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 .
Response to Amendments
Claims 1-20 remain pending in the application.
Claims 1, 5-6, 13, 15, and 18-19 have been amended.
The amendment filed 1/22/2026 is sufficient to overcome the 35 U.S.C. 101
rejections of claims 1-20. The previous rejections have been withdrawn.
The amendment filed 1/22/2026 is sufficient to overcome the 103 rejections over Lecue in view of Costa of claims 1-4, 6-9, 13-17, and 19-20, the 103 rejections of claims 5 and 18 over Lecue in view of Costa and further in view of Dey Thomas, the 103 rejection of claim 10 over Lecue in view of Costa and further in view of Elbsat, the 103 rejection of claim 11 over Lecue in view of Costa and further in view of Mintz, and the 103 rejection of claim 12 over Lecue in view of Costa and further in view of Chen. The previous rejections have been withdrawn.
Response to Arguments
Argument 1, regarding the 101 rejections, applicant argues that the claims integrate the abstract ideas into a practical application of improved inference performance and scalability in graph-based computing systems. Applicant argues that the claims reflect improving forecasting performance and computational efficiency by reducing the impact of noisy or low-utility edges by selecting the appropriate subgraph. This is reflected in the claims by requiring the use of a selected subgraph, generated through structure learning scores, to perform time-series forecasting operation using sensor data. Examiner agrees and the 101 rejections have been withdrawn.
Argument 2, regarding the 103 rejections, applicant argues that none of the cited art teaches “wherein generating the plurality of structure learning scores comprises, for each edge: forming an edge representation by concatenating the node representations for respective nodes on either side of the edge and processing the edge representation using a shared linear layer to produce the structure learning score for the edge;… wherein the inferencing operation comprises performing a forecasting operation using time-series data collected from a plurality of sensors”.
Examiner respectfully disagrees because Lecue teaches wherein generating the plurality of structure learning scores comprises, for each edge: forming an edge representation by concatenating the node representations for respective nodes on either side of the edge (“A given edge is represented as a triple indicating a type of relation connecting two entities”, C1:L58-59. Combined graph 332 is formed by combining first graph 312 and second graph 211. A similarity may be assigned between each given element of the first and second graph on the combined graph, with the elements including nodes or edges, C24:L12-25) and processing the edge representation using a shared linear layer to produce the structure learning score for the edge (A similarity score is assigned to each vector corresponding to each edge in combined graph 332, C25:L24-33).
Lecue in view of Costa does not appear to explicitly teach “wherein the inferencing operation comprises performing a forecasting operation using time-series data collected from a plurality of sensors”.
Elbsat teaches wherein the inferencing operation comprises performing a forecasting operation using time-series data collected from a plurality of sensors (“time series data is received that provides information related to operation of various HVAC devices”, C5:L7-9. Subsystems 428 may include sensors that collect time series data, C10:L41-46, C14:L65-67).
The full prior art rejections are outlined below.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-4, 6-10, 13-17, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Lecue et al (Pub. No.: US 11442963 B1), hereafter Lecue, in view of Costa et al (Pub. No.: US 20240022593 A1), hereafter Costa and Elbsat et al (Pub. No.: US 12436528 B2), hereafter Elbsat.
Regarding claims 1, 13, and 15, Lecue teaches a method, a non-transitory computer readable medium comprising computer executable instructions, and an apparatus comprising: a memory; and at least one processor, coupled to said memory, and operative to perform operations comprising (“In the context of the present specification, “electronic device” is any computing apparatus”, C9:L7-8, “The processor is operatively connected to a non-transitory storage medium comprising instructions, the processor, upon executing the instructions, is configured for”, C4:L31-33, “there is disclosed a method for ranking subgraphs as potential explanations for a labelled edge type class,”, C2:L38-39, “Moreover, explicit use of the term “processor” or “controller” should not be construed to refer exclusively to hardware capable of executing software, and may implicitly include, without limitation, …, read-only memory (ROM) for storing software, random access memory (RAM)”, C12:L1-7): obtaining a graph with a plurality of nodes, a plurality of edges, and a plurality of node features (“obtaining a first graph, the first graph being a first type of representation of a set of labelled digital items, each labelled digital item being represented as a respective entity node connected via at least one respective labelled edge type to a respective value node”, C2:L41-46); generating node representations for the node features (“ the first graph being a first type of representation of a set of labelled digital items, each labelled digital item being represented as a respective entity node connected via at least one respective labelled edge type to a respective value node”, C4:L33-37. Node features are encoded, C20:L18-21); generating a plurality of structure learning scores based on the node representations xi, each structure learning score corresponding to one of the plurality of edges (a similarity score is generated indicating the similarity of a vector of unlabelled digital items and a vector of elements on a subgraph. Unlabeled digital items are linked to edges, C20:L42-51, C21:L4-14); wherein generating the plurality of structure learning scores comprises, for each edge: forming an edge representation by concatenating the node representations for respective nodes on either side of the edge (“A given edge is represented as a triple indicating a type of relation connecting two entities”, C1:L58-59. Combined graph 332 is formed by combining first graph 312 and second graph 211. A similarity may be assigned between each given element of the first and second graph on the combined graph, with the elements including nodes or edges, C24:L12-25) and processing the edge representation using a shared linear layer to produce the structure learning score for the edge (A similarity score is assigned to each vector corresponding to each edge in combined graph 332, C25:L24-33); selecting a subset of the plurality of edges that identify a subgraph, each edge of the subset having a structure learning score that is greater than a given threshold (“the subgraph identification 340 may generate vectors by embedding textual representations of each of the set of unlabelled digital items 306 and the combined graph 332 to perform the matching. In one or more embodiments, the semantic similarity and/or syntactic similarity may be in the form of a similarity score, and a given unlabelled digital item 644 may be mapped to a given subgraph based on the similarity score between the vector corresponding to the given unlabelled digital item 644 and the vector corresponding to elements of the given subgraph being above a threshold”, C21:L4-14).
Lecue does not appear to explicitly teach inputting the subgraph to a representation learner; and performing an inferencing operation using the representation learner based on the subgraph.
Costa teaches inputting the subgraph to a representation learner; and performing an inferencing operation using the representation learner based on the subgraph (graphs are used as input to a neural network to perform different tasks such as node classification and link predictions, P0040. These graphs may be represented in various ways, P0005. K-nearest neighbor algorithm is used on the graphs, which is an inferencing operation under the broadest reasonable interpretation, P0052, P0068).
Accordingly, it would have been obvious to a person having ordinary skill in the
art before the effective filing date of the claimed invention, having the teachings of
Lecue and Costa before them, to include Lecue’s specific teaching of inputting a graph into a neural network to perform node classification and link predictions in Lecue’s system of Ranking Subgraphs As Potential Explanations For Graph Classification. One would have been motivated to make such a combination of inputting graphs to a neural network for node classification and link predictions (see Costa P0040, P0005) and performing classification tasks in graphs by providing subgraphs as explanations for a given class to reduce errors and save computational resources (see Lecue C26:L6-15).
Lecue in view of Costa does not appear to explicitly teach “wherein the inferencing operation comprises performing a forecasting operation using time-series data collected from a plurality of sensors”.
Elbsat teaches wherein the inferencing operation comprises performing a forecasting operation using time-series data collected from a plurality of sensors (“time series data is received that provides information related to operation of various HVAC devices”, C5:L7-9. Subsystems 428 may include sensors that collect time series data, C10:L41-46, C14:L65-67).
Accordingly, it would have been obvious to a person having ordinary skill in the
art before the effective filing date of the claimed invention, having the teachings of
Lecue, Costa, and Elbsat before them, to include Elbsat’s specific teaching of analyzing time series data to control heating, ventilation, and air conditioning devices Lecue’s system of Ranking Subgraphs As Potential Explanations For Graph Classification. One would have been motivated to make such a combination of analyzing time series data to control heating, ventilation, and air conditioning devices (see Elbsat C5:L7-15) and using neural networks to analyze subgraphs and interpret data for mechanical means such as pressure-based, temperature based or any other suitable physical parameter based (see Lecue C26:L6-15, C26:L26-31).
Regarding claims 2 and 14, Lecue in view of Costa and Elbsat teaches the limitations of claims 1 and 13 as outlined above. Costa further teaches wherein the representation learner comprises a graph neural network (the neural network used is a graph neural network (GNN), P0039-P0040).
Regarding claims 3 and 16, Lecue in view of Costa and Elbsat teaches the limitations of claims 1 and 15 as outlined above. Costa further teaches wherein attention coefficients of the GAT are defined by:
PNG
media_image1.png
282
735
media_image1.png
Greyscale
(Graph attention network defined by first and fourth equations, P0055-P0057).
Regarding claims 4 and 17, Lecue in view of Costa and Elbsat teaches the limitations of claims 1 and 15 as outlined above. Costa further teaches wherein the generating of the node representations xi further
PNG
media_image2.png
138
711
media_image2.png
Greyscale
(Node features in vector h’ are multiplied by linear transformations parameterized by weight matrix WεRFxF applied to each node, P0054-P0055).
Regarding claims 6 and 19, Lecue in view of Costa and Elbsat teaches the limitations of claims 1 and 15 as outlined above. Costa further teaches wherein the structure learning scores
PNG
media_image3.png
76
730
media_image3.png
Greyscale
PNG
media_image4.png
70
742
media_image4.png
Greyscale
PNG
media_image5.png
202
725
media_image5.png
Greyscale
(The second and third equations recite || as a vector concatenation and a sigmoid activation used to predict node feature representations. Node representations are mapped onto node h’. P0055-P0058).
Regarding claims 7 and 20, Lecue in view of Costa and Elbsat teaches the limitations of claims 1 and 15 as outlined above. Costa further teaches further comprising adding K attention heads, wherein the node
PNG
media_image6.png
161
710
media_image6.png
Greyscale
(K amount of W weight matrices are produced and summed outputs are divided by K, with node representations denoted by αij, P0057).
Regarding claim 8, Lecue in view of Costa and Elbsat teaches the limitations of claim 1 as outlined above. Lecue further teaches wherein the method is performed without any exogenous regularizer and edge-selection heuristics (Generic convolutional neural network is recited, which is a regularized multilayer perceptron. No edge selection heuristics or exogenous regularizer are recited, C7:L2-14).
Regarding claim 9, Lecue in view of Costa and Elbsat teaches the limitations of claim 1 as outlined above. Costa further teaches identifying an attempted online financial fraud event based on a result of the inferencing operation (Graph neural network may be used to identify a fraudulent login attempt in an online system. The online system may include billing. P0007, P0025. Graphical neural network makes this identification based on a prediction of fraud using predicted virtual graphs and a K-nearest neighbor algorithm. K-nearest neighbor is an inferencing operation under the broadest reasonable interpretation, P0052, P0068); and blocking a completion of the attempted online financial fraud event by changing a network security parameter (After detecting the fraudulent activity, a mitigation action is taken to protect the online system, P0007).
Regarding claim 10, Lecue in view of Costa and Elbsat teaches the limitations of claim 1 as outlined above. Elbsat further teaches collecting a set of time series data from a plurality of environmental sensors (“time series data is received that provides information related to operation of various HVAC devices”, C5:L7-9); forecasting one or more environmental comfort indicators for a given time horizon based on the time series data and the subgraph; and controlling a heating, ventilation and air conditioning system based on the forecasted environmental comfort indicators (“A building controller may receive the time series data and determine, via several fault detection methods, several fault detection indications based on the time series data. A supervisory layer with neural network functionality may receive these fault detection indications and make a final control decision based on the several fault detection indications”, C5:L9-15).
Claims 5 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Lecue in view of Costa and Elbsat and further in view of Dey Thomas et al (Pub. No.: WO 2017174627 A1), hereafter Thomas.
Regarding claims 5 and 18, Lecue in view of Costa and Elbsat teaches the limitations of claims 1 and 15 as outlined above. Costa further teaches wherein the structure learning scores comprise probabilities for retaining each edge of the plurality of edges, the node representations are designated by xi and x1, and wherein the generating of the structure learning scores further comprises: mapping the edge representations onto structure learning scores
PNG
media_image7.png
39
98
media_image7.png
Greyscale
using a shared linear layer
PNG
media_image8.png
38
87
media_image8.png
Greyscale
, …, and performing a sigmoid function to map a result of the mapping to a range between zero and one inclusive:
PNG
media_image9.png
45
222
media_image9.png
Greyscale
where II is a vector concatenation, …, and - is a sigmoid activation allowing each ij to be interpreted as the probability of retaining a given edge of the plurality of edges that resides between nodes i and j (The second and third equations recite || as a vector concatenation and a sigmoid activation used to predict node feature representations. Node representations are mapped onto node h’. P0055-P0058).
Lecue in view of Costa and Elbsat does not appear to explicitly teach adding noise u… u ER' is independent and identically distributed noise drawn from a U(-0.5, 0.5) distribution centered about zero.
Thomas teaches adding noise u… u ER' is independent and identically distributed noise drawn from a U(-0.5, 0.5) distribution centered about zero (The probability of an edge being preserved may be calculated with a sigmoid function relative to noise, page 4, lines 28-32).
Accordingly, it would have been obvious to a person having ordinary skill in the
art before the effective filing date of the claimed invention, having the teachings of
Lecue, Costa, Elbsat, and Thomas before them, to include Thomas’s specific teaching of calculating the probability of an edge being preserved with a sigmoid function relative to noise in Lecue’s system of Ranking Subgraphs As Potential Explanations For Graph Classification. One would have been motivated to make such a combination of calculating the probability of an edge being preserved with a sigmoid function relative to noise (see Thomas page 4, lines 28-32) and ranking edge property types by respective probabilities to produce potential explanation subgraphs (see Lecue C23:L24-31).
Claim 11 is rejected under 35 U.S.C. 103 as being unpatentable over Lecue in view of Costa and Elbsat and further in view of Mintz et al (Pub. No.: US 20220319312 A1), hereafter Mintz.
Regarding claim 11, Lecue in view of Costa teaches the limitations of claim 1 as outlined above. Lecue in view of Costa and Elbsat does not appear to explicitly teach collecting a set of time series data from a network of traffic sensor stations; forecasting traffic for a given time horizon based on the time series data and the subgraph; and controlling an autonomous vehicle based on the forecast traffic.
Mintz teaches collecting a set of time series data from a network of traffic sensor stations (Time series data regarding traffic patterns is collected, P0134-P0135, P0295); forecasting traffic for a given time horizon based on the time series data and the subgraph (Time series data is analyzed for link level traffic such as lane changing behavior, P0134); and controlling an autonomous vehicle based on the forecast traffic (Traffic predictions may be incorporated when controlling autonomous vehicles, P0134).
Accordingly, it would have been obvious to a person having ordinary skill in the
art before the effective filing date of the claimed invention, having the teachings of
Lecue, Costa, Elbsat, and Mintz before them, to include Mintz’s specific teaching of analyzing time series data to predict traffic and control an autonomous vehicle Lecue’s system of Ranking Subgraphs As Potential Explanations For Graph Classification. One would have been motivated to make such a combination of analyzing time series data to predict traffic and control an autonomous vehicle (see Mintz P0134-P0135, P0295) and using neural networks to analyze subgraphs and interpret data including components of a vehicle (see Lecue C14:L25-55, C14:L63-67).
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Lecue in view of Costa and Elbsat and further in view of Chen et al (Pub. No.: US 20240046075 A1), hereafter Chen.
Regarding claim 12, Lecue in view of Costa teaches the limitations of claim 1 as outlined above. Lecue in view of Costa and Elbsat does not appear to explicitly teach wherein the inputting the subgraph to the representation learner further comprises incorporating the subgraph into a Graph Convolutional Network (GCN) by multiplying an adjacency matrix A by a mask M before an application of a renormalization trick.
Chen teaches wherein the inputting the subgraph to the representation learner further comprises incorporating the subgraph into a Graph Convolutional Network (GCN) by multiplying an adjacency matrix A by a mask M before an application of a renormalization trick (Subgraphs are incorporated in a graph convolutional network by applying respective stochastic binary masks to fixed adjacency matrices, P0103).
Accordingly, it would have been obvious to a person having ordinary skill in the
art before the effective filing date of the claimed invention, having the teachings of
Lecue, Costa, Elbsat, and Chen before them, to include Chen’s specific teaching of incorporating subgraphs in a graph convolutional neural network by applying stochastic binary masks to fixed adjacency matrices Lecue’s system of Ranking Subgraphs As Potential Explanations For Graph Classification. One would have been motivated to make such a combination of incorporating subgraphs in a graph convolutional neural network by applying stochastic binary masks to fixed adjacency matrices (see Chen P0103) and encoding a graph with a graph embedding ML model as a matrix to perform various machine learning tasks (see Lecue C15:L19-27).
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 nonprovisional extension fee (37 CFR 1.17(a)) 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 ISHAN MOUNDI whose telephone number is (703)756-1547. The examiner can normally be reached 8:30 A.M. - 5 P.M..
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, Matthew Ell can be reached at (571) 270-3264. 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.
/I.M./ Examiner, Art Unit 2141
/MATTHEW ELL/Supervisory Patent Examiner, Art Unit 2141