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 .
Priority
Applicant claims the benefit of prior-filed U.S. Provisional Application No. 63/406,777 filed on September 15, 2022, which is acknowledged.
Drawings
The drawings were received on 01/30/2023. These drawings are acceptable.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on the following date(s): 04/24/2023 has been considered by the examiner.
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 1-15 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e. an abstract idea) without significantly more.
Claim 1: Dose claim fall within a statutory category? Yes
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
and generating the prediction (Considered directed to a Mental Process: Making evaluations and judgements of observations for formulating observations, evaluations and judgements as claimed; see MPEP § 2106.04(a)(2), subsection III ; and1) Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
obtaining graph structured data from a technical application domain of the machine learning system; (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation is directed to insignificant solution activity, e.g. receiving or transmitting data over a network)
generating the prediction in the machine learning system… (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
training a graph neural network to learn logical rules using message passing; (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation merely include instructions to implement an abstract idea on a computer, or merely use a computer as a tool to perform an abstract idea; Thus claim limitations amount to mere instructions to apply the judicial exception using a computer/computing environment as a tool, as discussed in MPEP § 2106.05(f).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception.
Secondly, the limitations directed to insufficient to transform the judicial exception to a patentable invention because the recitation is directed to insignificant solution activity for as noted above.
The courts have deemed these types of activity as well-known routine and convectional, see evidences noted below:
Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network); but see DDR Holdings, LLC v. Hotels.com, L.P., 773 F.3d 1245, 1258, 113 USPQ2d 1097, 1106 (Fed. Cir. 2014) ("Unlike the claims in Ultramercial, the claims at issue here specify how interactions with the Internet are manipulated to yield a desired result‐‐a result that overrides the routine and conventional sequence of events ordinarily triggered by the click of a hyperlink." (emphasis added));
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 2: Dose claim fall within a statutory category? Yes:
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
Abstract idea from claim 1.
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
further comprising obtaining an initial set of logical (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation is directed to insignificant solution activity, e.g. receiving or transmitting data over a network)
rules that are usable to solve a satisfiability problem in the technical application domain of the machine learning application, … (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
wherein the graph neural network is trained to learn updates to the initial set of logical rules to provide new learned rules; (Deemed insufficient to transform the judicial exception to a patentable invention because the limitation generically recites an effect of the judicial exception or claims every mode of accomplishing that effect, amounts to a claim that is merely adding the words "apply it" to the judicial exception,, as discussed in MPEP § 2106.05(f).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application.
Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception.
Secondly, the limitations directed to insufficient to transform the judicial exception to a patentable invention because the recitation is directed to insignificant solution activity for as noted above.
The courts have deemed these types of activity as well-known routine and convectional, see evidences noted below:
Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network); but see DDR Holdings, LLC v. Hotels.com, L.P., 773 F.3d 1245, 1258, 113 USPQ2d 1097, 1106 (Fed. Cir. 2014) ("Unlike the claims in Ultramercial, the claims at issue here specify how interactions with the Internet are manipulated to yield a desired result‐‐a result that overrides the routine and conventional sequence of events ordinarily triggered by the click of a hyperlink." (emphasis added));
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 3: Dose claim fall within a statutory category? Yes:
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
Abstract idea noted in claim 2.
wherein the initial set of logical rules are predefined using domain knowledge (Considered directed to a Mental Process: Making evaluations and judgements of observations for formulating observations, evaluations and judgements as claimed; see MPEP § 2106.04(a)(2), subsection III)
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
Alternatively, wherein the initial set of logical rules are predefined using domain knowledge. (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, the additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 4: Dose claim fall within a statutory category? Yes
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
further comprising computing an attention bit that is used to decide whether a feature of a node of the graph (Considered directed to a Mental Process: Making evaluations and judgements of observations for formulating observations, evaluations and judgements as claimed; see MPEP § 2106.04(a)(2), subsection III)
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
(Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, the additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 5: Dose claim fall within a statutory category? Yes
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
wherein a plurality of attention bits are computed, each for a respective node of the graph neural network, and wherein the nodes are ordered prior to aggregation by the message passing based on the attention bits ( Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
(Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, the additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 6: Dose claim fall within a statutory category? Yes.
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
wherein the training is performed end-to-end with a differentiable satisfiability solver (Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, the additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 7: Dose claim fall within a statutory category? Yes
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
Abstract idea recited in claims 1.
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
wherein the training is performed using reinforcement learning. (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, the additional limitations are directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 8: Dose claim fall within a statutory category? Yes.
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
further comprising generating two graph sequences from the graph structured data by dropping edges or nodes randomly, (Considered directed to a Mental Process: Making evaluations and judgements of observations for formulating observations, evaluations and judgements as claimed; see MPEP § 2106.04(a)(2), subsection III) ;. (Considered directed to a Mental Process: Making evaluations and judgements of observations for formulating observations, evaluations and judgements as claimed; see MPEP § 2106.04(a)(2), subsection III ; and Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
wherein the graph neural network generates a representation for each of the graph sequences (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation merely include instructions to implement an abstract idea on a computer, or merely use a computer as a tool to perform an abstract idea; Thus claim limitations amount to mere instructions to apply the judicial exception using a computer/computing environment as a tool, as discussed in MPEP § 2106.05(f).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use and merely invoke the use of computer technology as a tool for applying the judicial exception.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 9: Dose claim fall within a statutory category? Yes.
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
wherein the loss is built by minimizing a Kullback-Leibler (KL) divergence of the representation, by maximizing mutual information and/or using a cosine similarity function. (Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 10: Dose claim fall within a statutory category? Yes.
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
further comprising ordering nodes of the graph neural network prior to aggregation by the message passing. (Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application. Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 11: Dose claim fall within a statutory category? Yes.
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
wherein the ordering is based on values of the features of the node (Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application.
Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 12: Dose claim fall within a statutory category? Yes.
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
(Mathematical concepts – mathematical relationships, (see MPEP § 2106.04(a)(2), subsection I))
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
wherein the message passing performed by a node of the graph neural network… (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application.
Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Claim 13: Dose claim fall within a statutory category? Yes.
Step 2A Prong 1: Evaluate whether the claim recites a judicial exception.
Abstract idea noted in claim 1.
Step 2A Prong 2: Evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception
The preamble is deemed insufficient to transform the judicial exception to a patentable invention because the preamble generally links the use of a judicial exception to a particular technological environment or field of use, see MPEP 2106.05(h).
wherein the technical application domain is in medical artificial intelligence, bioinformatics and/or knowledge graphs, and wherein the prediction is an output of the graph neural network trained on a machine learning task that is a node classification, a link prediction and/or a graph classification. (Deemed insufficient to transform the judicial exception to a patentable invention because the recitation simply link the judicial exception to a field of use and/or technology environment, see MPEP 2106.05(h).)
The additional elements do not appear to be sufficient to transform the judicial exception into a practical application at Step 2A as analyzed above.
Step 2B: Evaluates whether the claim as a whole/in combination integrates the recited judicial exception into a practical application of the exception
The claim does not include additional elements that are sufficient to amount to significantly more that the judicial exception and fail to integrate the abstract into practical application.
Specifically, first, the additional limitations directed to elements that generally link the use of a judicial exception to a particular technological environment or field of use.
These types of claimed elements cannot transform the judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. This claim is not patent eligible.
Regarding claims 14-15, The claim limitations are similar to those in claim 1 and thus rejected under the same rationale.
As shown above, claims 1-15 are rejected under 35 U.S.C. 101 because the claimed invention is directed a judicial exception and does not recite, when claim elements are examined individually and as a whole, elements that the courts have identified as "significantly more” than the recited judicial exception. The claims are therefore directed to an abstract idea.
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.
Claims 1, 7, and 10-15 are rejected under 35 U.S.C. 103 as being unpatentable over Ramamurthy et al. (US 20240028868, hereinafter ‘Ram’) in view of Orhan et al. (US 20220124543, hereinafter ‘Or’).
Regarding independent claim 1, Ram teaches a method for learning logical rules over graph structured data to generate a prediction in a machine learning system, the method comprising: (in [0005] A system provides logic rule induction on knowledge graphs of engineering systems by a first framework for searching disconnected knowledge graphs and a second framework for searching well connected knowledge graphs [a method for learning logical rules over graph structured data to generate a prediction in a machine learning system, the method comprising]. In the first framework, top ranked candidates of first-order logic rule formulas are generated to reduce the search space of knowledge graphs as a formula building process searches for longer formulas. The second framework applies a graph neural network (GNN) with a counterfactual solver engine to capture local topology patterns of knowledge graphs and to abstract first-order logic rule formulas based on atomic actions to the graphs. The induced first-order logic rules explain an optimum design for the engineering system.)
obtaining graph structured data from a technical application domain of the machine learning system; (in [0005] A system provides logic rule induction on knowledge graphs of engineering systems [obtaining graph structured data from a technical application domain of the machine learning system] by a first framework for searching disconnected knowledge graphs and a second framework for searching well connected knowledge graphs. In the first framework, top ranked candidates of first-order logic rule formulas are generated to reduce the search space of knowledge graphs as a formula building process searches for longer formulas…)
training a graph neural network to learn logical rules using message passing; and generating the prediction in the machine learning system based on the learned logical rules. (in Abstract: System and method for logic rule formula induction on knowledge graphs for engineering system designs include receiving plurality of knowledge graphs for an engineering system... Top ranked formulas are selected from the candidate formulas according to defined criteria. For well-connected graphs, a graph neural network is trained [training a graph neural network to learn logical rules using message passing] to predict first class for a query graph and second class for distractor graph [and generating the prediction in the machine learning system based on the learned logical rules]. Counterfactual solver engine solves for minimum number of edits to query graph toward distractor graph which transforms predicted first class of the query graph to predicted second class. )
Examiner notes that graphs can be interpreted as logical rules among nodes learned for making predictions/modeled outcomes and that training a graph neural network learns using massage passing with adjacent nodes to learn the linked relationships (e.g. logical rules).
Or expressly teaches a graph neural network learns using massage passing with adjacent nodes to learn the linked relationships, in [0059] Graph Neural Networks (GNNs) are a framework to capture the dependence of nodes in graphs via message passing between the nodes [training a graph neural network to learn logical rules using message passing]. Unlike deep neural networks (DNNs), GNNs directly operate on a graph to represent information from its neighborhood with arbitrary hops. Additionally, GNNs learn embeddings of various graph attributes (e.g., nodes, edges, and global attributes/parameters)…; And in [0021] ... In embodiments, a graph neural networks (GNN) is used to model or otherwise represent a network, including heterogeneous RANs with multiple types of network elements/nodes such as UEs, central units (CUs), distributed units (DUs), and/or radio units (RUs) and/or multiple RAT types such as one or more WLAN APs, one or more cellular RAN nodes, and the like… Each network element/node in the network is represented as a node in the graph or GNN, and each interface is represented as an edge in the graph/GNN. Representing such a network as a graph allows relevant features to be extracted from network logical entities using GNN tools [training a graph neural network to learn logical rules using message passing] such as graph convolutional neural network (CNN), spatial-temporal neural network, and/or the like…
Or and Ram are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art of processing information from knowledge graph analysis using machine learning models and techniques, as disclosed by Or with the method of developing information retrieval and processing techniques using as logic rule induction from knowledge graph analysis and machine learning techniques, as disclosed by Ram.
One of ordinary skill in the arts would have been motivated to combine the methods disclosed by Or and Ram as noted above. Doing so enables techniques for representing a network as a graph that allows relevant features to be extracted from network logical entities using graph neural network (GNN) tools, (Or, 0021).
Regarding claim 7, the rejection of claim 1 is incorporated and Or further teaches the method according to claim 1, wherein the training is performed using reinforcement learning. (in [0070] In general, the GNN can be trained using different reinforcement learning (RL) techniques…)
Regarding claim 10, the rejection of claim 1 is incorporated and Ram in combination with Or teaches the method according to claim 1, further comprising ordering nodes of the graph neural network prior to aggregation. (in [0016] … Knowledge graphs 150 are the accumulation of design data exported from engineering applications 112 and generated by knowledge graph algorithm that processes an ontology of the exported data... The ontology also describes properties of the elements and the element relationships, and may organize the element types into hierarchies [further comprising ordering nodes of the graph neural network prior to aggregation as ordered graph hierarchies], such as super-types and sub-types. A knowledge graph 150 represents the ontology as nodes and edges that correspond to a set of elements of the ontology and element relationships, respectively… [0017] AI module 125 is configured to perform first-order logic formula induction on knowledge graphs 150 using a plurality of modules including a filter 121, a beam search engine 122, a dynamic formula generator 123, a formula evaluation engine 124, a counterfactual solver engine 127, and a graph neural network module 128. AI module 125 analyzes one or more knowledge graphs to induce first-order logic rules that express the optimum design [[further comprising ordering nodes of the graph neural network prior to aggregation as ordered logic rules]… The first-order logic formula to be derived consists of a chain of terms that may relate to engineering system components, such as for instance, an engine, chassis, axel, and wheel extracted from the knowledge graph for the mechanical interconnection domain... [0043] In an embodiment, the GNN is trained to learn the permutation matrix P, which enables determination of the minimum number of edits directly from the knowledge graph [further comprising ordering nodes of the graph neural network prior to aggregationas training of knowledge graph using message passing of ordered input graph].)
Or teaches the training of knowledge graphs using message passing, in in [0059] Graph Neural Networks (GNNs) are a framework to capture the dependence of nodes in graphs via message passing between the nodes [further comprising ordering nodes of the graph neural network prior to aggregation]. Unlike deep neural networks (DNNs), GNNs directly operate on a graph to represent information from its neighborhood with arbitrary hops. Additionally, GNNs learn embeddings of various graph attributes (e.g., nodes, edges, and global attributes/parameters)…; And in [0021] ... In embodiments, a graph neural networks (GNN) is used to model or otherwise represent a network, including heterogeneous RANs with multiple types of network elements/nodes such as UEs, central units (… Each network element/node in the network is represented as a node in the graph or GNN, and each interface is represented as an edge in the graph/GNN. Representing such a network as a graph allows relevant features to be extracted from network logical entities using GNN tools [further comprising ordering nodes of the graph neural network prior to aggregation by the message passing] such as graph convolutional neural network (CNN), spatial-temporal neural network, and/or the like…
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Or and Ram for the same reasons disclosed above.
Regarding claim 11, the rejection of claim 10 is incorporated and Ram in combination with Or teaches the method according to claim 10, wherein the ordering is based on values of the features of the nodes. (in [0016] … Knowledge graphs 150 are the accumulation of design data exported from engineering applications 112 and generated by knowledge graph algorithm that processes an ontology of the exported data... The ontology also describes properties of the elements and the element relationships, and may organize the element types into hierarchies [wherein the ordering is based on values of the features of the nodes as ordered graph hierarchies of feature knowledge/ontology element values], such as super-types and sub-types. A knowledge graph 150 represents the ontology as nodes and edges that correspond to a set of elements of the ontology and element relationships, respectively… [0017] AI module 125 is configured to perform first-order logic formula induction on knowledge graphs 150 using a plurality of modules including a filter 121, a beam search engine 122, a dynamic formula generator 123, a formula evaluation engine 124, a counterfactual solver engine 127, and a graph neural network module 128. AI module 125 analyzes one or more knowledge graphs to induce first-order logic rules that express the optimum design [wherein the ordering is based on values of the features of the nodes]… The first-order logic formula to be derived consists of a chain of terms [wherein the ordering is based on values of the features of the nodes] that may relate to engineering system components, such as for instance, an engine, chassis, axel, and wheel extracted from the knowledge graph for the mechanical interconnection domain...
Or teaches the training of knowledge graphs using message passing, in in [0059] Graph Neural Networks (GNNs) are a framework to capture the dependence of nodes in graphs via message passing between the nodes. Unlike deep neural networks (DNNs), GNNs directly operate on a graph to represent information from its neighborhood with arbitrary hops. Additionally, GNNs learn embeddings of various graph attributes (e.g., nodes, edges, and global attributes/parameters) [wherein the ordering is based on values of the features of the nodes]…; And in [0021] ... In embodiments, a graph neural networks (GNN) is used to model or otherwise represent a network, including heterogeneous RANs with multiple types of network elements/nodes such as UEs, central units (… Each network element/node in the network is represented as a node in the graph or GNN, and each interface is represented as an edge in the graph/GNN. Representing such a network as a graph allows relevant features [wherein the ordering is based on values of the features of the nodes] to be extracted from network logical entities using GNN tools such as graph convolutional neural network (CNN), spatial-temporal neural network, and/or the like…
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Or and Ram for the same reasons disclosed above.
Regarding claim 12, the rejection of claim 1 is incorporated and Ram in combination with Or teaches the method according to claim 1, wherein . (in [0036] FIG. 5 shows an example of applying counterfactual with a GNN to induce logic rule formulas [wherein ] from large scale well-connected knowledge graphs in accordance with embodiments of this disclosure. In an embodiment, counterfactual solver 127 solves an optimization by minimizing the norm of binary gating vector a, which represents the minimum number of edits from graph G′ to query graph G to generate counterfactual graph G*. This can be expressed as follows:
Or teaches wherein the message passing performed by a node of the graph neural network uses the logical rules to aggregate information from other nodes and/or to transform information from the same node to another layer of the graph neural network. (in [0021] ... In embodiments, a graph neural networks (GNN) is used to model or otherwise represent a network, including heterogeneous RANs with multiple types of network elements/nodes such as UEs, central units (… Each network element/node in the network is represented as a node in the graph or GNN, and each interface is represented as an edge in the graph/GNN. Representing such a network as a graph allows relevant features [wherein the message passing performed by a node of the graph neural network uses the logical rules to aggregate information from other nodes] to be extracted from network logical entities using GNN tools such as graph convolutional neural network (CNN), spatial-temporal neural network, and/or the like… [0059] Graph Neural Networks (GNNs) are a framework to capture the dependence of nodes in graphs via message passing between the nodes [wherein the message passing performed by a node of the graph neural network uses the logical rules to aggregate information from other nodes]. Unlike deep neural networks (DNNs), GNNs directly operate on a graph to represent information from its neighborhood with arbitrary hops. Additionally, GNNs learn embeddings of various graph attributes (e.g., nodes, edges, and global attributes/parameters)...
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Or and Ram for the same reasons disclosed above.
Regarding claim 13, the rejection of claim 1 is incorporated and Ram in combination with Or teaches the method according to claim 1, wherein the technical application domain is in medical artificial intelligence, bioinformatics and/or knowledge graphs, (in [0002] The problem of learning first-order logic rules from data has been a long-standing challenge in machine learning and plays an important role in many applications. For example, for systems like gas turbines, electrical grid, or smart buildings, an enormous amount of data is recorded by sensors. For such systems, one can construct a knowledge graph to represent the domain knowledge of the system [wherein the technical application domain is ]…)
and wherein the prediction is an output of the graph neural network trained on a machine learning task that is a node classification, a link prediction and/or a graph classification. (in claim 5. The system of claim 1, wherein the filtering module determines that a knowledge graph satisfies a connectedness criteria and triggering a logic rule formula induction as a connected knowledge graph, the system further comprising: a graph neural network trained to predict a first class of a query graph and to predict a second class for a distractor graph [wherein the prediction is an output of the graph neural network trained on a machine learning task that is a node classification, ]; …)
Regarding independent claims 14 and 15, Ram teaches a system for learning logical rules over graph structured data to generate a prediction in a machine learning system, comprising one or more hardware processors, configured to provide for execution of the following steps: and a tangible, non-transitory computer-readable medium having instructions thereon which, upon being executed by one or more processors, provides for execution of a method for learning logical rules over graph structured data to generate a prediction in a machine learning system, the method comprising following steps: (in [0005] A system provides logic rule induction on knowledge graphs of engineering systems by a first framework for searching disconnected knowledge graphs and a second framework for searching well connected knowledge graphs [system for learning logical rules over graph structured data to generate a prediction in a machine learning system … a method for learning logical rules over graph structured data to generate a prediction in a machine learning system]. In the first framework, top ranked candidates of first-order logic rule formulas are generated to reduce the search space of knowledge graphs as a formula building process searches for longer formulas. The second framework applies a graph neural network (GNN) with a counterfactual solver engine to capture local topology patterns of knowledge graphs and to abstract first-order logic rule formulas based on atomic actions to the graphs. The induced first-order logic rules explain an optimum design for the engineering system.; And in [0053] As stated above, the computer system 710 may include at least one computer readable medium or memory for holding instructions programmed according to embodiments of the invention and for containing data structures, tables, records, or other data described herein. The term “computer readable medium” as used herein refers to any medium that participates in providing instructions to the processors 720 for execution [comprising one or more hardware processors, configured to provide for execution of the following steps…]. A computer readable medium may take many forms including, but not limited to, non-transitory, non-volatile media, volatile media, and transmission media [a tangible, non-transitory computer-readable medium having instructions thereon which, upon being executed by one or more processors, provides for execution of a method … the method comprising following steps]… [0059] It should further be appreciated that the computer system 710 may include alternate and/or additional hardware, software, or firmware components beyond those described or depicted without departing from the scope of the disclosure. More particularly, it should be appreciated that software, firmware, or hardware components depicted as forming part of the computer system 710 [comprising one or more hardware processors, configured to provide for execution of the following steps…And a tangible, non-transitory computer-readable medium having instructions thereon which, upon being executed by one or more processors, provides for execution of a method … the method comprising following steps] are merely illustrative and that some components may not be present or additional components may be provided in various embodiments... Moreover, while certain modules may be depicted and described as sub-modules of another module, in certain embodiments, such modules may be provided as independent modules or as sub-modules of other modules.)
The remaining limitations are similar with claim 1 limitations and are rejected under the same rationale.
Claims 2 and 3 are rejected under 35 U.S.C. 103 as being unpatentable over Ramamurthy et al. (US 20240028868, hereinafter ‘Ram’) in view of Orhan et al. (US 20220124543, hereinafter ‘Or’) in view further view of Liu et al. (NPL: Learning the Satisfiability of Pseudo-Boolean Problem with Graph Neural Networks, hereinafter ‘Liu’).
Regarding claim 2, the rejection of claim 1 is incorporated and Ram in combination with Or teaches the method according to claim 1, further comprising obtaining an initial set of logical rules that are usable to solve a satisfiability problem in the technical application domain of the machine learning application, (in [0018] Briefly, the filter 121 is used to determine knowledge graphs that are disconnected for performing formula induction according to a first process of this disclosure [further comprising obtaining an initial set of logical rules …]. For example, the filter 121 may detect distinct clusters in a knowledge graph and allow such a knowledge graph to pass through as a disconnected knowledge graph… To evaluate the generated formulas, the formula evaluation engine 124 maps the formulas according to edge type, forming sets of subgraphs and finds set intersections to determine which subgraphs satisfy the candidate formula being evaluated [that are usable to solve a satisfiability problem in the technical application domain of the machine learning application]. And in [0033] Given a query knowledge graph G for which the GNN 128 predicts class c, an objective is to produce a counterfactual explanation that finds minimum changes to graph G [that are usable to solve a satisfiability problem in the technical application domain of the machine learning application], looking to changes towards a distractor graph G′ which the GNN previously predicted as class c′. The solution is to perform a transformation from G to counterfactual G* such that G* appears to be an instance of class c′ to a trained GNN model g….)
wherein the graph neural network is trained to learn updates to the initial set of logical rules to provide new learned rules. (in [0043] In an embodiment, the GNN is trained to learn the permutation matrix P, which enables determination of the minimum number of edits directly from the knowledge graph [wherein the graph neural network is trained to learn updates to the initial set of logical rules to provide new learned rules]. This approach results in a faster processing algorithm that enables the training of the system in an end-to-end fashion. The result of this approach is the generation of a GNN that can interpret the learned parameters to discover human readable logic formulas on the large-scale knowledge graph. Since the GNN can capture local topology patterns in the graph, the knowledge embedded in the learned model can be abstracted and generalized to logic formulas [wherein the graph neural network is trained to learn updates to the initial set of logical rules to provide new learned rules].)
Additionally, Or teaches in [0508] The term “optimization” at least in some embodiments refers to an act, process, or methodology of making something (e.g., a design, system, or decision) as fully perfect, functional, or effective as possible. Optimization usually includes mathematical procedures such as finding the maximum or minimum of a function. The term “optimal” at least in some embodiments refers to a most desirable or satisfactory end, outcome, or output…; And in Abstract: The present disclosure provides connection management techniques based on graph neural networks (GNN) and deep reinforcement learning (DRL) to optimize user association and load balancing. A graph structure of a communication network is considered for the GNN architecture and DRL is used to learn parameters of the GNN algorithm/model [wherein the graph neural network is trained to learn updates to the initial set of logical rules to provide new learned rules]. Connection management is defined as a combinatorial graph optimization problem [further comprising obtaining an initial set of logical rules that are usable to solve a satisfiability problem in the technical application domain of the machine learning application],… [0496] The term “graph neural network” or “GNN” at least in some embodiments refers to a class of ANNs for processing data represented by graph data structures. Additionally or alternatively, the term “graph neural network” or “GNN” at least in some embodiments refers to an optimizable transformation on one or more attributes of a graph data structure [wherein the graph neural network is trained to learn updates to the initial set of logical rules to provide new learned rules](e.g., nodes, edges, global-context) that preserves graph symmetries (e.g., permutation invariances)…
Furthermore Liu teaches, in Fig. 1:
PNG
media_image1.png
260
1152
media_image1.png
Greyscale
Fig.1. The model for solving decision pseudo-Boolean problem based on graph neural network. The pipeline is as follows: (a) A set of constraints is given as input in which all variables [obtaining an initial set of logical rules that are usable to solve a satisfiability problem in the technical application domain of the machine learning application] can be assigned True or False, then (b) some normalization rules are applied to transform the constraints into normalized form, which is done by equivalent transformation and removing tautologies. From the normalized constraints we can (c) construct a weighted bipartite graph to represent the topological relationship between variables and constraints [obtaining an initial set of logical rules that are usable to solve a satisfiability problem in the technical application domain of the machine learning application]. Each node in the graph is represented as an embedding vector, which is updated [wherein the graph neural network is trained to learn updates to the initial set of logical rules to provide new learned rules] iteratively through (d) the message passing mechanism. Finally, the model (e) outputs prediction about the satisfiability [… logical rules that are usable to solve a satisfiability problem in the technical application domain of the machine learning application].
Liu, Or and Ram are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art of processing information from knowledge graph analysis using machine learning models and techniques, as disclosed by Liu with the method of developing information retrieval and processing techniques using as logic rule induction from knowledge graph analysis and machine learning techniques, as collectively disclosed by Or and Ram.
One of ordinary skill in the arts would have been motivated to combine the methods disclosed by Liu, Or and Ram as noted above. Doing so enables techniques for applying GNN models to solve constraint satisfaction and combinatorial optimization problems using non-zero constant terms, (Liu, Sec. 3.2).
Regarding claim 3, the rejection of claim 2 is incorporated and Ram in combination with Or and Liu teaches the method according to claim 2, wherein the initial set of logical rules are predefined using domain knowledge. (in [0018] Briefly, the filter 121 is used to determine knowledge graphs that are disconnected for performing formula induction according to a first process of this disclosure [wherein the initial set of logical rules are predefined using domain knowledge]. For example, the filter 121 may detect distinct clusters in a knowledge graph and allow such a knowledge graph to pass through as a disconnected knowledge graph… To evaluate the generated formulas, the formula evaluation engine 124 maps the formulas according to edge type, forming sets of subgraphs and finds set intersections to determine which subgraphs satisfy the candidate formula being evaluated [wherein the initial set of logical rules are predefined using domain knowledge]. And in [0033] Given a query knowledge graph G for which the GNN 128 predicts class c, an objective is to produce a counterfactual explanation that finds minimum changes to graph G, looking to changes towards a distractor graph G′ which the GNN previously predicted as class c′. The solution is to perform a transformation from G to counterfactual G* such that G* appears to be an instance of class c′ to a trained GNN model g….)
Additionally, Or teaches wherein the initial set of logical rules are predefined using domain knowledge, in [0508] The term “optimization” at least in some embodiments refers to an act, process, or methodology of making something (e.g., a design, system, or decision) as fully perfect, functional, or effective as possible… The term “optimal” at least in some embodiments refers to a most desirable or satisfactory end, outcome, or output…; And in Abstract: The present disclosure provides connection management techniques based on graph neural networks (GNN) and deep reinforcement learning (DRL) to optimize user association and load balancing. A graph structure of a communication network [wherein the initial set of logical rules are predefined using domain knowledge] is considered for the GNN architecture and DRL is used to learn parameters of the GNN algorithm/model. Connection management is defined as a combinatorial graph optimization problem,… [0496] The term “graph neural network” or “GNN” at least in some embodiments refers to a class of ANNs for processing data represented by graph data structures. Additionally or alternatively, the term “graph neural network” or “GNN” at least in some embodiments refers to an optimizable transformation on one or more attributes of a graph data structure (e.g., nodes, edges, global-context) that preserves graph symmetries (e.g., permutation invariances)…
And Liu teaches wherein the initial set of logical rules are predefined using domain knowledge, in Fig.1. The model for solving decision pseudo-Boolean problem based on graph neural network. The pipeline is as follows: (a) A set of constraints is given as input in which all variables [wherein the initial set of logical rules are predefined using domain knowledge] can be assigned True or False, then (b) some normalization rules are applied to transform the constraints into normalized form, which is done by equivalent transformation and removing tautologies. From the normalized constraints we can (c) construct a weighted bipartite graph to represent the topological relationship between variables and constraints. Each node in the graph is represented as an embedding vector, which is updated iteratively through (d) the message passing mechanism. Finally, the model (e) outputs prediction about the satisfiability.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Liu, Or and Ram for the same reasons disclosed above.
Claims 4 and 5 are rejected under 35 U.S.C. 103 as being unpatentable over Ramamurthy et al. (US 20240028868, hereinafter ‘Ram’) in view of Orhan et al. (US 20220124543, hereinafter ‘Or’) in view further view of Narasimha et al. (US 20230189317, hereinafter ‘Nara’).
Regarding claim 4, the rejection of claim 1 is incorporated and Ram in combination with Or teaches the method according to claim 1, further comprising computing. (in [0017] AI module 125 is configured to perform first-order logic formula induction on knowledge graphs 150 using a plurality of modules including a filter 121, a beam search engine 122, a dynamic formula generator 123, a formula evaluation engine 124, a counterfactual solver engine 127, and a graph neural network module 128. AI module 125 analyzes one or more knowledge graphs to induce first-order logic rules that express the optimum design. In an embodiment, induction of first-order logic rule formulas involves deriving a formula that is a chain of terms that represent component relationships of a system design… )
Nara teaches in [0139] As an exemplary representation of the attention scores, the input data 312 may include an attention matrix including the plurality of attention scores. The attention matrix may be a weighted version of the simple adjacency matrix with only 0s and 1s [further comprising computing an attention bit that is used to decide whether a feature of a node of the graph neural network is included in the message passing]. Illustratively, a matrix element i,j of the attention matrix may represent a weighted interaction between node i and node j of the graph representation 322 [further comprising computing an attention bit that is used to decide whether a feature of a node of the graph neural network is included in the message passing]… [0141] Going back to FIG. 3B, the processing via trained graph neural network model 314 may be illustratively described as a “graph-in, graph-out” architecture, in which the attributes of a graph representation 322 (e.g., of each sub-band graph representation), such as the input feature vectors, the attention scores, etc. are updated by passing through a plurality of graph neural network layers 324 [further comprising computing an attention bit that is used to decide whether a feature of a node of the graph neural network is included in the message passing], to provide an output graph representation 326 (e.g., a plurality of output sub-band graph representations) with updated attributes (e.g., updated feature vectors, updated attention scores, etc.)… [0143] As an example, the processor 302 may be configured to instruct user scheduling of the plurality of wireless communication devices 306 based on the edges 334o of the last graph neural network layer 324 of the trained graph neural network model 314 (illustratively, the edges 334o of the output graph representation 326), e.g. based on the attention scores associated with the edges [further comprising computing an attention bit that is used to decide whether a feature of a node of the graph neural network is included in the message passing] 334o of the last graph neural network layer of the trained graph neural network model 314. Illustratively, the attention scores of the edges 334o of the output graph representation 326 may define the order of the wireless communication devices 306 for scheduling, e.g. with wireless communication devices 306 associated with greater scores (illustrated as thicker edges 334o) having a greater scheduling priority.
Nara, Or and Ram are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art of processing information from knowledge graph analysis using machine learning models and techniques, as disclosed by Nara with the method of developing information retrieval and processing techniques using as logic rule induction from knowledge graph analysis and machine learning techniques, as collectively disclosed by Or and Ram.
One of ordinary skill in the arts would have been motivated to combine the methods disclosed by Liu, Or and Ram as noted above. Doing so enables techniques for applying GNN models to solve constraint satisfaction and combinatorial optimization problems using non-zero constant terms, (Liu, Sec. 3.2).
Regarding claim 5, the rejection of claim 4 is incorporated and Nara further teaches the method according to claim 4, wherein a plurality of attention bits are computed, each for a respective node of the graph neural network, and wherein the nodes are ordered prior to aggregation by the message passing based on the attention bits. (in [0153] Illustratively, the layer function (Z.sub.s.sup.l) of a graph neural network layer may include an attention matrix (A.sub.s.sup.l) [wherein a plurality of attention bits are computed, each for a respective node of the graph neural network for each node in a network layer where the layer is considered the claimed order wherein the nodes are ordered prior to aggregation by the message passing based on the attention bits], a combined feature vector (F.sub.s.sup.l), and a correlation matrix (C.sub.s)… Furthermore, the layer function may include first (learnable) weights (W.sub.a.sup.l) associated with the attention matrix and second (learnable) weights (W.sub.c.sup.l) associated with the correlation matrix of the graph neural network layer. The learnable weights may be parameters adapted during a training of the graph neural network model 314, as described in further detail below. The layer function (Z.sub.s.sup.l) of a graph neural network layer may thus include a sum between a first addend (a product of the attention matrix, the combined feature vector, and the first weights, A.sub.s.sub.lF.sub.s.sup.lW.sub.a.sup.l) and a second addend [wherein the nodes are ordered prior to aggregation by the message passing based on the attention bits] (a product of the correlation matrix, the combined feature vector, and the second weights, C.sub.sF.sub.s.sup.lW.sub.c.sup.l). [0155] An attention matrix update rule (function) is denoted as g(.Math.), and is described in further detail below. Illustratively, the trained graph neural network model 314 may be configured to update the attention matrix of a graph neural network layer 324 to provide an output attention matrix (the attention matrix of the subsequent layer 324) by an update function having as function parameter the combined feature vector of the subsequent graph neural network layer [wherein the nodes are ordered prior to aggregation by the message passing based on the attention bits]…; Examiner notes that information in a graph neural network is processed using message passing as noted in the Ram and Or references and in the Nara reference in Nara [0128] In the context of graph neural networks, in case a large number of nodes are present, as may be the case for nodes representing wireless communication devices in a wireless network, nodes that are far away from each other (illustratively, not directly connected, but perhaps connected through a plurality of other nodes) may not transfer information to one another in an efficient manner. Illustratively, the transfer of information between nodes that are far away from each other may occur over a large number of graph network layer updates (e.g., over a large number of message passing, as described below). Further illustratively, for a node, if the graph neural network has L layers, information may propagate L steps away (e.g., to other nodes that are L nodes away)…)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Nara, Or and Ram for the same reasons disclosed above.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Ramamurthy et al. (US 20240028868, hereinafter ‘Ram’) in view of Orhan et al. (US 20220124543, hereinafter ‘Or’) in view further view of Kotary et al. (NPL: End-to-End Constrained Optimization Learning: A Survey, hereinafter ‘Kot’).
Regarding claim 6, the rejection of claim 1 is incorporated and Or further teaches the method according to claim 1, wherein the training is performed end-to-end with a differentiable satisfiability solver. (in [0078] … The graph custom-character.sub.t is updated, so that the next step s.sub.t+1 is obtained. The new node input features X.sub.cl.sup.(0) and X.sub.ue.sup.(0) are calculated every time the graph is updated, and the reward r(s.sub.t, a.sub.t) is calculated for each selected action. The L-layer GNN computation provides the score (e.g., Q-score) for each state-action pair. Then, to learn the NN weights W.sub.k.sup.(l)), ∀k, l, and w.sub.5, the Q-learning mechanism updates parameters by performing stochastic gradient descent (SGD) [wherein the training is performed end-to-end with a differentiable satisfiability solver] to minimize the squared loss E{(y−Q(s.sub.t, a.sub.t)).sup.2}, with y being defined in equation (A1a),…)
Additionally, Kot teaches in Sec. 6.2 : By contrast to approaches learning solutions to unstructured CO problems, a variety of methods learn to solve CO cast on graphs. The development of deep learning architectures, such as sequence models and attention mechanisms, as well as GNNs, has provided a natural toolset for these tasks. The survey categorizes these modern approaches broadly based on whether they rely supervised learning or reinforcement learning [wherein the training is performed end-to-end with a differentiable satisfiability solver]… The transition to RL was motivated partly by the difficulties associated with obtaining target solutions that are optimal, and the existence of nonunique optimal solutions to TSP instances. Rather than generating and targeting precomputed solutions, the authors present an actor-critic RL framework using expected tour length L(π|g) as the re ward signal, where g and π represent a graph (problem in stance) and a permutation (a tour over the graph):… The policy gradient [, wherein the training is performed end-to-end with a differentiable satisfiability solver] calculation requires a baseline function b which estimates the expected reward. In this work, an auxiliary critic network with its own set of parameters is trained to predict the expected tour length for any graph in supervised fashion, using empirically observed tours from the most recent policy as training labels…
Kot, Or and Ram are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art of processing information and techniques to solve constrained optimization problems using machine learning models and techniques, as disclosed by Kot with the method of developing information retrieval and processing techniques using as logic rule induction from knowledge graph analysis and machine learning techniques, as collectively disclosed by Or and Ram.
One of ordinary skill in the arts would have been motivated to combine the methods disclosed by Liu, Or and Ram as noted above. Doing so enables techniques for developing hybrid machine learning and optimization methods to predict fast, approximate, solutions to combinatorial problems and to enable structural logical inference, (Kot, Abstract).
Claims 8 and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Ramamurthy et al. (US 20240028868, hereinafter ‘Ram’) in view of Orhan et al. (US 20220124543, hereinafter ‘Or’) in view further view of Zhu et al. (US 20220366231, hereinafter ‘Zhu’) and Qiu et al. (US 20230351572, hereinafter ‘Qiu’).
Regarding claim 8, the rejection of claim 1 is incorporated. Ram and Or further teaches the method according to claim 1, (in [0016] In an embodiment, engineering data generated by engineering applications 112 is monitored and organized into knowledge graphs 150 as semantic data. Knowledge graphs 150 are the accumulation of design data [wherein the graph neural network generates a representation for each of the graph sequences] exported from engineering applications 112 and generated by knowledge graph algorithm that processes an ontology of the exported data. In some embodiments, knowledge graphs are obtained from a supplier, such as a vendor or manufacturer of similar systems, subsystems, or components related to the system under design. The ontology governs what types of elements of a system and the relationships between the elements are present (e.g., motor control, logic function block, associated sensor signals). The ontology also describes properties of the elements and the element relationships, and may organize the element types into hierarchies, such as super-types and sub-types. A knowledge graph 150 represents the ontology as nodes and edges that correspond to a set of elements of the ontology and element relationships, respectively [wherein the graph neural network generates a representation for each of the graph sequences].…)
Zhu teaches further comprising generating two graph sequences from the graph structured data by dropping edges or nodes randomly, (in [0113] In an embodiment, the method can also include infusing sparsity, e.g., into GCN. For instance, sparsity can be infused into an adjacency matrix that represents relationships among nodes. In an embodiment, infusing sparsity can include removing some of the edges represented in an adjacency matrix, based on a predefined threshold. In another embodiment, sparsity can be infused by randomly dropping some of the edges [further comprising generating two graph sequences from the graph structured data by dropping edges or nodes randomly] represented in an adjacency matrix at each training epoch of the GCN.)
Zhu, Or and Ram are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art of processing information and techniques using graph neural network, as disclosed by Zhu with the method of developing information retrieval and processing techniques using as logic rule induction from knowledge graph analysis and machine learning techniques, as collectively disclosed by Or and Ram.
One of ordinary skill in the arts would have been motivated to combine the methods disclosed by Zhu, Or and Ram as noted above. Doing so enables techniques for developing a graph neural network that can leverage heterogenous relations and infuse sparsity into network layers to reduce impact of noisy relationships, (Zhu, [0108]).
Qiu teaches wherein the graph neural network generates a representation for each of the graph sequences, and wherein the training is performed using a contrastive loss based on the representations. (in [0044] FIG. 2 discloses an example of a graph neural network according to an embodiment below. The system may include one or more graphs 201 that are utilized as a data set. The graphs 201 may be various nodes or representations of various data, including images (e.g. with pixels), financial data, etc. Each graph may be embedded by a set of GNNs 203. The GNNS may have up to K representations [wherein the graph neural network generates a representation for each of the graph sequences]. The GNNs 203 may be configured to embed each graph 201 to obtain representations, which may be diverse representations.… [0046] The OCGTL architecture (such as the embodiment shown in FIG. 2) may include K+1 GNNs 203. These networks may produce K+1 different embeddings given a graph 201 as input. The OCGTL may use two loss contributions that complement each other... It may be a contrastive loss [and wherein the training is performed using a contrastive loss based on the representations] that ensures that all the embeddings of the same graph are different from each other while still capturing important characteristics of the original input.)
Qiu, Zhu, Or and Ram are analogous art because both involve developing information retrieval and processing techniques using machine learning systems and algorithms.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of the prior art of processing information and techniques using graph neural networks, as disclosed by Qiu with the method of developing information retrieval and processing techniques using as logic rule induction from knowledge graph analysis and machine learning techniques, as collectively disclosed by Zhu, Or and Ram.
One of ordinary skill in the arts would have been motivated to combine the methods disclosed by Qiu, Zhu, Or and Ram as noted above. Doing so enables techniques for utilizing a plurality of graph neural networks (GNNs) to identify an aggregate loss, (Qiu, Abstract).
Regarding claim 9, the rejection of claim 8 is incorporated and Qiu further teaches the method according to claim 8, wherein the loss is built by minimizing a Kullback-Leibler (KL) divergence of the representation, by maximizing mutual information and/or using a cosine similarity function. (in [0046] The OCGTL architecture (such as the embodiment shown in FIG. 2) may include K+1 GNNs 203... This may be balanced by the transformation learning objective. It may be a contrastive loss that ensures that all the embeddings of the same graph are different from each other while still capturing important characteristics of the original input… [0051] where τ denotes a temperature parameter. The similarity here may be defined as the cosine similarity [wherein the loss is built bysimilarity function] sim(z, z.sup.1):=z.sup.Tz′/∥z∥ ∥z′|… [0072] Graph transformation learning (GTL) is an end-to-end self-supervised detection method using neural transformations. K GNNs, f.sub.k for k=1, . . . , K in addition to the reference feature extractor f are trained on custom-character.sub.GTL (eq. (2)). The loss is used directly to score anomalies. While this method works well in practice, it is not sensitive to the norm of the graph embeddings in eq. (2). The normalization step in computing the cosine similarity [wherein the loss is built by] makes mean and add pooling equivalent when aggregating the graph representations, and therefore loses the consideration of the norms of graph embedding)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine the teachings of Qiu, Zhu, Or and Ram for the same reasons disclosed above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Chen et al. (US 20210067549): teaches wherein the graph neural network generates a representation for each of the graph sequences, and wherein the training is performed using a contrastive loss based on the representations. (in [0087] Given a sequence of attributed graphs 1100 for a set of nodes 1104, where each node 1104 has a unique class label over a period of time [wherein the graph neural network generates a representation for each of the graph sequences], the present embodiments predict the labels of unlabeled nodes 1102 by learning from the labeled ones. In some embodiments, this can be used to detect anomalous network traffic…. And in [0027] Referring now to FIG. 3, additional detail on the adversarial GNN defense 43 is shown. Adversarial sample detection 302 leverages adversarial training to improve the robustness of detectors, by using adversarial samples. These samples may be generated by, e.g., gradient-based adversarial attacks.… [0040] Referring now to FIG. 5, additional detail is shown for adversarial contrastive learning 304. Adversarial contrastive learning 304 is used to generate good adversarial graph samples. An objective function for contrastive learning [wherein the training is performed using a contrastive loss based on the representations] can be expressed as: L=E.sub.x+.sub.,y+.sub.,y−[l(x.sup.+, y.sup.+, y.sup.−)]… )
Orhan et al. (US 20240049272): teaches in [0031] In various examples, the GNN may include a graph attention network (GAT) configured to apply attention (e.g. attention weight) parameters indicative or representative of assigned importance between nodes. In various examples, the determined score for each communication device may include the attention score associated for the communication device… [0085] In order to represent a graph with data items, an adjacency matrix A may be used to indicate or represent the presence of edges between the nodes in a manner that every node indexes a particular row and column in the adjacency matrix. In such example, a value of the adjacency matrix A associated with nodes (u,v), where u and v denote a first node and a second node, may be denoted with 1 if an edge is present between the first node u and the second node v, or the value of the adjacency matrix A associated with nodes (u,v) becomes 0 if an edge is not present between the nodes. In various aspects, an adjacency matrix may include a weighted adjacency matrix to represent weighted edges of the graph… Wang et al. (US 20230142254): teaches
Any inquiry concerning this communication or earlier communications from the examiner should be directed to OLUWATOSIN ALABI whose telephone number is (571)272-0516. The examiner can normally be reached Monday-Friday, 8:00am-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, Michael Huntley can be reached at (303) 297-4307. 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.
/OLUWATOSIN ALABI/ Primary Examiner, Art Unit 2129
Prosecution Insights