurality of subdomain-specific graphs comprising respective nodes and weighted edges for the prediction domain); generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs); generating a composite graph embedding based on the subdomain-specific embeddings and a designated predictive task (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the subdomain-specific embeddings and a designated predictive task to generate a composite graph embedding); and initiating the performance of the designated predictive task based on the composite graph embedding (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the composite graph embedding to initiate the performance of the designated predictive task).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) or insignificant extra-solution activity (See MPEP 2106.05(g)). The limitations:
“memory”
“one or more processors communicatively coupled to the memory”
“the one or more processors”
“using a graph-based machine learning model”
“using the graph-based machine learning model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic memory, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 15,
Claim 15 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 15 is directed to a system, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding”
As drafted, is part of the abstract idea of claim 14 of generating a plurality of subdomain-specific embeddings. The limitation of claim 15 further limits the limitation of claim 14 by further defining the subdomain-specific embeddings. The above limitation in the context of this claim encompasses generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph, the subdomain-specific embedding being based on attention weights assigned to graph nodes and weighted edges of a corresponding subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs based on attention weights assigned to the graph nodes and weighted edges of the subdomain-specific graph).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The recitation of additional elements in claim 14 of a generic memory, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic memory, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 16,
Claim 16 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 16 is directed to a system, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein the plurality of attention weights comprises a plurality of node-level weights”
As drafted, is part of the abstract idea of claim 15 of generating subdomain-specific embeddings based on attention weights. The limitation of claim 16 further limits the limitation of claim 15 by further defining what the attention weights comprise. The above limitation in the context of this claim encompasses generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph, the subdomain-specific embedding being based on attention weights comprising node-level weights assigned to graph nodes and weighted edges of a corresponding subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs based on attention weights comprising node-level weights assigned to the graph nodes and weighted edges of the subdomain-specific graph). The limitation:
“generate … the plurality of node-level weights for the plurality of graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the plurality of graph nodes”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses generating the node-level weights for the graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the graph nodes (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate the node-level weights based on a plurality of node attributes corresponding to the graph nodes).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“the one or more processors”
“using the graph-based machine learning model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 15 of a generic memory, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic memory, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 17,
Claim 17 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 17 is directed to a system, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein the plurality of node attributes comprises one or more node labels for the designated predictive task”
As drafted, is part of the abstract idea of claim 16 of generating the plurality of node-level weights based on a plurality of node attributes. The limitation of claim of claim 17 further limits the limitation of claim 16 by further defining what the node attributes comprise. The above limitation in the context of this claim encompasses generating the node-level weights for the graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the graph nodes, the node attributes comprising labels for the designated predictive task (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate the node-level weights based on a plurality of node attributes comprising labels for the designated predictive task corresponding to the graph nodes). The limitations:
“generate, using a semi-supervised loss function, a model loss for the graph-based machine learning model based on the composite graph embedding”
“update, using a machine learning training technique, the composite graph embedding based on the model loss”
As drafted, under their broadest reasonable interpretations, cover mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitations in the context of this claim encompass generating a model loss for the graph-based machine learning model based on the composite graph embedding using a semi-supervised loss function (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use a semi-supervised loss function to generate a model loss for the graph-based machine learning model based on the composite graph embedding); and updating the composite graph embedding based on the model loss using a machine learning training technique (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the model loss to update the composite graph embedding).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“the one or more processors”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 16 of a generic memory, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic memory, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 18,
Claim 18 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 18 is directed to a system, which is directed to a machine, one of the statutory categories.
Step 2A Prong One Analysis: The limitation:
“wherein the plurality of attention weights comprises a plurality of semantic-level weights”
As drafted, is part of the abstract idea of claim 15 of generating subdomain-specific embeddings based on attention weights. The limitation of claim 18 further limits the limitation of claim 15 by further defining what the attention weights comprise. The above limitation in the context of this claim encompasses generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph, the subdomain-specific embedding being based on attention weights comprising semantic-level weights assigned to graph nodes and weighted edges of a corresponding subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs based on attention weights comprising semantic-level weights assigned to the graph nodes and weighted edges of the subdomain-specific graph). The limitation:
“generate … the plurality of semantic-level weights for the plurality of weighted edges of the subdomain-specific graph based on one or more metapaths within the subdomain-specific graph”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses generating the semantic-level weights for the weighted edges of the subdomain-specific graph based on metapaths within the subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate the semantic-level weights based on metapaths within the subdomain-specific graph).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitations:
“the one or more processors”
“using the graph-based machine learning model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 15 of a generic memory, processors, and machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe a generic memory, processors, and machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 19,
Claim 19 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 19 is directed to non-transitory computer-readable storage media, which is directed to an article of manufacture, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“generate, using a plurality of source tables for a prediction domain, a plurality of subdomain-specific graphs for the prediction domain, each comprising a respective plurality of graph nodes and a respective plurality of weighted edges between the respective plurality of graph nodes”
“generate … a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs”
“generate … a composite graph embedding based on the plurality of subdomain-specific embeddings and a designated predictive task”
“initiate the performance of the designated predictive task based on the composite graph embedding”
As drafted, under their broadest reasonable interpretations, cover mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitations in the context of this claim encompass generating a plurality of subdomain-specific graphs for the prediction domain using a plurality of source tables, each subdomain-specific graph comprising nodes and weighted edges (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use source tables for a prediction domain to generate a plurality of subdomain-specific graphs comprising respective nodes and weighted edges for the prediction domain); generating subdomain-specific embeddings comprising a respective subdomain-specific embedding for each subdomain-specific graph (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate subdomain-specific embeddings for each of the subdomain-specific graphs); generating a composite graph embedding based on the subdomain-specific embeddings and a designated predictive task (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the subdomain-specific embeddings and a designated predictive task to generate a composite graph embedding); and initiating the performance of the designated predictive task based on the composite graph embedding (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the composite graph embedding to initiate the performance of the designated predictive task).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)) or insignificant extra-solution activity (See MPEP 2106.05(g)). The limitations:
“one or more processors”
“the one or more processors”
“using a graph-based machine learning model”
“using the graph-based machine learning model”
As drafted, are additional elements that amount to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe generic processors and a machine learning model for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Regarding Claim 20,
Claim 20 is rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1 Analysis: Claim 20 is directed to non-transitory computer-readable storage media, which is directed to an article of manufacture, one of the statutory categories.
Step 2A Prong One Analysis: The limitations:
“wherein the designated predictive task is a machine learning classification task”
As drafted, is part of the abstract idea of claim 19 of generating a composite graph embedding based on the subdomain-specific embeddings and a designated predictive task. The limitation of claim 20 further limits the limitation of claim 19 by further defining what the designated predictive task comprises. The above limitation in the context of this claim encompasses generating a composite graph embedding based on the subdomain-specific embeddings and a designated predictive task that is a machine learning classification task (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can use the subdomain-specific embeddings and a designated predictive task that is a machine learning classification task to generate a composite graph embedding). The limitation:
“generating … a predictive classification for an unlabeled graph node associated with the plurality of subdomain-specific graphs”
As drafted, under its broadest reasonable interpretation, covers mental processes (concepts performed in the human mind (including an observation, evaluation, judgement, opinion)) but for the recitation of mere instructions to apply language (See MPEP 2106.05(f)). The above limitation in the context of this claim encompasses generating a predictive classification for an unlabeled graph node associated with the subdomain-specific graphs (corresponds to evaluation and judgement; in particular, a human, with the assistance of pen and paper, can generate a predictive classification for an unlabeled graph node associated with the subdomain-specific graphs).
Step 2A Prong Two Analysis: The judicial exceptions are not integrated into a practical application. In particular, the claim recites additional elements that are mere instructions to apply (See MPEP 2106.05(f)). The limitation:
“using a machine learning classification model”
As drafted, is an additional element that amounts to no more than mere instructions to apply the exception for the abstract ideas. See MPEP 2106.05(f). In addition, the recitation of additional elements in claim 19 of generic processors and a machine learning model, as drafted, are reciting mere instructions to apply language such that it amounts to no more than mere instructions to apply the exceptions. Therefore, the additional elements do not integrate the abstract ideas into a practical application.
Step 2B Analysis: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, all of the additional elements are “mere instructions to apply an exception” (I.e. the additional elements describe generic processors and machine learning models for applying the abstract ideas) Mere instructions to apply an exception cannot provide an inventive concept. The claim is not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1, 6-10, 14, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Gandhi et al. (US 11,593,622 B1) in view of Luo et al. (US 2022/0180240 A1).
Regarding Claim 1,
Gandhi et al. teaches a computer-implemented method (Fig. 1; Col. 2, lines 40-45: "The present disclosure relates to methods and apparatus for training and using graph convolutional network (GCN) models to generate predictions with respect to data sets representing instances of multiple entity types, with multiple types of relationships among the instances of the entity types" teaches a method for using graph based machine learning for generating predictions. Fig. 10; Col. 20, lines 51-53: "Portions or all of multiple computing devices such as that illustrated in FIG. 10 may be used to implement the described functionality in various embodiments" teaches a computing device (computer) comprising processors for implementing the embodied method) comprising:
generating, by one or more processors and using a plurality of source tables for a prediction domain, a plurality of subdomain-specific graphs for the prediction domain, each comprising a respective plurality of graph nodes and a respective plurality of weighted edges between the respective plurality of graph nodes (Col. 3, lines 28-56: "The computing devices may include instructions that when executed on or across one or more processors cause the devices to obtain an indication of (a) a plurality of entity types associated with a problem domain and (b) a plurality of relationship types among the plurality of entity types … A source data set comprising records pertaining to the entity and relationship types may also be obtained in at least some embodiments. From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches one or more processors and a source data set (plurality of source tables) for a problem domain for generating a graph representation comprising subsets (subgraphs) of respective nodes (graph nodes) and edges (weighted edges) between the nodes for the problem to be solved (prediction domain));
generating, by the one or more processors and using the graph-based machine learning model, a composite graph embedding based on the plurality of subdomain-specific embeddings and a designated predictive task (Col. 3, line 66 - Col. 4, line 59: "Using the graph representation and the respective initial feature sets, a graph convolutional network (GCN) model may be trained to generate one or more types of predictions of interest in various embodiments. The types of predictions may also be specified by clients programmatically in some embodiments. Within the GCN model, a representation of a particular node at a particular hidden layer of the model may be based at least in part on aggregated representations of neighbor nodes of the particular node in various embodiments. The representations of the neighbor nodes may be aggregated at least across (a) a set of one-hop neighbors of the particular node with respect to a first relationship type and (b) a set of one-hop neighbors of the particular node with respect to a second relationship type in some embodiments. A learned embedding corresponding to the particular node, obtained from a final hidden layer of the GCN model may be provided as input to a prediction layer of the GCN model and used to obtain the desired types of predictions with respect to the particular node. Training may be conducted using mini-batches of the input data in some embodiments, with the size of the mini-batches and the specific loss function (e.g., cross-entropy loss) comprising hyper-parameters of the training procedure. A trained version of the GCN model may be stored, and used to generate a prediction with respect to one or more instances of the entity types represented in the graph … Within the GCN model, convolution operations transform and combine information from a given node's neighborhood, e.g., using information from one-hop neighbors at a time, to generate high-quality embeddings or representations of nodes for various types of problems. When multiple such convolutions are in effect stacked on top of each other, using respective layers of the model, information can be captured from relatively distant parts of the graph. In at least some embodiments, each node's convolutional module may have a different neural network architecture depending on the details of the neighborhood of the node; however, the modules may share the same set of parameters across all their nodes, thus making the parameter complexity of the GCN independent of graph size" teaches using the GCN model (graph-based machine learning model) to generate a final learned embedding for the graph nodes(composite graph embedding) based on aggregated embeddings of node neighborhoods (subgraph subdomain embeddings) and a designated prediction task); and
initiating, by the one or more processors, the performance of the designated predictive task based on the composite graph embedding (Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that the GCN model (includes composite graph embedding) is used for initiating a performance of a prediction (designated predictive task) for unlabeled nodes and/or edges).
Gandhi et al. does not appear to explicitly teach generating, by the one or more processors and using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs.
However, Luo et al. teaches generating, by the one or more processors and using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs ([0062]: "Some embodiments of the present invention may include one, or more, of the following operations, features, characteristics and/or advantages: (i) a specific transaction composition graph node embedding method to embed each node/artifact on the graph to a latent feature vector; (ii) can be described as a refinement to general Node2Vec graph embedding methodology; (iii) determines the direction or sequence of data flow, since transactions normally have a starting node(s) and an end node(s); (iv) instead of adopting random walk sampling in Node2Vec on a graph, graph traversal from the starting node to the ending node can be started (following the data flow) to convert the full directed cyclic graph to multiple directed acyclic subgraphs; (v) a skip-diagram model will be applied to each subgraph to embed the artifact into a latent feature vector" teaches a skip-diagram model (graph-based machine learning model) for generating embeddings (subdomain-specific embeddings) for each subgraph (subdomain-specific graphs)).
Gandhi et al. and Luo et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate generating, by the one or more processors and using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs as taught by Luo et al. to the disclosed invention of Gandhi et al.
One of ordinary skill in the art would have been motivated to make this modification to provide "a specific transaction composition graph node embedding method to embed each node/artifact on the graph to a latent feature vector" because "in order to perform numeric or predictive analysis (such as calculating the risk or importance of each artifact/node and be able to rank them or classify each artifact to assign certain labels, such as belong to a business rule or clustering/grouping related or similar artifacts) each artifact should be able to be presented or described as a numeric feature vector" (Luo et al. [0061]-[0062]).
Regarding Claim 6,
Gandhi et al. in view of Luo et al. teaches the computer-implemented method of claim 1.
In addition, Gandhi et al. further teaches wherein the designated predictive task is a machine learning classification task (Col. 16, lines 16-31: "When the client wishes to obtain a prediction, e.g., for a set of new nodes (or for nodes/edges which were already present in the graph but were unlabeled), a GeneratePrediction request 730 specifying the targeted nodes may be submitted. In response the MLS may execute the trained version of the model, obtain the generated predictions" teaches that the predictive task is a machine learning task for predicting unlabeled nodes and/or edges. Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that the prediction of unlabeled nodes and/or edges is a classification task) and
initiating the performance of the designated predictive task based on the composite graph embedding comprises: generating, using a machine learning classification model, a predictive classification for an unlabeled graph node associated with the plurality of subdomain-specific graphs (Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that a neural network for classification (machine learning classification model) may be used to generate a classification for an unlabeled graph node associated with the graph used for training (i.e. the subgraphs are part of the graph)).
Regarding Claim 7,
Gandhi et al. in view of Luo et al. teaches the computer-implemented method of claim 6.
In addition, Gandhi et al. further teaches wherein: a plurality of graph nodes for a subdomain-specific graph of the plurality of subdomain-specific graphs comprises a set of common graph nodes that are within each of the plurality of subdomain-specific graphs and a set of subdomain-specific graph nodes specific to the subdomain-specific graph (Col. 3, lines 44-56: "From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches a graph representation comprising subsets of labeled nodes and edges (set of nodes specific to subdomain-specific graphs). Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that the graph representation includes unlabeled nodes (common nodes) (e.g. not specific to the subdomain-specific graphs)), and
the set of common graph nodes comprises the unlabeled graph node (Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that the graph representation includes unlabeled nodes (common nodes) (e.g. not specific to the subdomain-specific graphs)).
Regarding Claim 8,
Gandhi et al. in view of Luo et al. teaches the computer-implemented method of claim 7.
In addition, Gandhi et al. further teaches further comprising: modifying each of the plurality of subdomain-specific graphs by assigning the predictive classification to the unlabeled graph node (Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches the graph is modified with labels from the predictive classification for the unlabeled graph nodes).
Regarding Claim 9,
Gandhi et al. in view of Luo et al. teaches the computer-implemented method of claim 1.
In addition, Gandhi et al. further teaches wherein each of the plurality of subdomain-specific graphs comprises a separate heterogeneous and undirected graph data structure (Col. 3, lines 42-56: "A source data set comprising records pertaining to the entity and relationship types may also be obtained in at least some embodiments. From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches a source data set (plurality of source tables) for a problem domain for generating a graph representation comprising subsets (subgraphs) of respective nodes (graph nodes) with entity types and edges (weighted edges) between the nodes for the problem to be solved (prediction domain). Fig. 4; Col. 11, lines 51-53: "In at least some cases, multiple relationship types (whether directional or not) may exist between entities of a given pair of entity types" teaches that the node entities may be undirected and with multiple relationship types (i.e. heterogeneous)).
Regarding Claim 10,
Gandhi et al. in view of Luo et al. teaches the computer-implemented method of claim 1.
In addition, Gandhi et al. further teaches wherein each of the plurality of source tables comprise respective subdomain data for a subdomain of the prediction domain and a subdomain-specific graph of the plurality of subdomain-specific graphs is generated based on subdomain data from a corresponding source table of the plurality of source tables (Col. 3, lines 28-56: "The computing devices may include instructions that when executed on or across one or more processors cause the devices to obtain an indication of (a) a plurality of entity types associated with a problem domain and (b) a plurality of relationship types among the plurality of entity types … A source data set comprising records pertaining to the entity and relationship types may also be obtained in at least some embodiments. From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches a source data set (plurality of source tables) for a problem domain for generating a graph representation comprising respective subsets (subgraphs) of respective nodes (graph nodes) and edges (weighted edges) between the nodes with specific labels depending on the problem to be solved (subsets have respective subdomain data for the problem domain)).
Regarding Claim 14,
Gandhi et al. teaches a computing system comprising memory and one or more processors communicatively coupled to the memory, the one or more processors (Fig. 1; Col. 2, lines 40-45: "The present disclosure relates to methods and apparatus for training and using graph convolutional network (GCN) models to generate predictions with respect to data sets representing instances of multiple entity types, with multiple types of relationships among the instances of the entity types" teaches an apparatus (system) for using graph based machine learning for generating predictions. Fig. 10; Col. 20, lines 51-53: "Portions or all of multiple computing devices such as that illustrated in FIG. 10 may be used to implement the described functionality in various embodiments" teaches a computing device (computing system) for implementing the embodied apparatus. Fig. 10; Col. 19, lines 12-15: "FIG. 10 illustrates such a general-purpose computing device 9000. In the illustrated embodiment, computing device 9000 includes one or more processors 9010 coupled to a system memory 9020" teaches the computing device (computing system) comprising memory and processors communicatively coupled) configured to:
generate, using a plurality of source tables for a prediction domain, a plurality of subdomain-specific graphs for the prediction domain, each comprising a respective plurality of graph nodes and a respective plurality of weighted edges between the respective plurality of graph nodes (Col. 3, lines 28-56: "The computing devices may include instructions that when executed on or across one or more processors cause the devices to obtain an indication of (a) a plurality of entity types associated with a problem domain and (b) a plurality of relationship types among the plurality of entity types … A source data set comprising records pertaining to the entity and relationship types may also be obtained in at least some embodiments. From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches one or more processors and a source data set (plurality of source tables) for a problem domain for generating a graph representation comprising subsets (subgraphs) of respective nodes (graph nodes) and edges (weighted edges) between the nodes for the problem to be solved (prediction domain));
generate, using the graph-based machine learning model, a composite graph embedding based on the plurality of subdomain-specific embeddings and a designated predictive task (Col. 3, line 66 - Col. 4, line 59: "Using the graph representation and the respective initial feature sets, a graph convolutional network (GCN) model may be trained to generate one or more types of predictions of interest in various embodiments. The types of predictions may also be specified by clients programmatically in some embodiments. Within the GCN model, a representation of a particular node at a particular hidden layer of the model may be based at least in part on aggregated representations of neighbor nodes of the particular node in various embodiments. The representations of the neighbor nodes may be aggregated at least across (a) a set of one-hop neighbors of the particular node with respect to a first relationship type and (b) a set of one-hop neighbors of the particular node with respect to a second relationship type in some embodiments. A learned embedding corresponding to the particular node, obtained from a final hidden layer of the GCN model may be provided as input to a prediction layer of the GCN model and used to obtain the desired types of predictions with respect to the particular node. Training may be conducted using mini-batches of the input data in some embodiments, with the size of the mini-batches and the specific loss function (e.g., cross-entropy loss) comprising hyper-parameters of the training procedure. A trained version of the GCN model may be stored, and used to generate a prediction with respect to one or more instances of the entity types represented in the graph … Within the GCN model, convolution operations transform and combine information from a given node's neighborhood, e.g., using information from one-hop neighbors at a time, to generate high-quality embeddings or representations of nodes for various types of problems. When multiple such convolutions are in effect stacked on top of each other, using respective layers of the model, information can be captured from relatively distant parts of the graph. In at least some embodiments, each node's convolutional module may have a different neural network architecture depending on the details of the neighborhood of the node; however, the modules may share the same set of parameters across all their nodes, thus making the parameter complexity of the GCN independent of graph size" teaches using the GCN model (graph-based machine learning model) to generate a final learned embedding for the graph nodes(composite graph embedding) based on aggregated embeddings of node neighborhoods (subgraph subdomain embeddings) and a designated prediction task); and
initiate the performance of the designated predictive task based on the composite graph embedding (Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that the GCN model (includes composite graph embedding) is used for initiating a performance of a prediction (designated predictive task) for unlabeled nodes and/or edges).
Gandhi et al. does not appear to explicitly teach generate, using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs.
However, Luo et al. teaches generate, using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs ([0062]: "Some embodiments of the present invention may include one, or more, of the following operations, features, characteristics and/or advantages: (i) a specific transaction composition graph node embedding method to embed each node/artifact on the graph to a latent feature vector; (ii) can be described as a refinement to general Node2Vec graph embedding methodology; (iii) determines the direction or sequence of data flow, since transactions normally have a starting node(s) and an end node(s); (iv) instead of adopting random walk sampling in Node2Vec on a graph, graph traversal from the starting node to the ending node can be started (following the data flow) to convert the full directed cyclic graph to multiple directed acyclic subgraphs; (v) a skip-diagram model will be applied to each subgraph to embed the artifact into a latent feature vector" teaches a skip-diagram model (graph-based machine learning model) for generating embeddings (subdomain-specific embeddings) for each subgraph (subdomain-specific graphs)).
Gandhi et al. and Luo et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate generate, using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs as taught by Luo et al. to the disclosed invention of Gandhi et al.
One of ordinary skill in the art would have been motivated to make this modification to provide "a specific transaction composition graph node embedding method to embed each node/artifact on the graph to a latent feature vector" because "in order to perform numeric or predictive analysis (such as calculating the risk or importance of each artifact/node and be able to rank them or classify each artifact to assign certain labels, such as belong to a business rule or clustering/grouping related or similar artifacts) each artifact should be able to be presented or described as a numeric feature vector" (Luo et al. [0061]-[0062]).
Regarding Claim 19,
Gandhi et al. teaches one or more non-transitory computer-readable storage media including instructions that, when executed by one or more processors, cause the one or more processors (Fig. 1; Col. 2, lines 40-45: "The present disclosure relates to methods and apparatus for training and using graph convolutional network (GCN) models to generate predictions with respect to data sets representing instances of multiple entity types, with multiple types of relationships among the instances of the entity types" teaches a method for using graph based machine learning for generating predictions. Fig. 10; Col. 20, lines 32-46: "a computer-accessible medium may include non-transitory storage media or memory media such as magnetic or optical media, e.g., disk or DVD/CD coupled to computing device 9000 via I/O interface 9030. A non-transitory computer-accessible storage medium may also include any volatile or non-volatile media such as RAM (e.g. SDRAM, DDR SDRAM, RDRAM, SRAM, etc.), ROM, etc., that may be included in some embodiments of computing device 9000 as system memory 9020 or another type of memory. In some embodiments, a plurality of non-transitory computer-readable storage media may collectively store program instructions that when executed on or across one or more processors implement at least a subset of the methods and techniques described above" teaches a non-transitory computer-accessible medium (non-transitory computer-readable storage media) comprising instructions for execution on one or more processors for implementing the embodied method) to:
generate, using a plurality of source tables for a prediction domain, a plurality of subdomain-specific graphs for the prediction domain, each comprising a respective plurality of graph nodes and a respective plurality of weighted edges between the respective plurality of graph nodes (Col. 3, lines 28-56: "The computing devices may include instructions that when executed on or across one or more processors cause the devices to obtain an indication of (a) a plurality of entity types associated with a problem domain and (b) a plurality of relationship types among the plurality of entity types … A source data set comprising records pertaining to the entity and relationship types may also be obtained in at least some embodiments. From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches one or more processors and a source data set (plurality of source tables) for a problem domain for generating a graph representation comprising subsets (subgraphs) of respective nodes (graph nodes) and edges (weighted edges) between the nodes for the problem to be solved (prediction domain));
generate, using the graph-based machine learning model, a composite graph embedding based on the plurality of subdomain-specific embeddings and a designated predictive task (Col. 3, line 66 - Col. 4, line 59: "Using the graph representation and the respective initial feature sets, a graph convolutional network (GCN) model may be trained to generate one or more types of predictions of interest in various embodiments. The types of predictions may also be specified by clients programmatically in some embodiments. Within the GCN model, a representation of a particular node at a particular hidden layer of the model may be based at least in part on aggregated representations of neighbor nodes of the particular node in various embodiments. The representations of the neighbor nodes may be aggregated at least across (a) a set of one-hop neighbors of the particular node with respect to a first relationship type and (b) a set of one-hop neighbors of the particular node with respect to a second relationship type in some embodiments. A learned embedding corresponding to the particular node, obtained from a final hidden layer of the GCN model may be provided as input to a prediction layer of the GCN model and used to obtain the desired types of predictions with respect to the particular node. Training may be conducted using mini-batches of the input data in some embodiments, with the size of the mini-batches and the specific loss function (e.g., cross-entropy loss) comprising hyper-parameters of the training procedure. A trained version of the GCN model may be stored, and used to generate a prediction with respect to one or more instances of the entity types represented in the graph … Within the GCN model, convolution operations transform and combine information from a given node's neighborhood, e.g., using information from one-hop neighbors at a time, to generate high-quality embeddings or representations of nodes for various types of problems. When multiple such convolutions are in effect stacked on top of each other, using respective layers of the model, information can be captured from relatively distant parts of the graph. In at least some embodiments, each node's convolutional module may have a different neural network architecture depending on the details of the neighborhood of the node; however, the modules may share the same set of parameters across all their nodes, thus making the parameter complexity of the GCN independent of graph size" teaches using the GCN model (graph-based machine learning model) to generate a final learned embedding for the graph nodes(composite graph embedding) based on aggregated embeddings of node neighborhoods (subgraph subdomain embeddings) and a designated prediction task); and
initiate the performance of the designated predictive task based on the composite graph embedding (Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that the GCN model (includes composite graph embedding) is used for initiating a performance of a prediction (designated predictive task) for unlabeled nodes and/or edges).
Gandhi et al. does not appear to explicitly teach generate, using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs.
However, Luo et al. teaches generate, using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs ([0062]: "Some embodiments of the present invention may include one, or more, of the following operations, features, characteristics and/or advantages: (i) a specific transaction composition graph node embedding method to embed each node/artifact on the graph to a latent feature vector; (ii) can be described as a refinement to general Node2Vec graph embedding methodology; (iii) determines the direction or sequence of data flow, since transactions normally have a starting node(s) and an end node(s); (iv) instead of adopting random walk sampling in Node2Vec on a graph, graph traversal from the starting node to the ending node can be started (following the data flow) to convert the full directed cyclic graph to multiple directed acyclic subgraphs; (v) a skip-diagram model will be applied to each subgraph to embed the artifact into a latent feature vector" teaches a skip-diagram model (graph-based machine learning model) for generating embeddings (subdomain-specific embeddings) for each subgraph (subdomain-specific graphs)).
Gandhi et al. and Luo et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate generate, using a graph-based machine learning model, a plurality of subdomain-specific embeddings comprising a respective subdomain-specific embedding for each of the plurality of subdomain-specific graphs as taught by Luo et al. to the disclosed invention of Gandhi et al.
One of ordinary skill in the art would have been motivated to make this modification to provide "a specific transaction composition graph node embedding method to embed each node/artifact on the graph to a latent feature vector" because "in order to perform numeric or predictive analysis (such as calculating the risk or importance of each artifact/node and be able to rank them or classify each artifact to assign certain labels, such as belong to a business rule or clustering/grouping related or similar artifacts) each artifact should be able to be presented or described as a numeric feature vector" (Luo et al. [0061]-[0062]).
Regarding Claim 20,
Gandhi et al. in view of Luo et al. teaches the one or more non-transitory computer-readable storage media of claim 19.
In addition, Gandhi et al. further teaches wherein the designated predictive task is a machine learning classification task (Col. 16, lines 16-31: "When the client wishes to obtain a prediction, e.g., for a set of new nodes (or for nodes/edges which were already present in the graph but were unlabeled), a GeneratePrediction request 730 specifying the targeted nodes may be submitted. In response the MLS may execute the trained version of the model, obtain the generated predictions" teaches that the predictive task is a machine learning task for predicting unlabeled nodes and/or edges. Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that the prediction of unlabeled nodes and/or edges is a classification task) and
initiating the performance of the designated predictive task based on the composite graph embedding comprises: generating, using a machine learning classification model, a predictive classification for an unlabeled graph node associated with the plurality of subdomain-specific graphs (Col. 13, lines 50-66: "As shown in the Classify procedure of the pseudocode, the predictions generated by the GCN model comprise class labels for nodes and/or edges which are currently unlabeled, using the final embeddings zv generated at the Kth layer of the model and weight matrices Y for the prediction layer of the model (which is not shown in FIG. 5). The logic of the Classify procedure may be implemented using a filly connected neural network layer in at least some embodiments, with the number of neurons in that layer dependent on the kind of problem which is being addressed (e.g., for classification among C classes, the fully-connected layer may comprise C neurons). Note that in some cases, the model may be used to generate labels for unlabeled nodes/edges which were already part of the graph used in the training of the model; in other cases, the model may be used to generate labels for new nodes and/or predict new labeled edges (which were not part of the original graph)" teaches that a neural network for classification (machine learning classification model) may be used to generate a classification for an unlabeled graph node associated with the graph used for training (i.e. the subgraphs are part of the graph)).
Claims 2-5 and 15-18 are rejected under 35 U.S.C. 103 as being unpatentable over Gandhi et al. (US 11,593,622 B1) in view of Luo et al. (US 2022/0180240 A1) and further in view of Creed et al. (US 2021/0081717 A1).
Regarding Claim 2,
Gandhi et al. in view of Luo et al. teaches the computer-implemented method of claim 1.
Gandhi et al. in view of Luo et al. does not appear to explicitly teach wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding.
However, Creed et al. teaches wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding ([0087]: "The encoding network 104 of the GNN model is configured to generate an embedding that includes an attention weight assigned to each relationship edge of at least the portion of the entity-entity graph that is ingested. The attention weights are configured to indicate the relevancy of each corresponding relationship edge between entity nodes of the entity-entity graph and assist the GNN model in relationship prediction. The attention weights are learned by the GNN model during training on the entity-entity graph dataset" teaches generating embeddings for portions (subgraphs) of the entity-entity graph based on a plurality of attention weights for the entities (nodes) and relationship edges (weighted edges)).
Gandhi et al., Luo et al., and Creed et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding as taught by Creed et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "efficiently increase the robustness of the GCNN technique to generate an accurate GNN model for prediction/classification problems in light of large noisy graph datasets during training and in decision directed mode (test or prediction mode)" (Creed et al. [0082]).
Regarding Claim 3,
Gandhi et al. in view of Luo et al. and further in view of Creed et al. teaches the computer-implemented method of claim 2.
In addition, Creed et al. further teaches wherein the plurality of attention weights comprises a plurality of node-level weights (Fig. 2; [0123]: "Each attention weight αr,i,j is a single scalar that adjusts the amount of that input representation to add to the representation of the i-th entity node ei 202a2. For example, when αr,i,j 206 is large, then i-th entity node ei 202a2 will use more of the input from the j-th entity node e1 202b1. When αr,i,j 206 is small, then i-th entity node ei 202a2 will use less of the input from the j-th entity node e1 202b1" teaches the attention weights comprising node level weights) and the computer-implemented method further comprises:
generating, using the graph-based machine learning model, the plurality of node-level weights for the plurality of graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the plurality of graph nodes (Fig. 2; [0123]: "Each attention weight αr,i,j is a single scalar that adjusts the amount of that input representation to add to the representation of the i-th entity node ei 202a2. For example, when αr,i,j 206 is large, then i-th entity node ei 202a2 will use more of the input from the j-th entity node e1 202b1. When αr,i,j 206 is small, then i-th entity node ei 202a2 will use less of the input from the j-th entity node e1 202b1. The GCNN technique 200 is configured to minimize the GCNN loss function based on the embeddings and the relationship scores and update the weights of the encoder network including the attention weights αr,i,j for each edge of the graph and to update the weights of the scoring network (or decoder network). The attention weights αr,i,j (a.k.a. scaling factors) of the graph are learned through training GNN model based on the GCNN technique 200 and data representative of the input entity-entity knowledge graph as the training graph dataset" teaches the GCNN model (graph-based machine learning model) being used to generate attention weights for nodes (node-level weights) for the graph nodes based on data representative of the input entity-entity knowledge graph (node attributes of the entities in the graph)).
Gandhi et al., Luo et al., and Creed et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate wherein the plurality of attention weights comprises a plurality of node-level weights and the computer-implemented method further comprises: generating, using the graph-based machine learning model, the plurality of node-level weights for the plurality of graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the plurality of graph nodes as taught by Creed et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "efficiently increase the robustness of the GCNN technique to generate an accurate GNN model for prediction/classification problems in light of large noisy graph datasets during training and in decision directed mode (test or prediction mode)" (Creed et al. [0082]).
Regarding Claim 4,
Gandhi et al. in view of Luo et al. and further in view of Creed et al. teaches the computer-implemented method of claim 3.
In addition, Gandhi et al. further teaches wherein the plurality of node attributes comprises one or more node labels for the designated predictive task (Col. 3, lines 44-56: "From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches a graph representation comprising subsets of labeled nodes and edges (node attributes comprise node labels) for the problem to be solved (designated predictive task)) and the computer-implemented method further comprises:
generating, using a semi-supervised loss function, a model loss for the graph-based machine learning model based on the composite graph embedding (Col. 3, line 66 - Col. 4, line 59: "Using the graph representation and the respective initial feature sets, a graph convolutional network (GCN) model may be trained to generate one or more types of predictions of interest in various embodiments. The types of predictions may also be specified by clients programmatically in some embodiments. Within the GCN model, a representation of a particular node at a particular hidden layer of the model may be based at least in part on aggregated representations of neighbor nodes of the particular node in various embodiments. The representations of the neighbor nodes may be aggregated at least across (a) a set of one-hop neighbors of the particular node with respect to a first relationship type and (b) a set of one-hop neighbors of the particular node with respect to a second relationship type in some embodiments. A learned embedding corresponding to the particular node, obtained from a final hidden layer of the GCN model may be provided as input to a prediction layer of the GCN model and used to obtain the desired types of predictions with respect to the particular node. Training may be conducted using mini-batches of the input data in some embodiments, with the size of the mini-batches and the specific loss function (e.g., cross-entropy loss) comprising hyper-parameters of the training procedure. A trained version of the GCN model may be stored, and used to generate a prediction with respect to one or more instances of the entity types represented in the graph … Within the GCN model, convolution operations transform and combine information from a given node's neighborhood, e.g., using information from one-hop neighbors at a time, to generate high-quality embeddings or representations of nodes for various types of problems. When multiple such convolutions are in effect stacked on top of each other, using respective layers of the model, information can be captured from relatively distant parts of the graph. In at least some embodiments, each node's convolutional module may have a different neural network architecture depending on the details of the neighborhood of the node; however, the modules may share the same set of parameters across all their nodes, thus making the parameter complexity of the GCN independent of graph size" teaches a loss function being used to generate a model loss for the GCN model (graph-based machine learning model) and generated final learned embedding for the graph nodes (composite graph embedding)); and
updating, using a machine learning training technique, the composite graph embedding based on the model loss (Col. 3, line 66 - Col. 4, line 59: "Using the graph representation and the respective initial feature sets, a graph convolutional network (GCN) model may be trained to generate one or more types of predictions of interest in various embodiments. The types of predictions may also be specified by clients programmatically in some embodiments. Within the GCN model, a representation of a particular node at a particular hidden layer of the model may be based at least in part on aggregated representations of neighbor nodes of the particular node in various embodiments. The representations of the neighbor nodes may be aggregated at least across (a) a set of one-hop neighbors of the particular node with respect to a first relationship type and (b) a set of one-hop neighbors of the particular node with respect to a second relationship type in some embodiments. A learned embedding corresponding to the particular node, obtained from a final hidden layer of the GCN model may be provided as input to a prediction layer of the GCN model and used to obtain the desired types of predictions with respect to the particular node. Training may be conducted using mini-batches of the input data in some embodiments, with the size of the mini-batches and the specific loss function (e.g., cross-entropy loss) comprising hyper-parameters of the training procedure. A trained version of the GCN model may be stored, and used to generate a prediction with respect to one or more instances of the entity types represented in the graph … Within the GCN model, convolution operations transform and combine information from a given node's neighborhood, e.g., using information from one-hop neighbors at a time, to generate high-quality embeddings or representations of nodes for various types of problems. When multiple such convolutions are in effect stacked on top of each other, using respective layers of the model, information can be captured from relatively distant parts of the graph. In at least some embodiments, each node's convolutional module may have a different neural network architecture depending on the details of the neighborhood of the node; however, the modules may share the same set of parameters across all their nodes, thus making the parameter complexity of the GCN independent of graph size" teaches a loss function being used to generate a model loss for training (machine learning training technique) to update the GCN model (graph-based machine learning model) and generated final learned embedding for the graph nodes (composite graph embedding)).
Regarding Claim 5,
Gandhi et al. in view of Luo et al. and further in view of Creed et al. teaches the computer-implemented method of claim 2.
In addition, Gandhi et al. further teaches wherein the plurality of attention weights comprises a plurality of semantic-level weights (Col. 11, lines 39-53: "In some cases, a logical relationship may have directional semantics, further increasing the number of different relationship types which may be represented in a graph used for a GCN … In at least some cases, multiple relationship types (whether directional or not) may exist between entities of a given pair of entity types" teaches that the relationships between entities (nodes) may have directional semantics (semantic-level weights)).
Furthermore, Creed et al. further teaches the computer-implemented method further comprises: generating, using the graph-based machine learning model, the plurality of semantic-level weights for the plurality of weighted edges of the subdomain-specific graph based on one or more metapaths within the subdomain-specific graph (Fig. 1d; [0114]: "The entity-relationship graph includes a plurality of entity nodes in which each entity node is connected to one or more entity node of the plurality of entity nodes by a relationship edge. The entity-entity relationship graph or a portion thereof may be used to train a GNN model in which an attention weight is assigned to each relationship edge of the entity-entity graph or portion thereof. The process 130 further including the steps of: In step 132, receiving a set of attention weights from the GNN model, where the GNN model has been trained based on data representative of the entity-entity graph or a portion thereof with an attention weight assigned to each relationship edge of the entity-entity graph or a portion thereof in which the attention weights are updated during training and are indicative of the relevancy of each corresponding relationship edge between entity nodes of the entity-entity graph or portion thereof. In step 134, generating a filtered entity-entity relationship graph by weighting each edge with the corresponding trained attention weight. Filtering the entity-entity graph based on the trained attention weights may include removing those relationship edge(s) from the entity-entity graph or portion thereof having a corresponding trained attention weight that is below or equal to an attention relevancy threshold. Alternatively, filtering the entity-entity graph may include removing those relationship edge(s) from the entity-entity graph or portion thereof having a corresponding trained attention weight that is above or equal to another attention relevancy threshold. In step 136, the filtered entity-entity graph may be output. Additionally or alternatively, step 134 may further include identifying uncertain relationship edges in the entity-entity graph or portion thereof based on a set of attention weights retrieved from the trained GNN model associated with the entity-entity graph and an attention relevancy threshold. The identified uncertain relationship edges may be culled, removed, and/or rules may be derived defining actions for excluding the relationship edges, where the rules for part of the data representative of the entity-entity graph" teaches generating semantic-level attention weights for the relationship edges (weighted edges) of the graph based on relationship edges meeting a threshold to form a path (metapath) through the graph).
Gandhi et al., Luo et al., and Creed et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the computer-implemented method further comprises: generating, using the graph-based machine learning model, the plurality of semantic-level weights for the plurality of weighted edges of the subdomain-specific graph based on one or more metapaths within the subdomain-specific graph as taught by Creed et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "efficiently increase the robustness of the GCNN technique to generate an accurate GNN model for prediction/classification problems in light of large noisy graph datasets during training and in decision directed mode (test or prediction mode)" (Creed et al. [0082]).
Regarding Claim 15,
Gandhi et al. in view of Luo et al. teaches the computing system of claim 14.
Gandhi et al. in view of Luo et al. does not appear to explicitly teach wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding.
However, Creed et al. teaches wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding ([0087]: "The encoding network 104 of the GNN model is configured to generate an embedding that includes an attention weight assigned to each relationship edge of at least the portion of the entity-entity graph that is ingested. The attention weights are configured to indicate the relevancy of each corresponding relationship edge between entity nodes of the entity-entity graph and assist the GNN model in relationship prediction. The attention weights are learned by the GNN model during training on the entity-entity graph dataset" teaches generating embeddings for portions (subgraphs) of the entity-entity graph based on a plurality of attention weights for the entities (nodes) and relationship edges (weighted edges)).
Gandhi et al., Luo et al., and Creed et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate wherein a subdomain-specific embedding of the plurality of subdomain-specific embeddings is based on a plurality of attention weights assigned to a plurality of graph nodes and a plurality of weighted edges of a subdomain-specific graph corresponding to the subdomain-specific embedding as taught by Creed et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "efficiently increase the robustness of the GCNN technique to generate an accurate GNN model for prediction/classification problems in light of large noisy graph datasets during training and in decision directed mode (test or prediction mode)" (Creed et al. [0082]).
Regarding Claim 16,
Gandhi et al. in view of Luo et al. and further in view of Creed et al. teaches the computing system of claim 15.
In addition, Creed et al. further teaches wherein the plurality of attention weights comprises a plurality of node-level weights (Fig. 2; [0123]: "Each attention weight αr,i,j is a single scalar that adjusts the amount of that input representation to add to the representation of the i-th entity node ei 202a2. For example, when αr,i,j 206 is large, then i-th entity node ei 202a2 will use more of the input from the j-th entity node e1 202b1. When αr,i,j 206 is small, then i-th entity node ei 202a2 will use less of the input from the j-th entity node e1 202b1" teaches the attention weights comprising node level weights) and
the one or more processors are further configured to: generate, using the graph-based machine learning model, the plurality of node-level weights for the plurality of graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the plurality of graph nodes (Fig. 2; [0123]: "Each attention weight αr,i,j is a single scalar that adjusts the amount of that input representation to add to the representation of the i-th entity node ei 202a2. For example, when αr,i,j 206 is large, then i-th entity node ei 202a2 will use more of the input from the j-th entity node e1 202b1. When αr,i,j 206 is small, then i-th entity node ei 202a2 will use less of the input from the j-th entity node e1 202b1. The GCNN technique 200 is configured to minimize the GCNN loss function based on the embeddings and the relationship scores and update the weights of the encoder network including the attention weights αr,i,j for each edge of the graph and to update the weights of the scoring network (or decoder network). The attention weights αr,i,j (a.k.a. scaling factors) of the graph are learned through training GNN model based on the GCNN technique 200 and data representative of the input entity-entity knowledge graph as the training graph dataset" teaches the GCNN model (graph-based machine learning model) being used to generate attention weights for nodes (node-level weights) for the graph nodes based on data representative of the input entity-entity knowledge graph (node attributes of the entities in the graph)).
Gandhi et al., Luo et al., and Creed et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate wherein the plurality of attention weights comprises a plurality of node-level weights and the one or more processors are further configured to: generate, using the graph-based machine learning model, the plurality of node-level weights for the plurality of graph nodes of the subdomain-specific graph based on a plurality of node attributes corresponding to the plurality of graph nodes as taught by Creed et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "efficiently increase the robustness of the GCNN technique to generate an accurate GNN model for prediction/classification problems in light of large noisy graph datasets during training and in decision directed mode (test or prediction mode)" (Creed et al. [0082]).
Regarding Claim 17,
Gandhi et al. in view of Luo et al. and further in view of Creed et al. teaches the computing system of claim 16.
In addition, Gandhi et al. further teaches wherein the plurality of node attributes comprises one or more node labels for the designated predictive task (Col. 3, lines 44-56: "From the source data set, a graph representation comprising a plurality of nodes and a plurality of edges may be generated. The nodes may represent respective instance of an entity type, and an edge between a pair of nodes may indicate a relationship of a particular relationship type between the respective instances represented by the pair of nodes. In various embodiments, at least one node of the graph may be linked to a plurality of other nodes via a respective edge, indicating a plurality of relationships of the node. In various embodiments, the source data set may comprise labels for a subset of the nodes and/or edges, with the specific labels being dependent on the kind of problem which is to be solved" teaches a graph representation comprising subsets of labeled nodes and edges (node attributes comprise node labels) for the problem to be solved (designated predictive task)) and
the one or more processors are further configured to: generate, using a semi-supervised loss function, a model loss for the graph-based machine learning model based on the composite graph embedding (Col. 3, line 66 - Col. 4, line 59: "Using the graph representation and the respective initial feature sets, a graph convolutional network (GCN) model may be trained to generate one or more types of predictions of interest in various embodiments. The types of predictions may also be specified by clients programmatically in some embodiments. Within the GCN model, a representation of a particular node at a particular hidden layer of the model may be based at least in part on aggregated representations of neighbor nodes of the particular node in various embodiments. The representations of the neighbor nodes may be aggregated at least across (a) a set of one-hop neighbors of the particular node with respect to a first relationship type and (b) a set of one-hop neighbors of the particular node with respect to a second relationship type in some embodiments. A learned embedding corresponding to the particular node, obtained from a final hidden layer of the GCN model may be provided as input to a prediction layer of the GCN model and used to obtain the desired types of predictions with respect to the particular node. Training may be conducted using mini-batches of the input data in some embodiments, with the size of the mini-batches and the specific loss function (e.g., cross-entropy loss) comprising hyper-parameters of the training procedure. A trained version of the GCN model may be stored, and used to generate a prediction with respect to one or more instances of the entity types represented in the graph … Within the GCN model, convolution operations transform and combine information from a given node's neighborhood, e.g., using information from one-hop neighbors at a time, to generate high-quality embeddings or representations of nodes for various types of problems. When multiple such convolutions are in effect stacked on top of each other, using respective layers of the model, information can be captured from relatively distant parts of the graph. In at least some embodiments, each node's convolutional module may have a different neural network architecture depending on the details of the neighborhood of the node; however, the modules may share the same set of parameters across all their nodes, thus making the parameter complexity of the GCN independent of graph size" teaches a loss function being used to generate a model loss for the GCN model (graph-based machine learning model) and generated final learned embedding for the graph nodes (composite graph embedding)); and
update, using a machine learning training technique, the composite graph embedding based on the model loss (Col. 3, line 66 - Col. 4, line 59: "Using the graph representation and the respective initial feature sets, a graph convolutional network (GCN) model may be trained to generate one or more types of predictions of interest in various embodiments. The types of predictions may also be specified by clients programmatically in some embodiments. Within the GCN model, a representation of a particular node at a particular hidden layer of the model may be based at least in part on aggregated representations of neighbor nodes of the particular node in various embodiments. The representations of the neighbor nodes may be aggregated at least across (a) a set of one-hop neighbors of the particular node with respect to a first relationship type and (b) a set of one-hop neighbors of the particular node with respect to a second relationship type in some embodiments. A learned embedding corresponding to the particular node, obtained from a final hidden layer of the GCN model may be provided as input to a prediction layer of the GCN model and used to obtain the desired types of predictions with respect to the particular node. Training may be conducted using mini-batches of the input data in some embodiments, with the size of the mini-batches and the specific loss function (e.g., cross-entropy loss) comprising hyper-parameters of the training procedure. A trained version of the GCN model may be stored, and used to generate a prediction with respect to one or more instances of the entity types represented in the graph … Within the GCN model, convolution operations transform and combine information from a given node's neighborhood, e.g., using information from one-hop neighbors at a time, to generate high-quality embeddings or representations of nodes for various types of problems. When multiple such convolutions are in effect stacked on top of each other, using respective layers of the model, information can be captured from relatively distant parts of the graph. In at least some embodiments, each node's convolutional module may have a different neural network architecture depending on the details of the neighborhood of the node; however, the modules may share the same set of parameters across all their nodes, thus making the parameter complexity of the GCN independent of graph size" teaches a loss function being used to generate a model loss for training (machine learning training technique) to update the GCN model (graph-based machine learning model) and generated final learned embedding for the graph nodes (composite graph embedding)).
Regarding Claim 18,
Gandhi et al. in view of Luo et al. and further in view of Creed et al. teaches the computing system of claim 15.
In addition, Gandhi et al. further teaches wherein the plurality of attention weights comprises a plurality of semantic-level weights (Col. 11, lines 39-53: "In some cases, a logical relationship may have directional semantics, further increasing the number of different relationship types which may be represented in a graph used for a GCN … In at least some cases, multiple relationship types (whether directional or not) may exist between entities of a given pair of entity types" teaches that the relationships between entities (nodes) may have directional semantics (semantic-level weights)).
Furthermore, Creed et al. further teaches the one or more processors are further configured to: generate, using the graph-based machine learning model, the plurality of semantic-level weights for the plurality of weighted edges of the subdomain-specific graph based on one or more metapaths within the subdomain-specific graph (Fig. 1d; [0114]: "The entity-relationship graph includes a plurality of entity nodes in which each entity node is connected to one or more entity node of the plurality of entity nodes by a relationship edge. The entity-entity relationship graph or a portion thereof may be used to train a GNN model in which an attention weight is assigned to each relationship edge of the entity-entity graph or portion thereof. The process 130 further including the steps of: In step 132, receiving a set of attention weights from the GNN model, where the GNN model has been trained based on data representative of the entity-entity graph or a portion thereof with an attention weight assigned to each relationship edge of the entity-entity graph or a portion thereof in which the attention weights are updated during training and are indicative of the relevancy of each corresponding relationship edge between entity nodes of the entity-entity graph or portion thereof. In step 134, generating a filtered entity-entity relationship graph by weighting each edge with the corresponding trained attention weight. Filtering the entity-entity graph based on the trained attention weights may include removing those relationship edge(s) from the entity-entity graph or portion thereof having a corresponding trained attention weight that is below or equal to an attention relevancy threshold. Alternatively, filtering the entity-entity graph may include removing those relationship edge(s) from the entity-entity graph or portion thereof having a corresponding trained attention weight that is above or equal to another attention relevancy threshold. In step 136, the filtered entity-entity graph may be output. Additionally or alternatively, step 134 may further include identifying uncertain relationship edges in the entity-entity graph or portion thereof based on a set of attention weights retrieved from the trained GNN model associated with the entity-entity graph and an attention relevancy threshold. The identified uncertain relationship edges may be culled, removed, and/or rules may be derived defining actions for excluding the relationship edges, where the rules for part of the data representative of the entity-entity graph" teaches generating semantic-level attention weights for the relationship edges (weighted edges) of the graph based on relationship edges meeting a threshold to form a path (metapath) through the graph).
Gandhi et al., Luo et al., and Creed et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the one or more processors are further configured to: generate, using the graph-based machine learning model, the plurality of semantic-level weights for the plurality of weighted edges of the subdomain-specific graph based on one or more metapaths within the subdomain-specific graph as taught by Creed et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "efficiently increase the robustness of the GCNN technique to generate an accurate GNN model for prediction/classification problems in light of large noisy graph datasets during training and in decision directed mode (test or prediction mode)" (Creed et al. [0082]).
Claims 11 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Gandhi et al. (US 11,593,622 B1) in view of Luo et al. (US 2022/0180240 A1) and further in view of Zhang et al. (US 2021/0125344 A1).
Regarding Claim 11,
Gandhi et al. in view of Luo et al. teaches the computer-implemented method of claim 10.
Gandhi et al. in view of Luo et al. does not appear to explicitly teach further comprising: receiving one or more modification data objects associated with the corresponding source table; and in response to the one or more modification data objects, regenerating the subdomain-specific graph.
However, Zhang et al. teaches further comprising: receiving one or more modification data objects associated with the corresponding source table (Fig. 2; [0038]-[0039]: "To help clarify terminology, the subset of nodes used to modify the appearance embedding vector (e0i) of another node is herein referred to as a “modification node set.” Thus, in the illustrated example, the node being modified corresponding to detection F is modified by a modification node set including nodes corresponding to detections D1, D2, D3, D4, and D5 (e.g., collectively—nodes corresponding to the subjective past relative to the node being modified, D7) … During actual post-training instances of multi-object tracking, the graph network 212 modifies of the appearance embedding vector (e0i) for each different detection based on the learned function G and also based on the edge weights connecting the node being modified to each node in its associated modification node set. In general, the modification node set for each update includes nodes with detection times corresponding to a time interval that excludes the detection time of the node being modified" teaches receiving modification node sets (modification data objects) associated with the detector sources (source tables)); and
in response to the one or more modification data objects, regenerating the subdomain-specific graph (Fig. 2; Fig. 3A; Fig. 3B; [0044]: "FIGS. 3A-3B detail an exemplary sequence of operations that a graph network, such as the graph network 212 of FIG. 2, that may be performed to update the appearance embedding vector corresponding to each individual detection of a scene. In this example, the graph network performs three iterations of node updates. In the first iteration described with respect to FIG. 3A, a first subset of the graph nodes are individually updated based on a first modification node set; in the second iteration described with respect to FIG. 3B, a second subset of the graph nodes are updated based on a second modification node set; and in a third iteration also described with respect to FIG. 3B, a third and final subset of the graph nodes is updated based on a third modification node set" teaches updating subsets of graph nodes (subdomain graphs) based on the modification node sets (modification data objects)).
Gandhi et al., Luo et al., and Zhang et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate further comprising: receiving one or more modification data objects associated with the corresponding source table; and in response to the one or more modification data objects, regenerating the subdomain-specific graph as taught by Zhang et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "reduce the computational complexity needed to simultaneously track multiple objects by eliminating the need to track objects independent of one another and instead, using the inter-relations between detections to enhance the distinguishing characteristics of each individual detection" (Zhang et al. [0013]).
Regarding Claim 12,
Gandhi et al. in view of Luo et al. and further in view of Zhang et al. teaches the computer-implemented method of claim 11.
In addition, Zhang et al. teaches wherein the one or more modification data objects are received at a defined time interval (Fig. 2; [0038]-[0039]: "To help clarify terminology, the subset of nodes used to modify the appearance embedding vector (e0i) of another node is herein referred to as a “modification node set.” Thus, in the illustrated example, the node being modified corresponding to detection F is modified by a modification node set including nodes corresponding to detections D1, D2, D3, D4, and D5 (e.g., collectively—nodes corresponding to the subjective past relative to the node being modified, D7) … During actual post-training instances of multi-object tracking, the graph network 212 modifies of the appearance embedding vector (e0i) for each different detection based on the learned function G and also based on the edge weights connecting the node being modified to each node in its associated modification node set. In general, the modification node set for each update includes nodes with detection times corresponding to a time interval that excludes the detection time of the node being modified" teaches that the modification node sets (modification data objects) are received at a corresponding time interval).
Gandhi et al., Luo et al., and Zhang et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate wherein the one or more modification data objects are received at a defined time interval as taught by Zhang et al. to the disclosed invention of Gandhi et al. in view of Luo et al.
One of ordinary skill in the art would have been motivated to make this modification to "reduce the computational complexity needed to simultaneously track multiple objects by eliminating the need to track objects independent of one another and instead, using the inter-relations between detections to enhance the distinguishing characteristics of each individual detection" (Zhang et al. [0013]).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Gandhi et al. (US 11,593,622 B1) in view of Luo et al. (US 2022/0180240 A1) in view of Zhang et al. (US 2021/0125344 A1) and further in view of Morato et al. (US 2024/0256857 A1).
Regarding Claim 13,
Gandhi et al. in view of Luo et al. and further in view of Zhang et al. teaches the computer-implemented method of claim 12.
Gandhi et al. in view of Luo et al. and further in view of Zhang et al. does not appear to explicitly teach wherein the prediction domain comprises a clinical domain and the defined time interval is associated with a claim aggregation frequency.
However, Morato et al. teaches wherein the prediction domain comprises a clinical domain and the defined time interval is associated with a claim aggregation frequency ([0094]: "When applied in a clinical prediction domain, the multi-headed composite model may accurately and efficiently transform data recorded with a wide variety of robust, dynamically changing coding sets to form a universal ontology of information aggregated across a plurality of disparate, incompatible, third-party datasets" teaches that the prediction domain may comprise a clinical domain with dynamically changing coding sets (i.e. is modified over time). [0067]-[0068]: "the term “third party” refers to an entity that is associated with a third-party coding set. The third party may be configured to generate, maintain, store, and/or the like, data that is defined by a third-party coding set ... The third party may be based on the prediction domain. In some examples, the third party may include a clinical provider that is configured to generate, maintain, store, and/or the like, medical data, such as medical claims, and/or the like, that is coded according to a medical, external, coding set" teaches that the dynamically changing coding sets for the clinical domain may include medical claims (i.e. the modification time interval is associated with claim frequency)).
Gandhi et al., Luo et al., and Zhang et al. are analogous to the claimed invention because they are directed towards graph-based machine learning.
Morato et al. is analogous to the claimed invention because it is directed towards machine learning classification for a clinical domain.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate wherein the prediction domain comprises a clinical domain and the defined time interval is associated with a claim aggregation frequency as taught by Morato et al. to the disclosed invention of Gandhi et al. in view of Luo et al. and further in view of Zhang et al.
One of ordinary skill in the art would have been motivated to make this modification to "provide improved performance, and reduced costs and latencies relative to traditional machine learning techniques" (Morato et al. [0196]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRIAN J HALES whose telephone number is (571)272-0878. The examiner can normally be reached M-F 9:00am - 5:00pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kamran Afshar can be reached at (571) 272-7796. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BRIAN J HALES/Examiner, Art Unit 2125
/KAMRAN AFSHAR/Supervisory Patent Examiner, Art Unit 2125
Prosecution Insights