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 .
Remarks
This Office Action is responsive to Applicants' Amendment filed on November 4, 2025, in which claims 1, 11, 15, and 18 are currently amended. Claims 1-20 are currently pending.
Response to Arguments
The rejection to claim 11 under 35 U.S.C. § 112(b) is hereby withdrawn, as necessitated by applicant's amendments and remarks made to the rejections.
Applicant’s arguments with respect to rejection of claims 1-20 under 35 U.S.C. 101 based on amendment have been considered, however, are not persuasive.
With respect to Applicant's arguments on p. 9 of the Remarks submitted 11/4/2025 that the claims are integrated into a practical application, Examiner respectfully disagrees. The human mind is readily capable of data generation as well as graph model creation and evaluation which Applicant admits on p. 9 of the Remarks are what the claimed invention is directed towards. With respect to the model training, the instant specification explicitly ties the model training to mathematical calculations and relationships which is also a judicial exception ([¶0094] "various training or learning techniques, such as, for example, backwards propagation of errors. For example, a loss function can be backpropagated through the model(s) to update one or more parameters of the model(s) (e.g., based on a gradient of the loss function). Various loss functions can be used such as mean squared error, likelihood loss, cross entropy loss, hinge loss, and/or various other loss functions. Gradient descent techniques can be used to iteratively update the parameters over a number of training iterations."). The additional elements are seen as gathering and outputting data which is insignificant extra solution activity (See MPEP 2106.05(g)) which is well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i) and MPEP 2106.05(d)(II)(iv)) and does not integrate the judicial exception into a practical application. With respect to Applicant's arguments on p. 9 of The Remarks that the claims "rely on and leverage technical architectures", Examiner notes that mere instructions to apply the judicial exception using generic computer components does not integrate the judicial exception into a practical application (MPEP 2106.07(a)(II) "employing well-known computer functions to execute an abstract idea, even when limiting the use of the idea to one particular environment, does not integrate the exception into a practical application"). Examiner also notes (MPEP 2106.05(a) "An important consideration in determining whether a claim improves technology is the extent to which the claim covers a particular solution to a problem or a particular way to achieve a desired outcome, as opposed to merely claiming the idea of a solution or outcome."). Examiner asserts that aside from the recitation of generic computer components and insignificant extra-solution activity, there is nothing in the claims that could not be performed entirely in the human mind with or without the assistance of tools such as pen and paper. For these reasons Examiner asserts that it is reasonable and appropriate to maintain the rejection under 35 USC 101.
Applicant’s arguments with respect to rejection of claims 1-20 under 35 U.S.C. 102/103 based on amendment have been considered and are persuasive. The argument is moot in view of a new ground of rejection set forth below.
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.
Regarding claims 1, 15, and 18, "the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises varying graph model architectures" is indefinite. First, "diverse" is a relative term without any basis for relative comparison to limit the scope of the claim. Secondly, it's unclear what the set relationship between "the plurality of graph models" and "diverse selection of graph models" is: is the plurality itself the "diverse" selection, does the plurality include a subset that is the diverse selection, does the plurality include another collection called "a diverse selection", does each of the graph models comprise a selection of graph models (subgraphs), or something else altogether? Finally, it's unclear what satisfies "varying graph model architectures": different hidden dimension, different activation function, any varied hyperparameter, etc. For these reasons the scope of the claim cannot reasonably be determined. In the interest of further examination the claim is interpreted as each of the plurality of graph models having a unique respective combination of hyperparameters (in other words if the plurality of graph models is two models and the second model has a different activation function (or any hyperparameter) than the first it satisfies the claim limitation).
The remaining claims are rejected with respect to their dependence on the rejected independent claims.
Claim Rejections - 35 USC § 101
101 Rejection
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 USC § 101 because the claimed invention is directed to non-statutory subject matter.
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 system, which is directed to a product, one of the statutory categories.
Step 2A Prong One Analysis: Claim 1 under its broadest reasonable interpretation is a series of mental processes and mathematical calculations and relationships. For example, but for the generic computer components language, the above limitations in the context of this claim encompass machine learning processing, including the following:
generating, by one or more generators, a plurality of synthetic graph datasets, wherein the plurality of synthetic graph datasets comprise structured-graph data (observation, evaluation, and judgement),
“training the plurality of graph models with at least a subset of the plurality of synthetic graph datasets to generate a plurality of trained graph models” (mathematical calculations and relationships in view of the instant specification at [¶0094])
processing one or more inputs from the plurality of synthetic graph datasets with the plurality of trained graph models to generate a plurality of graph outputs (observation, evaluation, and judgement)
determining a particular graph model of the plurality of graph models based on a comparison between the plurality of graph outputs (observation, evaluation, and judgement)
Therefore, claim 1 recites an abstract idea which is a judicial exception.
Step 2A Prong Two Analysis: Claim 1 recites additional elements “A computing system, the system comprising: one or more processors; and one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising”. However, these additional features are computer components recited at a high-level of generality, such that they amount to no more than mere instructions to apply the judicial exception using a generic computer component. An additional element that merely recites the words “apply it” (or an equivalent) with the judicial exception, or merely includes instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, does not integrate the judicial exception into a practical application (See MPEP 2106.05(f)). Claim 1 also recites additional elements “obtaining a plurality of graph models, wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises varying graph model architectures”, “obtaining one or more inputs associated with a particular task from the plurality of synthetic graph datasets”, and “storing data associated with the particular graph model with data associated with the particular task” which amounts to gathering and outputting data which is insignificant extra-solution activity (See MPEP 2106.05(g)). Therefore, claim 1 is directed to a judicial exception.
Step 2B Analysis: Claim 1 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the lack of integration of the abstract idea into a practical application, the additional elements recited in claim 1 amount to no more than mere instructions to apply the judicial exception using a generic computer component and insignificant extra-solution activity. The gathering and outputting of data is considered well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i) and 2106.05(d)(II)(iv)).
For the reasons above, claim 1 is rejected as being directed to non-patentable subject matter under §101. This rejection applies equally to dependent claims 2-14. The additional limitations of the dependent claims are addressed briefly below:
Dependent claim 2 recites additional observation, evaluation, and judgement “generating an evaluation representation associated with the plurality of graph models based on the plurality of graph outputs” as well as additional insignificant extra-solution activity of gathering and outputting data (See MPEP 2106.05(g)) “providing the evaluation representation for display” which is well-understood, routine, and conventional in the art (see MPEP 2106.05(d)(II)(i))
Dependent claim 3 recites additional mathematical calculations and relationships “each of the plurality of synthetic graph datasets comprise a realization of a parameterized probability distribution”
Dependent claim 4 recites additional observation, evaluation, and judgement “each of the plurality of synthetic graph datasets comprise one or more training graphs, one or more training features, and one or more training labels”
Dependent claim 5 recites additional observation, evaluation, and judgement “the one or more generators comprise one or more attributed-graph generators”
Dependent claim 6 recites additional observation, evaluation, and judgement “the one or more generators comprise one or more label generators”
Dependent claim 7 recites additional observation, evaluation, and judgement “each of the plurality of graph models comprise a graph neural network”
Dependent claim 8 recites additional observation, evaluation, and judgement “the subset of the plurality of synthetic graph datasets and the one or more inputs from the plurality of synthetic graph datasets differ”
Dependent claim 9 recites additional insignificant extra-solution activity of gathering and outputting data (See MPEP 2106.05(g)) “obtaining, from a user computing device, a user-input graph model” which is well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i)) as well as additional mathematical calculations and relationships “training the user-input graph model with a first synthetic graph dataset of the plurality of synthetic graph datasets to generate a first trained graph model; training the user-input graph model with a second synthetic graph dataset of the plurality of synthetic graph datasets to generate a second trained graph model;” (See the instant specification at [¶0094] which explicitly describes the training as mathematical calculations). Claim 9 also recites additional observation, evaluation, and judgement “processing a test portion of the plurality of synthetic graph datasets with the first trained graph model to generate a plurality of first user-model outputs; processing the test portion of the plurality of synthetic graph dataset with the second trained graph model to generate a plurality of second user-model outputs; and comparing the plurality of first user-model outputs and plurality of second user-model outputs”
Dependent claim 10 recites additional observation, evaluation, and judgement “generating evaluation data based at least in part on the plurality of first user-model outputs and plurality of second user-model outputs” as well as additional insignificant extra-solution activity of gathering and outputting data (See MPEP 2106.05(g)) “providing the evaluation data to the user computing device” which is well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i))
Dependent claim 11 recites additional observation, evaluation, and judgement “generating comparison data based on the plurality of first user-model outputs, the plurality of second user-model outputs, and the plurality of graph output” as well as additional insignificant extra-solution activity of gathering and outputting data (See MPEP 2106.05(g)) “providing the comparison data to the user computing device” which is well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i))
Dependent claim 12 recites additional insignificant extra-solution activity of gathering and outputting data (See MPEP 2106.05(g)) “obtaining input data associated with a specific task” which is well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i)), as well as additional mathematical calculations and relationships “wherein training the plurality of graph models with at least the subset of the plurality of synthetic graph datasets to generate the plurality of trained graph models comprises: training the plurality of graph models to perform the specific task” (See the instant specification at [¶0094] which explicitly describes the training as mathematical calculations and relationships).
Dependent claim 13 recites additional observation, evaluation, and judgement “the plurality of synthetic graph datasets are generated based on the input data; and wherein the plurality of synthetic graph datasets comprise a plurality of labels associated with the specific task”
Dependent claim 14 recites additional mathematical calculations and relationships “training the plurality of graph models with at least the subset of the plurality of synthetic graph datasets to generate the plurality of trained graph models comprises: training a first graph model of the plurality of graph models with a first synthetic graph dataset of the plurality of synthetic graph datasets to generate a first trained graph model; training a first graph model of the plurality of graph models with a second synthetic graph dataset of the plurality of synthetic graph datasets to generate a second trained graph model; training a second graph model of the plurality of graph models with a first synthetic graph dataset of the plurality of synthetic graph datasets to generate a third trained graph model; training a second graph model of the plurality of graph models with a second synthetic graph dataset of the plurality of synthetic graph datasets to generate a fourth trained graph model; and wherein the plurality of trained graph models comprises the first trained graph model, the second trained graph model, the third trained graph model, and the fourth trained graph model.”
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 method, which is directed to a process, one of the statutory categories.
Step 2A Prong One Analysis: Claim 15 under its broadest reasonable interpretation is a series of mental processes and mathematical calculations and relationships. For example, but for the generic computer components language, the above limitations in the context of this claim encompass machine learning processing, including the following:
generating, […], a plurality of synthetic graph datasets using a plurality of sampled configurations, wherein each sampled configuration is generated by sampling a generator parameter set of a generator parameter space, wherein the plurality of synthetic graph datasets comprise structured-graph data (observation, evaluation, and judgement),
training, […], the plurality of graph models with at least a subset of the plurality of synthetic graph datasets to generate a plurality of trained graph models (mathematical calculations and relationships in view of the instant specification at [¶0094])
processing, […], the one or more inputs from the plurality of synthetic graph datasets with the plurality of trained graph models to generate a plurality of graph outputs (observation, evaluation, and judgement)
generating, […], an evaluation representation associated with the plurality of graph models based on the plurality of graph outputs (observation, evaluation, and judgement)
determining, […], a particular graph model of the plurality of graph models based on the evaluation representation (observation, evaluation, and judgement)
Therefore, claim 15 recites an abstract idea which is a judicial exception.
Step 2A Prong Two Analysis: Claim 15 recites additional elements “A computer-implemented method”, “by a computing system comprising one or more processors”. However, these additional features are computer components recited at a high-level of generality, such that they amount to no more than mere instructions to apply the judicial exception using a generic computer component. An additional element that merely recites the words “apply it” (or an equivalent) with the judicial exception, or merely includes instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, does not integrate the judicial exception into a practical application (See MPEP 2106.05(f)). Claim 15 also recites additional elements “obtaining, by the computing system, a plurality of graph models, wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises differing model architectures,” “obtaining, by the computing system, one or more inputs associated with a particular task from the plurality of synthetic graph datasets;”, “providing, by the computing system, the evaluation representation for display,” and “storing, by the computing system, data associated with the particular graph model with data associated with the particular task” which amounts to gathering and outputting data which is insignificant extra-solution activity (See MPEP 2106.05(g)). Therefore, claim 15 is directed to a judicial exception.
Step 2B Analysis: Claim 15 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the lack of integration of the abstract idea into a practical application, the additional elements recited in claim 15 amount to no more than mere instructions to apply the judicial exception using a generic computer component and insignificant extra-solution activity. The gathering and outputting of data is considered well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i) and 2106.05(d)(II)(iv)).
For the reasons above, claim 15 is rejected as being directed to non-patentable subject matter under §101. This rejection applies equally to dependent claims 16-17. The additional limitations of the dependent claims are addressed briefly below:
Dependent claim 16 recites additional observation, evaluation, and judgement “wherein the evaluation representation comprises evaluation data descriptive of node classification for the plurality of graph models”
Dependent claim 17 recites additional mathematical calculations and relationships “the evaluation representation comprises evaluation data descriptive of link prediction for the plurality of graph models”
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 computer readable media, which is directed to a product, one of the statutory categories.
Step 2A Prong One Analysis: Claim 18 under its broadest reasonable interpretation is a series of mental processes and mathematical calculations and relationships. For example, but for the generic computer components language, the above limitations in the context of this claim encompass machine learning processing, including the following:
generating, […], a plurality of synthetic graph datasets based at least in part on the input data, wherein the plurality of synthetic graph datasets comprise structured-graph data (observation, evaluation, and judgement),
training a plurality of graph models with at least a subset of the plurality of synthetic graph datasets to generate a plurality of trained graph models (mathematical calculations and relationships in view of the instant specification at [¶0094])
processing, […], the one or more inputs from the plurality of synthetic graph datasets with the plurality of trained graph models to generate a plurality of graph outputs (observation, evaluation, and judgement)
processing one or more inputs associated with a particular task from the plurality of synthetic graph datasets with the plurality of trained graph models to generate a plurality of graph outputs (observation, evaluation, and judgement)
generating an output representation associated with the plurality of graph models based on the plurality of graph outputs (observation, evaluation, and judgement)
determining a particular graph model of the plurality of graph models based on the output representation (observation, evaluation, and judgement)
Therefore, claim 18 recites an abstract idea which is a judicial exception.
Step 2A Prong Two Analysis: Claim 18 recites additional elements “One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices, cause the one or more computing devices to perform operations”. However, these additional features are computer components recited at a high-level of generality, such that they amount to no more than mere instructions to apply the judicial exception using a generic computer component. An additional element that merely recites the words “apply it” (or an equivalent) with the judicial exception, or merely includes instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea, does not integrate the judicial exception into a practical application (See MPEP 2106.05(f)). Claim 18 also recites additional elements “obtaining input data associated with a user, “obtaining a plurality of graph models, wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises a variety of graph model architectures”, “providing the output representation for display”, and “storing data associated with the particular graph model with data associated with the particular task” which amounts to gathering and outputting data which is insignificant extra-solution activity (See MPEP 2106.05(g)). Therefore, claim 18 is directed to a judicial exception.
Step 2B Analysis: Claim 18 does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to the lack of integration of the abstract idea into a practical application, the additional elements recited in claim 18 amount to no more than mere instructions to apply the judicial exception using a generic computer component and insignificant extra-solution activity. The gathering and outputting of data is considered well-understood, routine, and conventional in the art (See MPEP 2106.05(d)(II)(i) and 2106.05(d)(II)(iv)).
For the reasons above, claim 18 is rejected as being directed to non-patentable subject matter under §101. This rejection applies equally to dependent claims 19-20. The additional limitations of the dependent claims are addressed briefly below:
Dependent claim 16 recites additional observation, evaluation, and judgement “the output representation comprises a graphical depiction of a feature center distance based on the plurality of graph outputs associated with the plurality of graph models”
Dependent claim 20 recites additional mathematical calculations and relationships “the output representation comprises vector graph statistic data and hyperparameter evaluation data”
Therefore, when considering the elements separately and in combination, they do not add significantly more to the inventive concept. Accordingly, claims 1-8 are rejected under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 103
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.
Claims 1-20 are rejected under U.S.C. §103 as being unpatentable over the combination of Zheng (US20210374279A1) and You (“Design Space for Graph Neural Networks”, 2020).
Regarding claim 1, Zheng teaches A computing system, the system comprising: one or more processors; and one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising:([¶0082] "A processor 407 includes a central processing unit, a graphics processing unit, and/or the like. A memory 408 includes random-access memory, read-only memory, and/or the like. The memory 408 may store a set of instructions (e.g., one or more instructions) for execution by the processor 407. The processor 407 may execute the set of instructions to perform one or more operations or processes described herein. In some implementations, execution of the set of instructions, by one or more processors 407, causes the one or more processors 407 and/or the synthetic data modeling system 401 to perform one or more operations or processes described herein.")
generating, by one or more generators, a plurality of synthetic graph datasets, ([¶0034] " the generator may be referred to as a graph generator model." [¶0048] " the generator model may select a random (or latent) value (e.g., using a Gaussian distribution), and may use the random value to generate a synthetic knowledge graph (e.g., a synthetic adjacency matrix and/or a synthetic attribute matrix)")
wherein the plurality of synthetic graph datasets comprise structured-graph data;([¶0013] "A knowledge graph includes a set of nodes and interconnections between nodes. The nodes may represent the same type of entity (e.g., people), or a node may have a node type that indicates a type of entity represented by the node. A node may be associated with a set of attributes that define characteristics of the node (e.g., height, weight, and gender for a node that represents a person). Interconnections between nodes may be referred to as edges. In some cases, edges may have different edge types that represent different relationships between nodes.")
obtaining a plurality of graph models, ([¶0031] "the synthetic data modeling system may train a set of teacher models (e.g., machine learning models) using the true data partitions of the true knowledge graph and the synthetic data partitions of the synthetic knowledge graph, which may be generated by the synthetic data modeling system as described above")
training the plurality of graph models with at least a subset of the plurality of synthetic graph datasets to generate a plurality of trained graph models;([¶0031] " the synthetic data modeling system may train a set of teacher models (e.g., machine learning models) using the true data partitions of the true knowledge graph and the synthetic data partitions of the synthetic knowledge graph, which may be generated by the synthetic data modeling system as described above")
obtaining one or more inputs associated with a particular task from the plurality of synthetic graph datasets;([¶0073] "FIG. 3 is a diagram illustrating an example 300 of applying a trained machine learning model to a new observation associated with differentially private dataset generation and modeling for knowledge graphs. ")
processing the one or more inputs from the plurality of synthetic graph datasets with the plurality of trained graph models to generate a plurality of graph outputs;([¶0039] "As shown by reference number 124, the trained teacher models may generate predictions based on the synthetic data partitions. As described above, a prediction may indicate a probability (sometimes referred to as a confidence score) that the data is true data and/or a probability that the data is synthetic data." [¶0073] "FIG. 3 is a diagram illustrating an example 300 of applying a trained machine learning model to a new observation associated with differentially private dataset generation and modeling for knowledge graphs. ")
determining a particular graph model of the plurality of graph models based on a comparison between the plurality of graph outputs; and([¶0071] "The machine learning model may compare the performance scores for each machine learning model, and may select the machine learning model with the best (e.g., highest accuracy, lowest error, closest to a desired threshold, and/or the like) performance score as the trained machine learning model 245.")
storing data associated with the particular graph model with data associated with the particular task.([¶0070] "If the machine learning model performs adequately (e.g., with a performance score that satisfies a threshold), then the machine learning system may store that machine learning model as a trained machine learning model 245 to be used to analyze new observations, as described below in connection with FIG. 3." [¶0082] "A storage component 409 includes a hard disk or another type of storage device that stores information, data, and/or software (e.g., code, instructions, and/or the like) related to the operation and use of the synthetic data modeling system 401").
However, Zheng does not explicitly teach wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises varying graph model architectures.
You, in the same field of endeavor, teaches wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises varying graph model architectures([Abstract] "current research focuses on proposing and evaluating specific architectural designs of GNNs, such as GCN,GIN, or GAT [...] we define and systematically study the architectural design space for GNNs which consists of 315,000 different designs over 32 different predictive tasks" [p. 2] "The GNN task space with a task similarity metric allows us to identify novel tasks and effectively transfer GNN architectural designs between similar tasks").
Zheng as well as You are directed towards dataset generation and graph modeling. Therefore, Zheng as well as You are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention implement Zheng's teacher/student models as a diverse set of GNN architectures described by You. You provides as additional motivation for combination ([p. 8] “The best model in our design space is better than the best GCN model in 24 out of 32 tasks”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 2, the combination of Zheng, and You teaches The computing system of claim 1, wherein the operations further comprising: generating an evaluation representation associated with the plurality of graph models based on the plurality of graph outputs; and(Zheng FIG. 6 step 640 shows that the results of the teacher models are aggregated l step 650 shows that the results and the output of the student model which is also a trained graph model, is compared based on a classification loss, thus corresponding to “a comparison between the plurality of graph outputs”; step 655 shows that based on a condition, “output…a model used to generate the synthetic knowledge graph”; [¶0050] “…. the synthetic data modeling system may perform the statistical analysis on the final synthetic knowledge graph, and may output a result of performing the statistical analysis. For example, the synthetic data modeling system may output the result for display, may output the result to another device (e.g., a user device) for display by the other device, and/or the like.”; “statistical analysis” corresponds to “generating an evaluation representation”)
providing the evaluation representation for display.(Zheng [¶0050] “…. the synthetic data modeling system may perform the statistical analysis on the final synthetic knowledge graph, and may output a result of performing the statistical analysis. For example, the synthetic data modeling system may output the result for display, may output the result to another device (e.g., a user device) for display by the other device, and/or the like.”).
Regarding claim 3, the combination of Zheng, and You teaches The computing system of claim 1, wherein each of the plurality of synthetic graph datasets comprise a realization of a parameterized probability distribution.(Zheng [¶0048] “The generator model may process the input to generate an output (e.g., a synthetic knowledge graph), and may output the result (e.g., to the device). In some implementations, the generator model may select a random (or latent) value (e.g., using a Gaussian distribution), and may use the random value to generate a synthetic knowledge graph”).
Regarding claim 4, the combination of Zheng, and You teaches The computing system of claim 1, wherein each of the plurality of synthetic graph datasets comprise one or more training graphs, one or more training features, and one or more training labels.(Zheng [¶0028] “As shown by reference number 112, using attribute partitioning, the synthetic data modeling system may partition a knowledge graph such that each partition includes an unpartitioned adjacency matrix (e.g., from the original knowledge graph that is being partitioned) and includes only the attribute vectors for the nodes that are included in that partition. As a simple example, for a knowledge graph with four nodes (e.g., 1, 2, 3, and 4), attribute partitioning can be used to create a first partition that includes the entire adjacency matrix from the unpartitioned knowledge graph (e.g., indicating all relationships among nodes 1, 2, 3, and 4) and includes only the attribute vectors for nodes 1 and 2 (shown as x1 and x2) and not the attribute vectors for nodes 3 and 4 (shown as x3 and x4), and to create a second partition that includes the entire adjacency matrix from the unpartitioned knowledge graph and includes only the attribute vectors for nodes 3 and 4 (shown as x3 and x4) and not the attribute vectors for nodes 1 and 2 (shown as x1 and x2).”).
Regarding claim 5, the combination of Zheng, and You teaches The computing system of claim 1, wherein the one or more generators comprise one or more attributed-graph generators.(Zheng [¶0003] "generate, using a graph generator model, a synthetic knowledge graph based on a true knowledge graph that is representative of a dataset; partition the synthetic knowledge graph into a set of synthetic data partitions; partition the true knowledge graph into a set of true data partitions”; FIG. 6 step 620 states “Generate…a synthetic knowledge graph that includes a synthetic adjacency matrix and a synthetic attribute matrix” since the generator generates the synthetic knowledge graph including its attribute matrix, the generator is an “attribute-graph generator”, thus “the generator” corresponds to “one or more generators” that “comprise attribute-graph generators”).
Regarding claim 6, the combination of Zheng, and You teaches The computing system of claim 1, wherein the one or more generators comprise one or more label generators.(Zheng [¶0035] “the generator may determine the values and positions of those values (e.g., in elements of the adjacency matrix and/or the attribute matrix) that tend to result in a larger classification error, and may generate a new synthetic knowledge graph that includes those values in those positions to attempt to increase the classification error. By receiving a label for data and/or comparing a prediction to the label, a teacher model and/or a student model may determine values” Since the generator can also generate a synthetic graph and include labels, “the generator” is also a “label generator”, thus “the generator” corresponds to “one or more generators” that “comprise one or labeled generators").
Regarding claim 7, the combination of Zheng, and You teaches The computing system of claim 1, wherein each of the plurality of graph models comprise a graph neural network.(You [Abstract] "current research focuses on proposing and evaluating specific architectural designs of GNNs, such as GCN,GIN, or GAT [...] we define and systematically study the architectural design space for GNNs which consists of 315,000 different designs over 32 different predictive tasks" [p. 2] "The GNN task space with a task similarity metric allows us to identify novel tasks and effectively transfer GNN architectural designs between similar tasks").
Regarding claim 8, the combination of Zheng, and You teaches The computing system of claim 1, wherein the subset of the plurality of synthetic graph datasets and the one or more inputs from the plurality of synthetic graph datasets differ.(Zheng [¶0028] “As shown by reference number 112, using attribute partitioning, the synthetic data modeling system may partition a knowledge graph such that each partition includes an unpartitioned adjacency matrix (e.g., from the original knowledge graph that is being partitioned) and includes only the attribute vectors for the nodes that are included in that partition. As a simple example, for a knowledge graph with four nodes (e.g., 1, 2, 3, and 4), attribute partitioning can be used to create a first partition that includes the entire adjacency matrix from the unpartitioned knowledge graph (e.g., indicating all relationships among nodes 1, 2, 3, and 4) and includes only the attribute vectors for nodes 1 and 2 (shown as x1 and x2) and not the attribute vectors for nodes 3 and 4 (shown as x3 and x4), and to create a second partition that includes the entire adjacency matrix from the unpartitioned knowledge graph and includes only the attribute vectors for nodes 3 and 4 (shown as x3 and x4) and not the attribute vectors for nodes 1 and 2 (shown as x1 and x2).”; [0032] “As shown, the synthetic data modeling system may label (e.g., categorize, classify, and/or the like) the partition from the synthetic knowledge graph with a first label (shown as 0) to indicate that the partition includes synthetic data (e.g., from the synthetic knowledge graph), and may label the partition from the true knowledge graph with a second label (shown as 1) to indicate that the partition includes true data (e.g., from the true knowledge graph). The teacher model may be trained, using the labeled inputs” “first partition” corresponds to “the subset of the plurality of synthetic graph datasets”; “first label” and “second label” corresponds to “one or more inputs” from the plurality of synthetic graph datasets”; the partition includes matrices whereas the labels are numerical data type, hence this corresponds to “the subset of the plurality of synthetic graph datasets and the one or more inputs from the plurality of synthetic graph datasets differ”).
Regarding claim 9, the combination of Zheng, and You teaches The computing system of claim 1, wherein the operations further comprise: obtaining, from a user computing device, a [user-input] graph model;(Zheng [¶0048] “Additionally, or alternatively, the synthetic data modeling system may provide access to the generator model via a website, an application, and/or the like. In this case, a user may interact with a device (e.g., a user device) to provide input to the generator model (e.g., a dataset or true knowledge graph) “a user device” corresponds to “a user computing device”)
training the user-input graph model with a first synthetic graph dataset of the plurality of synthetic graph datasets to generate a first trained graph model;(Zheng FIGs. 1C and 1D shows a “teacher model” which corresponds to “the…graph model” is being trained on “Synthetic Data Partition 1” which corresponds to “a first synthetic graph dataset of the plurality of synthetic graph datasets” to generate “Trained Model 1” which corresponds to “a first trained graph model”)
training the [user-input] graph model with a second synthetic graph dataset of the plurality of synthetic graph datasets to generate a second trained graph model;(Zheng FIGs. 1C and 1D shows a “teacher model” which corresponds to “the…graph model” is being trained on “Synthetic Data Partition 2” which corresponds to “a second synthetic graph dataset of the plurality of synthetic graph datasets” to generate “Trained Model 2” which corresponds to “a second trained graph model”);)
processing a test portion of the plurality of synthetic graph datasets with the first trained graph model to generate a plurality of first [user-model] outputs;(Zheng [¶0065] “As further shown, the machine learning system may partition the set of observations into a .. test set that includes a second subset of observations… test set 225 may be used to test whether the trained model accurately predicts target variables in the second subset of observations. " [¶0031] “For example, the synthetic data modeling system may partition the true knowledge graph into a quantity of partitions, may partition the synthetic knowledge graph into the same quantity of partitions, and may use those partitions to train the same quantity of teacher models.”; [¶0032] “As shown, the synthetic data modeling system may label (e.g., categorize, classify, and/or the like) the partition from the synthetic knowledge graph with a first label (shown as 0) to indicate that the partition includes synthetic data (e.g., from the synthetic knowledge graph), and may label the partition from the true knowledge graph with a second label (shown as 1) to indicate that the partition includes true data (e.g., from the true knowledge graph). The teacher model may be trained, using the labeled inputs, to differentiate between (or determine probability or confidence scores for) true data and synthetic data by applying a machine learning algorithm, as described in more detail below in connection with FIG. 2.” based on FIG. 2 and [0032] each of the “machine learning model” mentioned in [0071] refer to the teacher and student models; FIG. 2 shows “a test set” which comprises “Synthetic Data Partition X”, thus it corresponds to “a test portion of the plurality of the synthetic graph datasets”; since the machine learning system tests each “machine learning model using the test set 225” which includes “Trained Model 1” from FIGs. 1C and 1D; thus overall, this corresponds to “processing a test portion of the plurality of synthetic graph datasets with the first trained graph model”; since “the trained model” including “Trained Model 1” predicts “target variables” in the test set, this corresponds to “generate a plurality of first…outputs”)
processing the test portion of the plurality of synthetic graph dataset with the second trained graph model to generate a plurality of second [user-model] outputs; and(Zheng [0065] “As further shown, the machine learning system may partition the set of observations into a .. test set that includes a second subset of observations… test set 225 may be used to test whether the trained model accurately predicts target variables in the second subset of observations. " [0031] “For example, the synthetic data modeling system may partition the true knowledge graph into a quantity of partitions, may partition the synthetic knowledge graph into the same quantity of partitions, and may use those partitions to train the same quantity of teacher models.”; [0032] “As shown, the synthetic data modeling system may label (e.g., categorize, classify, and/or the like) the partition from the synthetic knowledge graph with a first label (shown as 0) to indicate that the partition includes synthetic data (e.g., from the synthetic knowledge graph), and may label the partition from the true knowledge graph with a second label (shown as 1) to indicate that the partition includes true data (e.g., from the true knowledge graph). The teacher model may be trained, using the labeled inputs, to differentiate between (or determine probability or confidence scores for) true data and synthetic data by applying a machine learning algorithm, as described in more detail below in connection with FIG. 2.” based on FIG. 2 and [0032] each of the “machine learning model” mentioned in [0071] refer to the teacher and student models; FIG. 2 shows “a test set” which comprises “Synthetic Data Partition X”, thus it corresponds to “a test portion of the plurality of the synthetic graph datasets”; since the machine learning system tests each “machine learning model using the test set 225” which includes “Trained Model 2” from FIGs. 1C and 1D; thus overall, this corresponds to “processing a test portion of the plurality of synthetic graph datasets with the second trained graph model”; since “the trained model” including “Trained Model 2” predicts “target variables” in the test set, this corresponds to “generate a plurality of second…outputs”))
comparing the plurality of first [user-model] outputs and plurality of second [user-model] outputs.(Zheng [0065] “As further shown, the machine learning system may partition the set of observations into a .. test set that includes a second subset of observations… test set 225 may be used to test whether the trained model accurately predicts target variables in the second subset of observations.”; [0071] “The machine learning system may test each machine learning model using the test set 225 to generate a corresponding performance score for each machine learning model. The machine learning model may compare the performance scores for each machine learning model….”; [0044] “As shown in FIG. 1E, the synthetic data modeling system may determine a classification error associated with the prediction of the student model, sometimes referred to as a classification loss. In some implementations, the classification loss may be calculated as a difference between the aggregated prediction (shown as AP) and the student prediction (shown as SP). The difference may be an absolute difference (e.g., an absolute value) without an accompanying positive or negative sign, or may be a relative difference with an accompanying positive or negative sign (e.g., indicating whether the student prediction was greater than or less than the aggregated prediction, which indicates whether the aggregated prediction or the student prediction was more accurate). Alternatively, the classification loss may be calculated as an error metric other than a difference, such as a percentage error. In example 100, the aggregated prediction is 1, the student prediction is 0, and the classification loss is 1. As another example, if the aggregated prediction is 0.493 and the student prediction is 0.293, then the classification loss would be 0.200 as stated above in FIGs. 1C and 1D “Trained Model 1” corresponds to “first trained graph model” and produces “the plurality of first…outputs” and “Trained Model 2” corresponds to “second trained graph model” and produces “the plurality of second…outputs”; since “Trained Model 1” and “Trained Model 2” can be any teacher and student model, “teacher model” corresponds to a first graph model producing “first…outputs” and “student model” corresponds to a second graph model producing “second…outputs” ; since the classification loss uses prediction results from the student and teacher models, “classification loss” correspond to “comparison data”)
user-input graph model(You [p. 6] "users can easily import new design dimensions to GraphGym, such as new types of GNN layers or new connectivity patterns across layers. We provide an example of using ATTENTION as a new intra-layer design dimension in the Appendix")
user-model(You [p. 6] "users can easily import new design dimensions to GraphGym, such as new types of GNN layers or new connectivity patterns across layers. We provide an example of using ATTENTION as a new intra-layer design dimension in the Appendix").
Regarding claim 10, the combination of Zheng, and You teaches The computing system of claim 9, wherein the operations further comprise: generating evaluation data based at least in part on the plurality of first [user-model] outputs and plurality of second [user-model] outputs; and(Zheng [0065] “As further shown, the machine learning system may partition the set of observations into a .. test set that includes a second subset of observations… test set 225 may be used to test whether the trained model accurately predicts target variables in the second subset of observations.”; [0071] “The machine learning system may test each machine learning model using the test set 225 to generate a corresponding performance score for each machine learning model. The machine learning model may compare the performance scores for each machine learning model….”; [0044] “As shown in FIG. 1E, the synthetic data modeling system may determine a classification error associated with the prediction of the student model, sometimes referred to as a classification loss. In some implementations, the classification loss may be calculated as a difference between the aggregated prediction (shown as AP) and the student prediction (shown as SP). The difference may be an absolute difference (e.g., an absolute value) without an accompanying positive or negative sign, or may be a relative difference with an accompanying positive or negative sign (e.g., indicating whether the student prediction was greater than or less than the aggregated prediction, which indicates whether the aggregated prediction or the student prediction was more accurate). Alternatively, the classification loss may be calculated as an error metric other than a difference, such as a percentage error. In example 100, the aggregated prediction is 1, the student prediction is 0, and the classification loss is 1. As another example, if the aggregated prediction is 0.493 and the student prediction is 0.293, then the classification loss would be 0.200 as stated above in FIGs. 1C and 1D “Trained Model 1” corresponds to “first trained graph model” and produces “the plurality of first…outputs” and “Trained Model 2” corresponds to “second trained graph model” and produces “the plurality of second…outputs”; since “Trained Model 1” and “Trained Model 2” can be any teacher and student model, “teacher model” corresponds to a first graph model producing “first…outputs” and “student model” corresponds to a second graph model producing “second…outputs” ; since the classification loss uses prediction results from the student and teacher models, “classification loss” correspond to “comparison data”)
providing the evaluation data to the user computing device.(Zheng [¶0075] “the machine learning system may provide output, such as outputting the value of the prediction, calculating a classification loss based on the prediction, providing feedback to a generator to generate a new synthetic knowledge graph, an indication that a stopping condition has been satisfied, and/or the like.”)
user-model(You [p. 6] "users can easily import new design dimensions to GraphGym, such as new types of GNN layers or new connectivity patterns across layers. We provide an example of using ATTENTION as a new intra-layer design dimension in the Appendix").
Regarding claim 11, the combination of Zheng, and You teaches The computing system of claim 9, wherein the operations further comprise: generating comparison data based on the plurality of first [user-model] outputs, the plurality of second [user-model] outputs, and the plurality of graph outputs; and(Zheng [0065] “As further shown, the machine learning system may partition the set of observations into a .. test set that includes a second subset of observations… test set 225 may be used to test whether the trained model accurately predicts target variables in the second subset of observations.”; [0071] “The machine learning system may test each machine learning model using the test set 225 to generate a corresponding performance score for each machine learning model. The machine learning model may compare the performance scores for each machine learning model….”; [0044] “As shown in FIG. 1E, the synthetic data modeling system may determine a classification error associated with the prediction of the student model, sometimes referred to as a classification loss. In some implementations, the classification loss may be calculated as a difference between the aggregated prediction (shown as AP) and the student prediction (shown as SP). The difference may be an absolute difference (e.g., an absolute value) without an accompanying positive or negative sign, or may be a relative difference with an accompanying positive or negative sign (e.g., indicating whether the student prediction was greater than or less than the aggregated prediction, which indicates whether the aggregated prediction or the student prediction was more accurate). Alternatively, the classification loss may be calculated as an error metric other than a difference, such as a percentage error. In example 100, the aggregated prediction is 1, the student prediction is 0, and the classification loss is 1. As another example, if the aggregated prediction is 0.493 and the student prediction is 0.293, then the classification loss would be 0.200)
providing the comparison data to the user computing device.(Zheng [¶0075] “the machine learning system may provide output, such as outputting the value of the prediction, calculating a classification loss based on the prediction, providing feedback to a generator to generate a new synthetic knowledge graph, an indication that a stopping condition has been satisfied, and/or the like.”)
user-model(You [p. 6] "users can easily import new design dimensions to GraphGym, such as new types of GNN layers or new connectivity patterns across layers. We provide an example of using ATTENTION as a new intra-layer design dimension in the Appendix").
Regarding claim 12, the combination of Zheng, and You teaches The computing system of claim 1, wherein the operations further comprise: obtaining input data associated with a specific task; and(Zheng [¶0049] “…. a user may interact with a device (e.g., a user device) to provide input to the generator model (e.g., a dataset or true knowledge graph). The generator model may process the input to generate an output (e.g., a synthetic knowledge graph), and may output the result (e.g., to the device).” [¶0032] “The teacher model may be trained, using the labeled inputs, to differentiate between (or determine probability or confidence scores for) true data and synthetic data by applying a machine learning algorithm, as described in more detail below in connection with FIG. 2. Each teacher model may be trained in this manner, either using the same machine learning model or different machine learning models for training on respective partitions.”; “labeled inputs” and “provide input” correspond to “obtaining input data”; “differentiate between…true data and synthetic data” corresponds to “a specific task”))
wherein training the plurality of graph models with at least the subset of the plurality of synthetic graph datasets to generate the plurality of trained graph models comprises:(Zheng [¶0031] “For example, the synthetic data modeling system may partition the true knowledge graph into a quantity of partitions, may partition the synthetic knowledge graph into the same quantity of partitions, and may use those partitions to train the same quantity of teacher models.” “teacher models” correspond to “a plurality of graph models” and “trained graph models”; “partitions” corresponds to “at least a subset of the plurality of synthetic graph datasets”)
training the plurality of graph models to perform the specific task.(Zheng [¶0049] "a user may interact with a device (e.g., a user device) to provide input to the generator model (e.g., a dataset or true knowledge graph). The generator model may process the input to generate an output (e.g., a synthetic knowledge graph), and may output the result (e.g., to the device).”; [¶0032] “The teacher model may be trained, using the labeled inputs, to differentiate between (or determine probability or confidence scores for) true data and synthetic data by applying a machine learning algorithm, as described in more detail below in connection with FIG. 2. Each teacher model may be trained in this manner, either using the same machine learning model or different machine learning models for training on respective partitions.”).
Regarding claim 13, the combination of Zheng, and You teaches The computing system of claim 12, wherein the plurality of synthetic graph datasets are generated based on the input data; (Zheng [¶0049] "a user may interact with a device (e.g., a user device) to provide input to the generator model (e.g., a dataset or true knowledge graph). The generator model may process the input to generate an output (e.g., a synthetic knowledge graph), and may output the result (e.g., to the device).”)
and wherein the plurality of synthetic graph datasets comprise a plurality of labels associated with the specific task.(Zheng [¶0049] “…. a user may interact with a device (e.g., a user device) to provide input to the generator model (e.g., a dataset or true knowledge graph). The generator model may process the input to generate an output (e.g., a synthetic knowledge graph), and may output the result (e.g., to the device).”; [¶0032] “The teacher model may be trained, using the labeled inputs, to differentiate between (or determine probability or confidence scores for) true data and synthetic data by applying a machine learning algorithm, as described in more detail below in connection with FIG. 2. Each teacher model may be trained in this manner, either using the same machine learning model or different machine learning models for training on respective partitions.”;).
Regarding claim 14, the combination of Zheng, and You teaches The computing system of claim 1, wherein training the plurality of graph models with at least the subset of the plurality of synthetic graph datasets to generate the plurality of trained graph models comprises: training a first graph model of the plurality of graph models with a first synthetic graph dataset of the plurality of synthetic graph datasets to generate a first trained graph model;(You [p. 5] "we collect a variety of 32 synthetic and real-world GNN tasks/datasets" [p. 6] "We evaluate the proposed GNN design space (Section 4) over the GNN task space (Section 5), using the proposed evaluation techniques (Section 6). For all the experiments in Sections 7.3 and 7.4, we use a consistent setup, where results on three random 80%/20% train/val splits are averaged" [p. 5] "the computation cost to compare T GNN tasks is to train and evaluate M ⇤ T GNN models. We show that M = 12 anchor models is sufficient" You explicitly performs three random splits of synthetic dataset for training of M GNN models)
training a first graph model of the plurality of graph models with a second synthetic graph dataset of the plurality of synthetic graph datasets to generate a second trained graph model;(You [p. 5] "we collect a variety of 32 synthetic and real-world GNN tasks/datasets" [p. 6] "We evaluate the proposed GNN design space (Section 4) over the GNN task space (Section 5), using the proposed evaluation techniques (Section 6). For all the experiments in Sections 7.3 and 7.4, we use a consistent setup, where results on three random 80%/20% train/val splits are averaged" [p. 5] "the computation cost to compare T GNN tasks is to train and evaluate M ⇤ T GNN models. We show that M = 12 anchor models is sufficient" You explicitly performs three random splits of synthetic dataset for training of M GNN models)
training a second graph model of the plurality of graph models with a first synthetic graph dataset of the plurality of synthetic graph datasets to generate a third trained graph model;(You [p. 5] "we collect a variety of 32 synthetic and real-world GNN tasks/datasets" [p. 6] "We evaluate the proposed GNN design space (Section 4) over the GNN task space (Section 5), using the proposed evaluation techniques (Section 6). For all the experiments in Sections 7.3 and 7.4, we use a consistent setup, where results on three random 80%/20% train/val splits are averaged" [p. 5] "the computation cost to compare T GNN tasks is to train and evaluate M ⇤ T GNN models. We show that M = 12 anchor models is sufficient" You explicitly performs three random splits of synthetic dataset for training of M GNN models)
training a second graph model of the plurality of graph models with a second synthetic graph dataset of the plurality of synthetic graph datasets to generate a fourth trained graph model; and(You [p. 5] "we collect a variety of 32 synthetic and real-world GNN tasks/datasets" [p. 6] "We evaluate the proposed GNN design space (Section 4) over the GNN task space (Section 5), using the proposed evaluation techniques (Section 6). For all the experiments in Sections 7.3 and 7.4, we use a consistent setup, where results on three random 80%/20% train/val splits are averaged" [p. 5] "the computation cost to compare T GNN tasks is to train and evaluate M ⇤ T GNN models. We show that M = 12 anchor models is sufficient" You explicitly performs three random splits of synthetic dataset for training of M GNN models)
wherein the plurality of trained graph models comprises the first trained graph model, the second trained graph model, the third trained graph model, and the fourth trained graph model.(You [p. 5] "we collect a variety of 32 synthetic and real-world GNN tasks/datasets" [p. 6] "We evaluate the proposed GNN design space (Section 4) over the GNN task space (Section 5), using the proposed evaluation techniques (Section 6). For all the experiments in Sections 7.3 and 7.4, we use a consistent setup, where results on three random 80%/20% train/val splits are averaged" [p. 5] "the computation cost to compare T GNN tasks is to train and evaluate M ⇤ T GNN models. We show that M = 12 anchor models is sufficient" You explicitly performs three random splits of synthetic dataset for training of M GNN models).
Regarding claim 15, Zheng teaches A computer-implemented method, the method comprising: generating, by a computing system comprising one or more processors, ([¶0082] "A processor 407 includes a central processing unit, a graphics processing unit, and/or the like. A memory 408 includes random-access memory, read-only memory, and/or the like. The memory 408 may store a set of instructions (e.g., one or more instructions) for execution by the processor 407. The processor 407 may execute the set of instructions to perform one or more operations or processes described herein. In some implementations, execution of the set of instructions, by one or more processors 407, causes the one or more processors 407 and/or the synthetic data modeling system 401 to perform one or more operations or processes described herein.")
a plurality of synthetic graph datasets using a plurality of sampled configurations, ([¶0034] " the generator may be referred to as a graph generator model." [¶0048] " the generator model may select a random (or latent) value (e.g., using a Gaussian distribution), and may use the random value to generate a synthetic knowledge graph (e.g., a synthetic adjacency matrix and/or a synthetic attribute matrix)")
wherein the plurality of synthetic graph datasets comprise structured-graph data;([¶0013] "A knowledge graph includes a set of nodes and interconnections between nodes. The nodes may represent the same type of entity (e.g., people), or a node may have a node type that indicates a type of entity represented by the node. A node may be associated with a set of attributes that define characteristics of the node (e.g., height, weight, and gender for a node that represents a person). Interconnections between nodes may be referred to as edges. In some cases, edges may have different edge types that represent different relationships between nodes.")
training, by the computing system, a plurality of graph models with at least a subset of the plurality of synthetic graph datasets to generate a plurality of trained graph models;([¶0031] " the synthetic data modeling system may train a set of teacher models (e.g., machine learning models) using the true data partitions of the true knowledge graph and the synthetic data partitions of the synthetic knowledge graph, which may be generated by the synthetic data modeling system as described above")
obtaining, by the computing system, one or more inputs associated with a particular task from the plurality of synthetic graph datasets;([¶0073] "FIG. 3 is a diagram illustrating an example 300 of applying a trained machine learning model to a new observation associated with differentially private dataset generation and modeling for knowledge graphs. ")
processing, by the computing system, one or more inputs from the plurality of synthetic graph datasets with the plurality of trained graph models to generate a plurality of graph outputs;([¶0073] "FIG. 3 is a diagram illustrating an example 300 of applying a trained machine learning model to a new observation associated with differentially private dataset generation and modeling for knowledge graphs. ")
generating, by the computing system, an evaluation representation associated with the plurality of graph models based on the plurality of graph outputs; (FIG. 6 step 640 shows that the results of the teacher models are aggregated l step 650 shows that the results and the output of the student model which is also a trained graph model, is compared based on a classification loss, thus corresponding to “a comparison between the plurality of graph outputs”; step 655 shows that based on a condition, “output…a model used to generate the synthetic knowledge graph”; [¶0050] “…. the synthetic data modeling system may perform the statistical analysis on the final synthetic knowledge graph, and may output a result of performing the statistical analysis. For example, the synthetic data modeling system may output the result for display, may output the result to another device (e.g., a user device) for display by the other device, and/or the like.”; “statistical analysis” corresponds to “generating an evaluation representation” “statistical analysis” corresponds to “generating an evaluation representation”)
providing, by the computing system, the evaluation representation for display.([¶0050] “…. the synthetic data modeling system may perform the statistical analysis on the final synthetic knowledge graph, and may output a result of performing the statistical analysis. For example, the synthetic data modeling system may output the result for display, may output the result to another device (e.g., a user device) for display by the other device, and/or the like.”;)
determining, by the computing system, a particular graph model of the plurality of graph models based on the evaluation representation ([¶0071] "The machine learning model may compare the performance scores for each machine learning model, and may select the machine learning model with the best (e.g., highest accuracy, lowest error, closest to a desired threshold, and/or the like) performance score as the trained machine learning model 245.")
and storing, by the computing system, data associated with the particular graph model with data associated with the particular task.([¶0070] "If the machine learning model performs adequately (e.g., with a performance score that satisfies a threshold), then the machine learning system may store that machine learning model as a trained machine learning model 245 to be used to analyze new observations, as described below in connection with FIG. 3." [¶0082] "A storage component 409 includes a hard disk or another type of storage device that stores information, data, and/or software (e.g., code, instructions, and/or the like) related to the operation and use of the synthetic data modeling system 401").
However, Zheng does not explicitly teach wherein each sampled configuration is generated by sampling a generator parameter set of a generator parameter space,
obtaining, by the computing system, a plurality of graph models, wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises differing model architectures.
You, in the same field of endeavor, teaches wherein each sampled configuration is generated by sampling a generator parameter set of a generator parameter space, ([p. 5 §5.1] "we first sample D random GNN designs from the design space. Then, we apply these designs to a fixed set of GNN tasks, and record each GNN’s average performance across the tasks")
obtaining, by the computing system, a plurality of graph models, wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises differing model architectures([Abstract] "current research focuses on proposing and evaluating specific architectural designs of GNNs, such as GCN,GIN, or GAT [...] we define and systematically study the architectural design space for GNNs which consists of 315,000 different designs over 32 different predictive tasks" [p. 2] "The GNN task space with a task similarity metric allows us to identify novel tasks and effectively transfer GNN architectural designs between similar tasks").
Zheng as well as You are directed towards dataset generation and graph modeling. Therefore, Zheng as well as You are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention implement Zheng's teacher/student models as a diverse set of GNN architectures described by You. You provides as additional motivation for combination ([p. 8] “The best model in our design space is better than the best GCN model in 24 out of 32 tasks”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 16, the combination of Zheng, and You teaches The method of claim 15, wherein the evaluation representation comprises evaluation data descriptive of node classification for the plurality of graph models.(You [p. 2] "We consider 32 tasks consisting of 12 synthetic node classification tasks, 8 synthetic graph classification tasks, and 6 real-world node classification and 6 graph classification tasks" [p. 5] "the exploration of GNN design space can be efficiently conducted. We focus on node and graph level tasks; results for link prediction are in the Appendix").
Regarding claim 17, the combination of Zheng, and You teaches The method of claim 15, wherein the evaluation representation comprises evaluation data descriptive of link prediction for the plurality of graph models.(You [p. 2] "We consider 32 tasks consisting of 12 synthetic node classification tasks, 8 synthetic graph classification tasks, and 6 real-world node classification and 6 graph classification tasks" [p. 5] "the exploration of GNN design space can be efficiently conducted. We focus on node and graph level tasks; results for link prediction are in the Appendix").
Regarding claim 18, Zheng teaches One or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more computing devices, cause the one or more computing devices, cause the one or more computing devices to perform operations, the operations comprising:([¶0082] "A processor 407 includes a central processing unit, a graphics processing unit, and/or the like. A memory 408 includes random-access memory, read-only memory, and/or the like. The memory 408 may store a set of instructions (e.g., one or more instructions) for execution by the processor 407. The processor 407 may execute the set of instructions to perform one or more operations or processes described herein. In some implementations, execution of the set of instructions, by one or more processors 407, causes the one or more processors 407 and/or the synthetic data modeling system 401 to perform one or more operations or processes described herein.")
obtaining input data associated with a user;([¶0049] "a user may interact with a device (e.g., a user device) to provide input to the generator model (e.g., a dataset or true knowledge graph). The generator model may process the input to generate an output (e.g., a synthetic knowledge graph), and may output the result (e.g., to the device).”; [¶0032] “The teacher model may be trained, using the labeled inputs, to differentiate between (or determine probability or confidence scores for) true data and synthetic data by applying a machine learning algorithm, as described in more detail below in connection with FIG. 2. Each teacher model may be trained in this manner, either using the same machine learning model or different machine learning models for training on respective partitions.”)
generating, by one or more generators, a plurality of synthetic graph datasets based at least in part on the input data, ([¶0034] " the generator may be referred to as a graph generator model." [¶0048] " the generator model may select a random (or latent) value (e.g., using a Gaussian distribution), and may use the random value to generate a synthetic knowledge graph (e.g., a synthetic adjacency matrix and/or a synthetic attribute matrix)")
wherein the plurality of synthetic graph datasets comprise structured-graph data;([¶0013] "A knowledge graph includes a set of nodes and interconnections between nodes. The nodes may represent the same type of entity (e.g., people), or a node may have a node type that indicates a type of entity represented by the node. A node may be associated with a set of attributes that define characteristics of the node (e.g., height, weight, and gender for a node that represents a person). Interconnections between nodes may be referred to as edges. In some cases, edges may have different edge types that represent different relationships between nodes.")
training a plurality of graph models with at least a subset of the plurality of synthetic graph datasets to generate a plurality of trained graph models;([¶0031] " the synthetic data modeling system may train a set of teacher models (e.g., machine learning models) using the true data partitions of the true knowledge graph and the synthetic data partitions of the synthetic knowledge graph, which may be generated by the synthetic data modeling system as described above")
processing one or more inputs from the plurality of synthetic graph datasets with the plurality of trained graph models to generate a plurality of graph outputs;([¶0073] "FIG. 3 is a diagram illustrating an example 300 of applying a trained machine learning model to a new observation associated with differentially private dataset generation and modeling for knowledge graphs. ")
generating an output representation associated with the plurality of graph models based on the plurality of graph outputs; and([¶0073] "FIG. 3 is a diagram illustrating an example 300 of applying a trained machine learning model to a new observation associated with differentially private dataset generation and modeling for knowledge graphs. ")
providing the output representation for display.([¶0050] “…. the synthetic data modeling system may perform the statistical analysis on the final synthetic knowledge graph, and may output a result of performing the statistical analysis. For example, the synthetic data modeling system may output the result for display, may output the result to another device (e.g., a user device) for display by the other device, and/or the like.”;).
However, Zheng does not explicitly teach obtaining a plurality of graph models, wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises a variety of graph model architectures;.
You, in the same field of endeavor, teaches obtaining a plurality of graph models, wherein the plurality of graph models comprises a diverse selection of graph models, wherein the diverse selection of graph models comprises a variety of graph model architectures;([Abstract] "current research focuses on proposing and evaluating specific architectural designs of GNNs, such as GCN,GIN, or GAT [...] we define and systematically study the architectural design space for GNNs which consists of 315,000 different designs over 32 different predictive tasks" [p. 2] "The GNN task space with a task similarity metric allows us to identify novel tasks and effectively transfer GNN architectural designs between similar tasks").
Zheng as well as You are directed towards dataset generation and graph modeling. Therefore, Zheng as well as You are reasonably pertinent analogous art. It would have been obvious before the effective filing date of the claimed invention implement Zheng's teacher/student models as a diverse set of GNN architectures described by You. You provides as additional motivation for combination ([p. 8] “The best model in our design space is better than the best GCN model in 24 out of 32 tasks”). This motivation for combination also applies to the remaining claims which depend on this combination.
Regarding claim 19, the combination of Zheng, and You teaches The one or more non-transitory computer-readable media of claim 18, wherein the output representation comprises a graphical depiction of a feature center distance based on the plurality of graph outputs associated with the plurality of graph models.(You [p. 5] "The proposed task similarity metric consists of two components: (1) selection of anchor models and (2) measuring the rank distance of the performance of anchor models." See FIG. 5 graphical depiction of task similarities).
Regarding claim 20, the combination of Zheng, and You teaches The one or more non-transitory computer-readable media of claim 18, wherein the output representation comprises vector graph statistic data and hyperparameter evaluation data.(Zheng [¶0050] “…. the synthetic data modeling system may perform the statistical analysis on the final synthetic knowledge graph, and may output a result of performing the statistical analysis. For example, the synthetic data modeling system may output the result for display, may output the result to another device (e.g., a user device) for display by the other device, and/or the like.”).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Gao (“Graph Neural Architecture Search”, 2020) is directed towards a neural architecture search for graph neural networks.
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SIDNEY VINCENT BOSTWICK whose telephone number is (571)272-4720. The examiner can normally be reached M-F 7:30am-5:00pm EST.
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, Miranda Huang can be reached on (571)270-7092. 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.
/SIDNEY VINCENT BOSTWICK/Examiner, Art Unit 2124
Prosecution Insights