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 .
The action is in response to the application filed on 8/03/2023. Claims 1-20 are pending and have been examined.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 8/03/2023 and 3/28/2025 are in compliance with the provisions of 37 CFR 1.97, 1.98, and MPEP § 609. They has been placed in the application file, and the information referred to therein has been considered as to the merits.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-6, 8-13, and 15-20 are rejected under 35 U.S.C. 103 as being unpatentable over Xu et al. “Graphical Modeling for Multi-Source Domain Adaptation”, hereinafter “Xu” in view of Kipf et al. “Variational Graph Auto-Encoders”, hereinafter “Kipf”.
Regarding Claim 1, Xu teaches:
A computer-implemented (p. 13, col. 1, paragraph 1, “The hardware conditions for the experiments are Intel(R) Xeon(R) CPU E5-2620 v4@2.40 GHz with 8 processors and one NVIDIA TITAN Xp GPU”) method for employing a graph-based adaptive domain generation framework (p.1 Abstract, “The graphical structure among different domains is useful to tackle these challenges, in which the interdependency among various instances/categories can be effectively modeled”), the method comprising:
in a training phase:
performing domain prototypical network training on source domains (p. 8, col. 2, paragraph 1, “In the learning phase, since the prototypes are online updated, the graph construction involves the computation of all proto types and the query sample”, p. 4, col. 1, paragraph 3, “During the learning phase, the prototypes are updated… For the source domain Sm (1 m M), the estimated prototype cm k is defined as the mean embedding of all samples belonging to class k in the query sample set Sm”);
constructing an… domain relation graph… (p. 6, col. 2, paragraph 2, “G+ = (V,E+). In this network, the node set V is identical to the observation set X where the embeddings of all nodes are with the same dimension, and the edge set E+ = {(u,v)} reflects the desired relationships among observations as stated above”, p. 4, Fig. 1(b), “construct relational graph”); and
performing, via a domain-adaptive classifier, domain-adaptive classifier training to make an informed decision (p. 7, col. 2, paragraph 1, “the overall classification objective function combines two constraints as below: Lcls=Lsrc cls+Ltgt cls”, p. 2, col. 1, paragraph 4, “We then employ a graph neural network (GNN) to propagate the local messages on the graph and use a linear classifier to predict the label of each node”); and
in a testing phase, given testing samples from a new source domain (p. 4, col. 1, paragraph 1, “Given a query sample q from an arbitrary domain”):
computing a prototype by using a pretrained domain prototypical network (p. 6, col. 1, paragraph 3, “In the inference phase, given a query sample q, we first extract its embedding zq with the extractor f and combine the embedding with all prototypes to form the observation set”);
inferring node embedding; and making a prediction by the domain-adaptive classifier based on the domain node embeddings (p. 6, col. 1, paragraph 3, “Upon this graph, the GNN g and linear classifier c are consecutively applied to derive the label predictions for all nodes. Finally, we take the prediction for the node corresponding to the query sample as the output”).
Xu does not expressly teach:
constructing an autoencoder domain relation graph by applying a graph autoencoder to produce domain node embeddings
However, Kipf teaches:
constructing an autoencoder domain relation graph by applying a graph autoencoder to produce domain node embeddings (Kipf, p. 2, paragraph 2, “For a non-probabilistic variant of the VGAE model, we calculate embeddings Z and the reconstructed adjacency matrix ˆ A as follows: ˆ A=σZZ , with Z=GCN(X,A)”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use Kipf’s graph autoencoder on Xu’s domain relation graph. The motivation to do so would be to increase predictive performance by learning latent node representations (Kipf, p. 1, paragraph 1, “This model makes use of latent variables and is capable of learning interpretable latent representations for undirected graphs”, p. 1, paragraph 2, “Our model achieves competitive results on a link prediction task in citation networks. In contrast to most existing models for unsupervised learning on graph-structured data and link prediction [5, 6, 7, 8], our model can naturally incorporate node features, which significantly improves predictive performance”).
Regarding Claim 2, Xu in view of Kipf teaches the method of Claim 1. Xu further teaches:
wherein the domain prototypical network training results in extracting domain-specific features and capturing a similarity between the source domains (Xu, p. 4, col. 2, paragraph 4, “by the embedding vectors with the same dimension, and the edge set E = f(u; v;Auv)g describes the relations among observations”, p. 5, col. 1, paragraph 1, “Auv=K(Xu,Xv)=exp – (||Xu−Xv||2/2 2σ2) , (4) where Xu and Xv stand for the embedding of node u and v”, The nodes include prototype embeddings per source domain, similarity is through distance calculation between nodes when finding Auv which captures the similarity between source domains when Auv is found for all nodes each from a specific source domain”).
Regarding Claim 3, Xu in view of Kipf teaches the method of Claim 1. Xu further teaches:
wherein prior knowledge vectors serve as an initial node embedding for each source domain (initial node embeddings from source domains before further processing from Kipf’s graph autoencoder are estimated prototypes, Xu, p. 4, col. 1, paragraph 3, “the estimated prototype cmk is defined as the mean embedding of all samples belonging to class k in the query sample set Sm”).
Regarding Claim 4, Xu in view of Kipf teaches the method of Claim 1. Xu further teaches:
wherein an edge weight is computed according to a similarity between two source domains (Xu, p. 4, col. 2, paragraph 4, “where Auv denotes the adjacency weight between node u and v”, p. 5, Equation 4, similarity is found through distance calculation where nodes u and v can be from different source domains).
Regarding Claim 5, Xu in view of Kipf teaches the method of Claim 1. Xu further teaches:
wherein the inferring of new embeddings includes adding a new node to an existing domain graph to represent the new source domain (Xu, p. 7, col. 1, paragraph 1, “target domain is regarded as the (M +1)-th domain”, p. 6, col. 1, paragraph 3, “In the inference phase, given a query sample q, we first extract its embedding zq with the extractor f and combine the embedding with all prototypes to form the observation set… a graph G is constructed over the observations”, p. 4, col. 1, paragraph 1, “Given a query sample q from an arbitrary domain”).
Regarding Claim 6, Xu in view of Kipf teaches the method of Claim 1. Kipf further teaches:
wherein the graph autoencoder is trained to produce the node embeddings used to infer a linkage of each edge (Kipf, p. 2, paragraph 4, “The models are trained on an incomplete version of these datasets where parts of the citation links (edges) have been removed, while all node features are kept. We form validation and test sets from previously removed edges and the same number of randomly sampled pairs of unconnected nodes (non-edges). We compare models based on their ability to correctly classify edges and non-edges”).
Regarding Claim 8, Xu teaches:
A computer program product for employing a graph-based adaptive domain generation framework (p.1 Abstract, “The graphical structure among different domains is useful to tackle these challenges, in which the interdependency among various instances/categories can be effectively modeled”), the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising (Xu uses computer hardware components to perform their experiments, demonstrating that Xu performs their method on a computer in which processor, memory, and storage devices are inherent, p. 13, col. 1, paragraph 1, “The hardware conditions for the experiments are Intel(R) Xeon(R) CPU E5-2620 v4@2.40 GHz with 8 processors and one NVIDIA TITAN Xp GPU”):
in a training phase:
performing domain prototypical network training on source domains (p. 8, col. 2, paragraph 1, “In the learning phase, since the prototypes are online updated, the graph construction involves the computation of all proto types and the query sample”, p. 4, col. 1, paragraph 3, “During the learning phase, the prototypes are updated… For the source domain Sm (1 m M), the estimated prototype cm k is defined as the mean embedding of all samples belonging to class k in the query sample set Sm”);
constructing an… domain relation graph… (p. 6, col. 2, paragraph 2, “G+ = (V,E+). In this network, the node set V is identical to the observation set X where the embeddings of all nodes are with the same dimension, and the edge set E+ = {(u,v)} reflects the desired relationships among observations as stated above”, p. 4, Fig. 1(b), “construct relational graph”); and
performing, via a domain-adaptive classifier, domain-adaptive classifier training to make an informed decision (p. 7, col. 2, paragraph 1, “the overall classification objective function combines two constraints as below: Lcls=Lsrc cls+Ltgt cls”, p. 2, col. 1, paragraph 4, “We then employ a graph neural network (GNN) to propagate the local messages on the graph and use a linear classifier to predict the label of each node”); and
in a testing phase, given testing samples from a new source domain (p. 4, col. 1, paragraph 1, “Given a query sample q from an arbitrary domain”):
computing a prototype by using a pretrained domain prototypical network (p. 6, col. 1, paragraph 3, “In the inference phase, given a query sample q, we first extract its embedding zq with the extractor f and combine the embedding with all prototypes to form the observation set”);
inferring node embedding; and making a prediction by the domain-adaptive classifier based on the domain node embeddings (p. 6, col. 1, paragraph 3, “Upon this graph, the GNN g and linear classifier c are consecutively applied to derive the label predictions for all nodes. Finally, we take the prediction for the node corresponding to the query sample as the output”).
Xu does not expressly teach:
constructing an autoencoder domain relation graph by applying a graph autoencoder to produce domain node embeddings
However, Kipf teaches:
constructing an autoencoder domain relation graph by applying a graph autoencoder to produce domain node embeddings (Kipf, p. 2, paragraph 2, “For a non-probabilistic variant of the VGAE model, we calculate embeddings Z and the reconstructed adjacency matrix ˆ A as follows: ˆ A=σZZ , with Z=GCN(X,A)”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use Kipf’s graph autoencoder on Xu’s domain relation graph. The motivation to do so would be to increase predictive performance by learning latent node representations (Kipf, p. 1, paragraph 1, “This model makes use of latent variables and is capable of learning interpretable latent representations for undirected graphs”, p. 1, paragraph 2, “Our model achieves competitive results on a link prediction task in citation networks. In contrast to most existing models for unsupervised learning on graph-structured data and link prediction [5, 6, 7, 8], our model can naturally incorporate node features, which significantly improves predictive performance”).
Regarding Claim 9, the rejection of Claim 8 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 2.
Regarding Claim 10, the rejection of Claim 8 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 3.
Regarding Claim 11, the rejection of Claim 8 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 4.
Regarding Claim 12, the rejection of Claim 8 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 5.
Regarding Claim 13, the rejection of Claim 8 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 6.
Regarding Claim 15, Xu teaches:
A computer processing system for employing a graph-based adaptive domain generation framework (p.1 Abstract, “The graphical structure among different domains is useful to tackle these challenges, in which the interdependency among various instances/categories can be effectively modeled”), comprising: a memory device for storing program code; and a processor device, operatively coupled to the memory device, for running the program code to (Xu uses computer hardware components to perform their experiments, demonstrating that Xu performs their method on a computer in which processor, memory, and storage devices are inherent, p. 13, col. 1, paragraph 1, “The hardware conditions for the experiments are Intel(R) Xeon(R) CPU E5-2620 v4@2.40 GHz with 8 processors and one NVIDIA TITAN Xp GPU”):
in a training phase:
perform domain prototypical network training on source domains (p. 8, col. 2, paragraph 1, “In the learning phase, since the prototypes are online updated, the graph construction involves the computation of all proto types and the query sample”, p. 4, col. 1, paragraph 3, “During the learning phase, the prototypes are updated… For the source domain Sm (1 m M), the estimated prototype cm k is defined as the mean embedding of all samples belonging to class k in the query sample set Sm”);
construct an… domain relation graph… (p. 6, col. 2, paragraph 2, “G+ = (V,E+). In this network, the node set V is identical to the observation set X where the embeddings of all nodes are with the same dimension, and the edge set E+ = {(u,v)} reflects the desired relationships among observations as stated above”, p. 4, Fig. 1(b), “construct relational graph”); and
perform, via a domain-adaptive classifier, domain-adaptive classifier training to make an informed decision (p. 7, col. 2, paragraph 1, “the overall classification objective function combines two constraints as below: Lcls=Lsrc cls+Ltgt cls”, p. 2, col. 1, paragraph 4, “We then employ a graph neural network (GNN) to propagate the local messages on the graph and use a linear classifier to predict the label of each node”); and
in a testing phase, given testing samples from a new source domain (p. 4, col. 1, paragraph 1, “Given a query sample q from an arbitrary domain”):
compute a prototype by using a pretrained domain prototypical network (p. 6, col. 1, paragraph 3, “In the inference phase, given a query sample q, we first extract its embedding zq with the extractor f and combine the embedding with all prototypes to form the observation set”);
infer node embedding; and make a prediction by the domain-adaptive classifier based on the domain node embeddings (p. 6, col. 1, paragraph 3, “Upon this graph, the GNN g and linear classifier c are consecutively applied to derive the label predictions for all nodes. Finally, we take the prediction for the node corresponding to the query sample as the output”).
Xu does not expressly teach:
constructing an autoencoder domain relation graph by applying a graph autoencoder to produce domain node embeddings
However, Kipf teaches:
constructing an autoencoder domain relation graph by applying a graph autoencoder to produce domain node embeddings (Kipf, p. 2, paragraph 2, “For a non-probabilistic variant of the VGAE model, we calculate embeddings Z and the reconstructed adjacency matrix ˆ A as follows: ˆ A=σZZ , with Z=GCN(X,A)”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to use Kipf’s graph autoencoder on Xu’s domain relation graph. The motivation to do so would be to increase predictive performance by learning latent node representations (Kipf, p. 1, paragraph 1, “This model makes use of latent variables and is capable of learning interpretable latent representations for undirected graphs”, p. 1, paragraph 2, “Our model achieves competitive results on a link prediction task in citation networks. In contrast to most existing models for unsupervised learning on graph-structured data and link prediction [5, 6, 7, 8], our model can naturally incorporate node features, which significantly improves predictive performance”).
Regarding Claim 16, the rejection of Claim 15 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 2.
Regarding Claim 17, the rejection of Claim 15 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 3.
Regarding Claim 18, the rejection of Claim 15 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 4.
Regarding Claim 19, the rejection of Claim 15 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 5.
Regarding Claim 20, the rejection of Claim 15 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 6.
Claims 7 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Xu in view of Kipf, further in view of Yue et al. “Prototypical Cross-domain Self-supervised Learning for Few-shot Unsupervised Domain Adaptation”, hereinafter “Yue”.
Regarding Claim 7, Xu in view of Kipf teaches the method of Claim 1. Xu further teaches:
wherein the pretrained domain prototypical network learns a representation of each source domain by training on …data from the source domains (Xu, p. 4, col. 1, paragraph 2, “query samples from various domains are given, and these samples are mapped to the latent space by a feature extractor to update prototypes”, p. 6, col. 1, paragraph 2, “the overall learning objective with respect to feature extractor f, GNN g and classifier c is defined as below: min f;g;c Lcls + _1Lglobal + _2Llocal”).
Xu in view of Kipfs does not expressly teach:
…unlabeled data…
However, Yue teaches:
…unlabeled data… (Yue, p. 13836, col. 2, paragraph 1, “we are given a very limited number of labeled source images Ds = {(xsi, ysi )}Nsi=1, as well as unlabeled source images Dsu = {(xsui )}Nsui=1”, p. 13838, col. 1, paragraph 4, “We use the few-shot labeled data as well as samples with high-confident predictions to estimate the prototype for each class”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine Yue’s class assignment of unlabeled data using few shot labeled data with Xu’s source domain prototype training. The motivation to do so would to be able to use Xu’s method in source domains where it is expensive or impractical to collect labels (Yue, p. 13834, “In some applications, however, it is expensive even to collect labels in the source domain, making most previous works impractical” and Figure 1 description, “We address the task of few-shot unsupervised domain adaptation”).
Regarding Claim 14, the rejection of Claim 8 is incorporated and further, the claim is rejected
for the same reasons as set forth in Claim 7.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JESSE CHEN COULSON whose telephone number is (571)272-4716. The examiner can normally be reached Monday-Friday 8:30-5:30.
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, Kakali Chaki can be reached at (571) 272-3719. 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.
/JESSE C COULSON/
Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights