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 .
This action is responsive to the application filed 02/20/2024. Claims 1-20 are presented for examination.
Priority
Applicant’s claim for the benefit of a provisionally filed application, filed 04/14/2023, is acknowledged.
Information Disclosure Statement
The information disclosure statement (IDS) submitted 02/20/2024, 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 therefore, subject to the conditions and requirements of this title.
Claims 1-3, 5-10, 12-17, & 19-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1
Step 1: The claim recites “A computer-implemented method for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, comprising:”; therefore, it is directed to the statutory category of a process.
Step 2A Prong 1: The claim recites, inter alia:
generating a first user interaction graph for a first time window, the first user interaction graph representing the users as respective first user nodes, the items as respective first item nodes, and each user-item interaction that occurred during the first time window as a respective first edge connecting a first user node to a first item node; generating a second user interaction graph for a second time window that follows the first time window, the second user interaction graph representing the users as respective second user nodes, the items as respective second item nodes, and each user-item interaction that occurred during the second time window as a respective second edge connecting a second user node to a second item node; sampling second user node and second item node pairs from the second user interaction graph: These limitations recite mathematical relationships similar to organizing information and manipulating information through mathematical correlations per MPEP 2106.04(a)(2)(I)(A)(iv).
Thus, the claim recites a judicial exception.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The additional elements of the claim are as follows:
A computer-implemented method for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, comprising: These additional elements are recited at a high level of generality and merely indicate a field of use or technological environment in which to apply a judicial exception, e.g. a computer-implemented method for training a heterogeneous graph neural network (GNN), to a particular technological environment or field of use, e.g. to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items. See MPEP 2106.05(h). Elements that use or interact with the judicial exception do not integrate the judicial exception into a practical application.
by at least one computer processor: These additional elements are recited at a high level of generality and amount to invoking computers or other machinery merely as a tool to apply the underlying judicial exception. See MPEP § 2106.05(f).
and training the heterogeneous GNN based on first user node and first item node pairs from the first user interaction graph that respectively correspond to the sampled second user node and second item node pairs from the second user interaction graph: These additional elements recite only the idea of training the heterogeneous GNN based on first user node and first item node pairs from the first user interaction graph that respectively correspond to the sampled second user node and second item node pairs from the second user interaction graph and attempts to cover any implementation of the training without any restriction as to how the first user node and first item node pairs from the first user interaction graph are incorporated; the specific style of training, or how this style of training is carried out. These additional elements do not meaningfully limit the claim and does not integrate the judicial exception into a practical application because this type of recitation is equivalent to the words “apply it”. See MPEP 2106.05(f).
Thus, the way in which the additional elements use or interact with the judicial exception do not integrate the judicial exception into a practical application.
Step 2B: The additional elements from Step 2A Prong 2 include generally linking the use of the judicial exception to indicate a field of use or technological environment, invoking generic computer components to apply the underlying judicial exception, and adding words equivalent to "apply it" with the judicial exception. Thus, the additional elements, viewed individually or in combination, do not provide an inventive concept or otherwise amount to significantly more than the abstract idea itself. See MPEP 2106.05.
Claim 2
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the abstract ideas of claim 1 as well as, inter alia:
utilizing the trained heterogeneous GNN to generate a first graph user embedding corresponding to a first user and a first graph item embedding corresponding to a first item: These limitations recite mathematical relationships similar to organizing information and manipulating information through mathematical correlations per MPEP 2106.04(a)(2)(I)(A)(iv).
and determining a relevancy of the first item to the first user based on the first graph user embedding and the first graph item embedding: These limitations recite a mentally performable process with the aid of pen and paper of using observation and judgement to determine a relevancy of the first item to the first user based on the first graph user embedding and the first graph item embedding.
Thus, the claim recites a judicial exception.
Step 2A Prong 2 & Step 2B: There are no additional elements recited so the claim does not provide a
practical application and is not considered to be significantly more. As such, the claim is patent
ineligible.
Claim 3
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the abstract ideas as the judicial exception of claim 1.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The additional elements of the claim are as follows:
wherein the first time window and the second time window are overlapping: These additional elements are recited at a high level of generality and merely indicate a field of use or technological environment in which to apply a judicial exception, e.g. the first time window and the second time window, to a particular technological environment or field of use, e.g. are overlapping. See MPEP 2106.05(h). Elements that use or interact with the judicial exception do not integrate the judicial exception into a practical application.
Thus, the way in which the additional elements use or interact with the judicial exception do not integrate the judicial exception into a practical application.
Step 2B: The additional elements from Step 2A Prong 2 include generally linking the use of the judicial exception to indicate a field of use or technological environment. Thus, the additional elements, viewed individually or in combination, do not provide an inventive concept or otherwise amount to significantly more than the abstract idea itself. See MPEP § 2106.05.
Claim 5
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the abstract ideas of claim 1 as well as, inter alia:
generating the first user interaction graph comprises assigning a weight to each first edge of the first user interaction graph, wherein the weight represents a popularity of the item represented by the first item node to which the first edge is connected: These limitations recite mathematical relationships similar to organizing information and manipulating information through mathematical correlations per MPEP 2106.04(a)(2)(I)(A)(iv).
and training the heterogeneous GNN comprises minimizing a loss function that incurs a greater penalty for user-item interactions with items having a greater popularity as reflected by the weight: These limitations recite mathematical calculations similar to an act of calculating using mathematical methods to determine a variable or number per MPEP 2106.04(a)(2)(I)(C).
Thus, the claim recites a judicial exception.
Step 2A Prong 2 & Step 2B: There are no additional elements recited so the claim does not provide a
practical application and is not considered to be significantly more. As such, the claim is patent
ineligible.
Claim 6
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the abstract ideas as the judicial exception of claim 1.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The additional elements of the claim are as follows:
wherein the heterogeneous GNN comprises a heterogeneous GraphSAGE neural network: These additional elements are recited at a high level of generality and merely indicate a field of use or technological environment in which to apply a judicial exception, e.g. the heterogeneous GNN, to a particular technological environment or field of use, e.g. comprises a heterogeneous GraphSAGE neural network. See MPEP 2106.05(h). Elements that use or interact with the judicial exception do not integrate the judicial exception into a practical application.
Thus, the way in which the additional elements use or interact with the judicial exception do not integrate the judicial exception into a practical application.
Step 2B: The additional elements from Step 2A Prong 2 include generally linking the use of the judicial exception to indicate a field of use or technological environment. Thus, the additional elements, viewed individually or in combination, do not provide an inventive concept or otherwise amount to significantly more than the abstract idea itself. See MPEP § 2106.05.
Claim 7
Step 1: A process, as above.
Step 2A Prong 1: The claim recites the abstract ideas as the judicial exception of claim 1.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The additional elements of the claim are as follows:
wherein the user-item interactions comprises one or more of: a user being shown information about the item; a user interacting with a user interface to obtain information about the item; or a user launching the item for playback: These additional elements are recited at a high level of generality and merely indicate a field of use or technological environment in which to apply a judicial exception, e.g. the user-item interactions, to a particular technological environment or field of use, e.g. comprises one or more of: a user being shown information about the item; a user interacting with a user interface to obtain information about the item; or a user launching the item for playback. See MPEP 2106.05(h). Elements that use or interact with the judicial exception do not integrate the judicial exception into a practical application.
Thus, the way in which the additional elements use or interact with the judicial exception do not integrate the judicial exception into a practical application.
Step 2B: The additional elements from Step 2A Prong 2 include generally linking the use of the judicial exception to indicate a field of use or technological environment. Thus, the additional elements, viewed individually or in combination, do not provide an inventive concept or otherwise amount to significantly more than the abstract idea itself. See MPEP § 2106.05.
Claims 8-10 & 12-14
Step 1: These claims are directed to “A system for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, comprising:”; therefore, it is directed the statutory category of a machine.
Step 2A Prong 1: Claims 8-10 & 12-14 recite the same judicial exceptions as Claims 1-3 & 5-7, respectively.
Step 2A Prong 2: The judicial exception recited in these claims are not integrated into a practical application. The analysis at this step for Claims 8-10 & 12-14 mirrors that of Claims 1-3 & 5-7, respectively.
Step 2B: The additional elements from Step 2A Prong 2 do not contain significantly more than the judicial exception for these claims. The analysis at this step for Claims 8-10 & 12-14 mirrors that of Claims 1-3 & 5-7, respectively.
Claims 15-17 & 19-20
Step 1: This claim recites "A non-transitory computer-readable medium having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, the operations comprising:"; therefore, it is directed to the statutory category of an article of manufacture.
Step 2A Prong 1: Claims 15-17 & 19-20 recite the same judicial exception as Claims 1-3 & 5-7, respectively.
Step 2A Prong 2: The judicial exception recited in these claims are not integrated into a practical application. The only difference between Claims 15-17 & 19-20 and Claims 1-3 & 5-7, is that Claims 15-17 & 19-20 are directed to "A non-transitory computer-readable medium having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, the operations comprising”. However, mere recitation that a judicial exception is to be performed using generic computer equipment in their ordinary capacity, i.e. a non-transitory computer-readable medium having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, the operations comprising, cannot meaningfully integrate the judicial exception into a practical application. See MPEP 2106.05(f). With that exception, the analysis at this step for Claims 15-17 & 19-20 mirrors that of Claims 1-3 & 5-7, respectively.
Step 2B: The additional elements from Step 2A Prong 2 do not contain significantly more than the judicial exception for these claims. The only difference between Claims 15-17 & 19-20 and Claims 1-3 & 5-7, is that Claims 15-17 & 19-20 are directed to "A non-transitory computer-readable medium having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, the operations comprising”. However, mere recitation that a judicial exception is to be performed using generic computer equipment in their ordinary capacity, i.e. a non-transitory computer-readable medium having instructions stored thereon that, when executed by at least one computing device, cause the at least one computing device to perform operations for training a heterogeneous graph neural network (GNN) to generate graph user embeddings corresponding to users and graph item embeddings corresponding to items, the operations comprising, cannot amount to significantly more than the judicial exception. See MPEP 2106.05(f). With that exception, the analysis at this step for Claims 15-17 & 19-20 mirrors that of Claims 1-3 & 5-7, respectively.
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 1-2, 7-9, 14-16, & 20 are rejected under 35 U.S.C. 103 as being unpatentable over Joseph et al. (US 10650432 B1, published 05/12/2020), hereafter Joseph, in view of Sun et al. ("Multi-Graph Convolution Collaborative Filtering", Montreal Research Center, arXiv) (Year: 2020), hereafter Sun, and further in view of You et al. ("ROLAND: Graph Learning Framework for Dynamic Graphs", Conference on Knowledge Discovery and Data Mining, arXiv) (Year: 2022), hereafter You.
Regarding independent claim 1, Joseph teaches a method comprising:
generating, by at least one computer processor, a first time window, and each user-item interaction that occurred during the first time window ([Col. 2, Lines 46-50 & Col. 7, Lines 55-65] discusses building a first time-windowed representation of user-item interaction data from a subset of values being used as input);
generating a second time window that follows the first time window, and each user-item interaction that occurred during the second time window ([Col. 7, Lines 55-65] discusses building a second time-windowed representation, that is non-overlapping with the first window, using the rest of the values of the same repository);
and training a prediction model using the first time window inputs and the second time window inputs ([Col. 8, Lines 17-44] discusses training a prediction model using the first window inputs to predict future interactions corresponding to the second window).
Joseph does not explicitly teach generating, by at least one computer processor, a first user interaction graph for a first time window, the first user interaction graph representing the users as respective first user nodes, the items as respective first item nodes, and each user-item interaction that occurred during the first time window as a respective first edge connecting a first user node to a first item node; generating a second user interaction graph for a second time window that follows the first time window, the second user interaction graph representing the users as respective second user nodes, the items as respective second item nodes, and each user-item interaction that occurred during the second time window as a respective second edge connecting a second user node to a second item node; sampling second user node and second item node pairs from the second user interaction graph; and training the heterogeneous GNN based on first user node and first item node pairs from the first user interaction graph that respectively correspond to the sampled second user node and second item node pairs from the second user interaction graph.
However, in a similar field of endeavor, Sun teaches a method generating an interaction graph, comprising connected user nodes and item nodes ([Sec. 3 & Fig. 2] discusses generating a bipartite graph with two types of nodes, user nodes and item nodes, connected by edges representing the interaction events between nodes); and sampling user nodes and item node pairs from a user interaction graph ([Sec. 3] discusses sampling user and item pairs from the graph).
Because Joseph teaches a method for generating a first and second time window containing user-item interactions that occurred within each time window, and training a model using both time windows; and Sun teaches a method for generating an interaction graph wherein user nodes and item nodes are connected by edges representing their interaction events, and sampling item node pairs from the user interaction graph, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate generating an interaction graph wherein user nodes and item nodes are connected by edges representing their interaction events, and sampling item node pairs from the user interaction graph as taught by Sun into Joseph’s computer-implemented method, with a reasonable expectation of success, to teach generating, by at least one computer processor, a first user interaction graph for a first time window, the first user interaction graph representing the users as respective first user nodes, the items as respective first item nodes, and each user-item interaction that occurred during the first time window as a respective first edge connecting a first user node to a first item node; generating a second time window that follows the first time window, and each user-item interaction that occurred during the second time window; and sampling second user node and second item node pairs from the second user interaction graph. This combination would have been motivated by the desire to implement time windows onto an item interaction graph, wherein the graph effectively captures the importance of the user item relationship structure (Sun [Abstract]).
The combination of Joseph and Sun does not explicitly teach generating a second user interaction graph for a second time window that follows the first time window, the second user interaction graph representing the users as respective second user nodes, the items as respective second item nodes, and each user-item interaction that occurred during the second time window as a respective second edge connecting a second user node to a second item node; and training the heterogeneous GNN based on first user node and first item node pairs from the first user interaction graph that respectively correspond to the sampled second user node and second item node pairs from the second user interaction graph.
However, in a similar field of endeavor, You teaches a method for generating a second interaction graph ([Sec. 3] discusses a sequence of discrete graphs built from edges of interactions in each respective time window; thus, there exists a second user interaction graph built from second user nodes and second item nodes); and training a GNN based on the first user node and item nodes corresponding to second user nodes and item nodes of the second interaction graph ([Sec. 3 & Fig. 1] discusses training a GNN using a first interaction graph and its users and items to predict edges corresponding to the specific node pairs of the second interaction graph).
Because the combination of Joseph and Sun teaches a method for generating an interaction graph within a first time window wherein user nodes and item nodes are connected by edges representing their interaction events, generating a second time window containing user-item interactions that occurred within that time window, training a model using both time windows, and sampling item node pairs from the user interaction graph; and You teaches a method for generating a second interaction graph, and training a GNN based on the first user node and item nodes corresponding to second user nodes and item nodes of the second interaction graph, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate generating a second interaction graph, and training a GNN based on the first user node and item nodes corresponding to second user nodes and item nodes of the second interaction graph as taught by You into the combination of Joseph and Sun’s computer-implemented method, with a reasonable expectation of success, to teach generating, by at least one computer processor, a first user interaction graph for a first time window, the first user interaction graph representing the users as respective first user nodes, the items as respective first item nodes, and each user-item interaction that occurred during the first time window as a respective first edge connecting a first user node to a first item node; generating a second user interaction graph for a second time window that follows the first time window, the second user interaction graph representing the users as respective second user nodes, the items as respective second item nodes, and each user-item interaction that occurred during the second time window as a respective second edge connecting a second user node to a second item node; sampling second user node and second item node pairs from the second user interaction graph; and training the heterogeneous GNN based on first user node and first item node pairs from the first user interaction graph that respectively correspond to the sampled second user node and second item node pairs from the second user interaction graph. This combination would have been motivated by the desire to implement the first and second time windows into separate non-overlapping interaction graphs in order train the GNN using the separated graphs (You [Alg. 1-2 & Sec. 3]).
Regarding dependent claim 2, the combination of Joseph, Sun, and You teaches the claimed invention as claimed in claim 1, including utilizing the trained heterogeneous GNN to generate a first graph user embedding corresponding to a first user and a first graph item embedding corresponding to a first item (You [Sec. 3, Alg. 1-2, & Fig. 1] discusses generating a graph user and item embedding for a first graph and item); and determining a relevancy of the first item to the first user based on the first graph user embedding and the first graph item embedding (Sun [Abstract & Sec. 4] discusses the goal of determining relevancy of items to a user based on the embeddings of the users and items).
Regarding dependent claim 7, the combination of Joseph, Sun, and You teaches the claimed invention as claimed in claim 1, including wherein the user-item interactions comprises one or more of: a user being shown information about the item; a user interacting with a user interface to obtain information about the item; or a user launching the item for playback (Joseph [Col. 7, Lines 5-14] discusses user-item interactions comprising the user viewing information about the item or clicking on and interacting with the item).
Regarding claims 8-9 & 14, claims 8-9 & 14 are system claims that are substantially the same as the method of claims 1-2 & 7, respectively. Therefore, claims 8-9 & 14 are rejected for the same reasons as claims 1-2 & 7, respectively.
Regarding claims 15-16 & 20, they are computer-readable storage medium claims that are
substantially the same as the method of claims 1-2 & 7, respectively. Therefore, claims 15-16 & 20 are rejected for the same reasons as claims 1-2 & 7, respectively.
Claims 3, 10, & 17 are rejected under 35 U.S.C. 103 as being unpatentable over Joseph, in view of Sun, in view of You, as applied in claim 1, and further in view of Rybakov et al. (US 10824940 B1), hereafter Rybakov.
Regarding dependent claim 3, the combination of Joseph, Sun, and You teaches the claimed invention as claimed in claim 1, including:
generating, by at least one computer processor, a first time window, and each user-item interaction that occurred during the first time window (Joseph [Col. 2, Lines 46-50 & Col. 7, Lines 55-65] discusses building a first time-windowed representation of user-item interaction data from a subset of values being used as input);
generating a second time window that follows the first time window, and each user-item interaction that occurred during the second time window (Joseph [Col. 7, Lines 55-65] discusses building a second time-windowed representation, that is non-overlapping with the first window, using the rest of the values of the same repository);
The combination of Joseph, Sun, and You does not explicitly teach wherein the first time window and the second time window are overlapping.
However, Rybakov teaches a method for a recommendation system using a neural network in which multiple windows overlap during training ([Col. 2, Lines 24-41 & Col. 6, Lines 49-63] discusses the use of multiple datasets that are overlapping, wherein during training, sliding windows allow for one dataset to overlap with a separate dataset; thus, the first time window and second time window are overlapping).
Because the combination of Joseph, Sun, and You teaches generating a first and second time window; and Rybakov teaches the use of overlapping first and second time windows, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the use of overlapping first and second time windows as taught by Rybakov into the combination of Joseph, Sun, and You’s computer-implemented method, with a reasonable expectation of success, to teach wherein the first time window and the second time window are overlapping. This combination would have been motivated by the desire to accommodate to changes in item interaction behavior across the separate time windows (Rybakov [Col. 4-5, Lines 66-3]).
Regarding claim 10, claim 10 is a system claim that is substantially the same as the method of claim 3. Therefore, claim 10 is rejected for the same reasons as claim 3.
Regarding claim 17, claim 17 is a computer-readable storage medium claim that is substantially the same as the method of claim 3. Therefore, claim 17 is rejected for the same reasons as claim 3.
Claims 4, 11, & 18 are rejected under 35 U.S.C. 103 as being unpatentable over Joseph, in view of Sun, in view of You, as applied in claim 1, further in view of He et al. ("PinSage: A new graph convolutional neural network for web-scale recommender systems", Interest Engineering Blog, August 15, 2018, 14 pages), hereafter He, and further in view of Yin et al. ("Challenging the Long Tail Recommendation", Proceedings of the VLDB Endowment, Vol. 5, No. 9, arXiv) (Year: 2012), hereafter Yin.
He was cited in the IDS submitted 02/20/2024.
Regarding dependent claim 4, the combination of Joseph, Sun, and You teaches the claimed invention as claimed in claim 1, including wherein sampling the second user node and second item node pairs from the second user interaction graph comprises identifying a most recent user-item interaction (Joseph [Col. 12, Lines 46-52] discusses ordering item interaction events by weight depending on how recent they occurred); and identifying a most and least popular item (Sun [Sec. 3] discusses the most frequent items attract many interactions while the least popular items attract few interactions; thus, there exists a popularity skew in the bipartite graph).
The combination of Joseph, Sun, and You does not explicitly teach conducting a predetermined number of walks of a predetermined number of hops from each second user node and second item node in the second user interaction graph and identifying the second user node and second item node pairs based on the walks, wherein: a first hop of a first walk is along a second edge representing a most recent user-item interaction; a first hop of a third walk is along a second edge representing a user-item interaction with a most popular item; and a first hop of a fourth walk is along a second edge that is selected at random.
However, in a similar field of endeavor, He teaches a method for a web-scale recommender system in wherein sampling is done by conducting a fixed number of walks of a fixed number of hops ([Sec. 3.2-3.3] discusses conducting a fixed length of walks of a predetermined number of hops from each node to construct the node’s neighborhood of item node pairs); conducting a first hop of a first walk along an edge representing a most recent user-item interaction ([Sec. 3.2-3.3] discusses the neighborhood constructed from the process of walks comprises the top nodes; thus, the most recent, popular items are visited in the first hop); conducting a first hop of a third walk along an edge representing a user-item interaction with a most popular item ([Sec. 3.2-3.3] discusses the neighborhood constructed from the process of walks comprises the top nodes with the highest visit counts indicating the most popular items); and conducting a first hop of a fourth walk along an edge that is selected at random ([Sec. 3.2-3.3] discusses conducting random walks of a fixed walk length from each node to construct the node’s neighborhood).
Because the combination of Joseph, Sun, and You teaches sampling the second user node and second item node pairs from the second user interaction graph comprising identifying a most recent user-item interaction, and identifying a most and least popular item; and He teaches sampling executed by conducting a fixed number of walks of a fixed number of hops, conducting a first hop of a first walk along an edge representing a most recent user-item interaction, conducting a first hop of a third walk along an edge representing a user-item interaction with a most popular item, and conducting a first hop of a fourth walk along an edge that is selected at random, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate sampling executed by conducting a fixed number of walks of a fixed number of hops, conducting a first hop of a first walk along an edge representing a most recent user-item interaction, conducting a first hop of a third walk along an edge representing a user-item interaction with a most popular item, and conducting a first hop of a fourth walk along an edge that is selected at random as taught by He into the combination of Joseph, Sun, and You’s computer-implemented method, with a reasonable expectation of success, to teach conducting a predetermined number of walks of a predetermined number of hops from each second user node and second item node in the second user interaction graph and identifying the second user node and second item node pairs based on the walks, wherein: a first hop of a first walk is along a second edge representing a most recent user-item interaction; a first hop of a third walk is along a second edge representing a user-item interaction with a most popular item; and a first hop of a fourth walk is along a second edge that is selected at random. This combination would have been motivated by the desire to implement efficient random walks in order to design a training strategy that relies on harder training examples to improve scalability, robustness, and convergence of the model (He [Abstract & Sec. 1]).
The combination of Joseph, Sun, You, and He does not explicitly teach conducting a first hop of a second walk along an edge representing a user-item interaction with a least popular item.
However, in a similar field of endeavor, Yin teaches a method for graph-based representation of user-item information in which the least popular item is identified, and a walk is conducted along the corresponding edge ([Sec. 3.1-3.3] discusses a random walk process in which the least popular item is visited first by being selected given their low probability).
Because the combination of Joseph, Sun, You, and He teaches sampling the second user node and second item node pairs from the second user interaction graph comprising conducting a fixed number of walks of a fixed number of hops, conducting a first hop of a first walk along an edge representing a most recent user-item interaction, conducting a first hop of a third walk along an edge representing a user-item interaction with a most popular item, and conducting a first hop of a fourth walk along an edge that is selected at random; and Yin teaches conducting a first hop of a second walk along an edge representing a user-item interaction with a least popular item, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate conducting a first hop of a second walk along an edge representing a user-item interaction with a least popular item as taught by Yin into the combination of Joseph, Sun, You, and He’s computer-implemented method, with a reasonable expectation of success, to teach wherein sampling the second user node and second item node pairs from the second user interaction graph comprises: conducting a predetermined number of walks of a predetermined number of hops from each second user node and second item node in the second user interaction graph and identifying the second user node and second item node pairs based on the walks, wherein: a first hop of a first walk is along a second edge representing a most recent user-item interaction; a first hop of a second walk is along a second edge representing a user-item interaction with a least popular item; a first hop of a third walk is along a second edge representing a user-item interaction with a most popular item; and a first hop of a fourth walk is along a second edge that is selected at random. This combination would have been motivated by the desire to implement recommendation of less popular items while improving recommendation diversity and accuracy (Yin [Abstract]).
Regarding claim 11, claim 11 is a system claim that is substantially the same as the method of claim 4. Therefore, claim 11 is rejected for the same reasons as claim 4.
Regarding claim 18, claim 18 is a computer-readable storage medium claim that is substantially the same as the method of claim 4. Therefore, claim 18 is rejected for the same reasons as claim 4.
Claims 5, 12, & 19 are rejected under 35 U.S.C. 103 as being unpatentable over Joseph, in view of Sun, in view of You, as applied in claim 1, and further in view of Yi et al. ("Sampling-Bias-Corrected Neural Modeling for Large Corpus Item Recommendations", ACM Conference on Recommender Systems, ACM) (Year: 2019), hereafter Yi.
Regarding dependent claim 5, the combination of Joseph, Sun, and You teaches the claimed invention as claimed in claim 1, including generating the first user interaction graph comprises assigning a weight to each first edge of the first user interaction graph, wherein the weight represents a popularity of the item represented by the first item node to which the first edge is connected (Joseph [Col. 8, Lines 23-28 & Lines 2-20] discusses assigning a weight to each connection edge based on a pattern identified between them, the pattern being popularity or recency).
The combination of Joseph, Sun, and You does not explicitly teach training the heterogeneous GNN comprises minimizing a loss function that incurs a greater penalty for user-item interactions with items having a greater popularity as reflected by the weight.
However, in a similar field of endeavor, Yi teaches a method for a recommendation system in which bias is corrected by minimizing a loss function that generates a penalty for user-item interactions that have a greater popularity as reflected by the weight ([Sec. 3] discusses a softmax loss function that generates a high penalty for candidates that have a high popularity).
Because the combination of Joseph, Sun, and You teaches assigning a weight to each first edge of the first user interaction graph, wherein the weight represents a popularity of the item; and Yi teaches minimizing a loss function that incurs a penalty scaled with the popularity of each respective user-item interactions, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate minimizing a loss function that incurs a penalty scaled with the popularity of each respective user-item interactions as taught by Yi into the combination of Joseph, Sun, and You’s computer-implemented method, with a reasonable expectation of success, to teach generating the first user interaction graph comprising assigning a weight to each first edge of the first user interaction graph, wherein the weight represents a popularity of the item represented by the first item node to which the first edge is connected; and training the heterogeneous GNN comprises minimizing a loss function that incurs a greater penalty for user-item interactions with items having a greater popularity as reflected by the weight. This combination would have been motivated by the desire to implement an algorithm that can work without requiring fixed item vocabulary and is capable of producing unbiased estimation and being adaptive to item distribution change (Yi [Abstract]).
Regarding claim 12, claim 12 is a system claim that is substantially the same as the method of claim 5. Therefore, claim 12 is rejected for the same reasons as claim 5.
Regarding claim 19, claim 19 is a computer-readable storage medium claim that is substantially the same as the method of claim 5. Therefore, claim 19 is rejected for the same reasons as claim 5.
Claims 6 & 13 are rejected under 35 U.S.C. 103 as being unpatentable over Joseph, in view of Sun, in view of You, as applied in claim 1, and further in view of Hamilton et al. ("Inductive Representation Learning on Large Graphs," 31st Conference on Neural Information Processing Systems (NIPS 2017), Long Beach, CA, USA, December 2017, 19 pages), hereafter Yi.
Hamilton was cited in the IDS submitted 02/20/2024.
Regarding dependent claim 6, the combination of Joseph, Sun, and You teaches the claimed invention as claimed in claim 1, including training the heterogeneous GNN (You [Sec. 3 & Fig. 1] discusses training a heterogenous GNN).
The combination of Joseph, Sun, and You does not explicitly teach wherein the heterogeneous GNN comprises a heterogeneous GraphSAGE neural network.
However, in a similar field of endeavor, Hamilton teaches a method for generating node embeddings wherein the method utilizes a GraphSAGE neural network ([Sec. 3] discusses a method for aggregating information from a local neighborhood utilizing a heterogenous GraphSAGE neural network).
Because the combination of Joseph, Sun, and You teaches training a heterogenous GNN; and Yi teaches the utilization of a GraphSAGE neural network, accordingly, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate utilization of a GraphSAGE neural network as taught by Yi into the combination of Joseph, Sun, and You’s computer-implemented method, with a reasonable expectation of success, to teach wherein the heterogeneous GNN comprises a heterogeneous GraphSAGE neural network. This combination would have been motivated by the desire to implement an algorithm that generates embeddings by sampling and aggregating features from a node’s local neighborhood which is proven to operate well with unseen data and outperform baselines on benchmarks (Hamilton [Abstract]).
Regarding claim 13, claim 13 is a system claim that is substantially the same as the method of claim 6. Therefore, claim 13 is rejected for the same reasons as claim 6.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Schnabel et al. ("Recommendations as Treatments: Debiasing Learning and Evaluation", Proceedings of the 33rd International Conference on Machine Learning, arXiv) (Year: 2016) ([Abstract] Most data for evaluating and training recommender systems is subject to selection biases, either through self-selection by the users or through the actions of the recommendation system itself. In this paper, we provide a principled approach to handle selection biases by adapting models and estimation techniques from causal inference. The approach leads to unbiased performance estimators despite biased data, and to a matrix factorization method that provides substantially improved prediction performance on real-world data. We theoretically and empirically characterize the robustness of the approach, and find that it is highly practical and scalable).
Wang et al. ("Neural Graph Collaborative Filtering", 2019 Association for Computing Machinery, arXiv) (Year: 2020) ([Abstract] In this work, we propose to integrate the user-item interactions —more specifically the bipartite graph structure—into the embedding process. We develop a new recommendation framework Neural Graph Collaborative Filtering (NGCF), which exploits the user item graph structure by propagating embeddings on it. This leads to the expressive modeling of high-order connectivity in user item graph, effectively injecting the collaborative signal into the embedding process in an explicit manner. We conduct extensive experiments on three public benchmarks, demonstrating significant improvements over several state-of-the-art models like HOPRec [40] and Collaborative Memory Network [5]. Further analysis verifies the importance of embedding propagation for learning better user and item representations, justifying the rationality and effectiveness of NGCF).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RILEY S ACOSTA whose telephone number is (571)272-8714. The examiner can normally be reached Monday-Thursday 6am-4pm.
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, Jennifer N Welch can be reached at (571)272-7212. 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.
/RILEY S ACOSTA/Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143
Prosecution Insights