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 in response to the application and claims filed 06/27/2024. Claims 1-9 are pending and have been examined. Claims 1-9 are rejected.
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The present application claims foreign priority based on Korean application KR10-2024-0010364 filed 23 Jan 2024. The examiner notes that a certified copy (in Korean) of the above-noted application was retrieved on 08/06/2024. Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 06/06/2025, 02/20/2025, 10/18/2024, 06/27/2024 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claim 2 is objected to because of the following informalities:
“the unconnected node pair”. Draws implicit antecedent from claim 1’s “at least one node pair unconnected in the structure…”. Suggestion: recite “the at least one unconnected node pair” in claim 2 to fix the singular/plural mismatch.
Appropriate correction is required.
Claim 7 is objected to because of the following informalities:
“the randomly sampled edges” in claim (ii) reads back to the edges sampled in clause (i), but the spec uses two distinct sample sets (ℰ′ᵤ for the dual loss, ℰ″ᵤ for the correction loss). See specification page 13, equations 3 and 4.
Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim 7 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The term “excessive” in claim 7 is a relative term which renders the claim indefinite. The term “excessive” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-9 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 method; therefore, it is directed to the statutory category of processes.
Step 2A Prong 1: The claim recites the following abstract ideas:
"predicting one or more edges having a probability of being connected in a structure of an edge-incomplete graph (...)": This limitation is a mental process because a person mentally or with a pen and paper can observe the structure of a graph in which one or more edges are missing and predict which node pairs have a probability of being connected.
"(…) performs binary classification by processing at least one edge observed in the structure of the edge-incomplete graph as positive data and processing at least one node pair unconnected in the structure of the edge-incomplete graph as unlabeled data.": This limitation is a mental process because a person mentally or with a pen and paper can perform binary classification by designating (labeling) at least one edge observed in the structure of the graph as positive data and designating (labeling) at least one node pair unconnected in the structure of the graph as unlabeled data.
Step 2A Prong 2: This judicial exception is not integrated into a practical application because the additional elements are as follows:
"A link prediction method, the link prediction method being performed by a link prediction apparatus, the link prediction method comprising:": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim recites a generic, off the shelf "link prediction apparatus" as a tool to perform the recited abstract ideas.
"...by entering the edge-incomplete graph into a link prediction model;": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model. The "link prediction model" is interpreted as applying a generic model on the abstract idea of "predicting one or more edges having a probability of being connected."
"wherein the link prediction model is a model that...": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The "link prediction model" in the claim is interpreted as applying a generic, off the shelf model on the abstract idea of "(...) performs binary classification."
Step 2B:
"A link prediction method, the link prediction method being performed by a link prediction apparatus, the link prediction method comprising:": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim recites a generic, off the shelf "link prediction apparatus" as a tool to perform the recited abstract ideas.
"...by entering the edge-incomplete graph into a link prediction model;": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model. The "link prediction model" is interpreted as applying a generic model on the abstract idea of "predicting one or more edges having a probability of being connected."
"wherein the link prediction model is a model that...": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The "link prediction model" in the claim is interpreted as applying a generic, off the shelf model on the abstract idea of "(...) performs binary classification."
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 2
Step 1: A process, as above.
Step 2A Prong 1: See the rejection of Claim 1 above, which claim 2 depends on. Claim 2 further recites:
"(...) a parameter (...) is updated by using an expected edge-incomplete graph to which a random variable representing a connection state of the unconnected node pair in the structure of the edge-incomplete graph is applied.": This limitation is a mental process. A person mentally or with a pen and paper can apply a random variable representing a connection state (e.g., connected or unconnected) to the unconnected node pair of the graph, form the resulting expected graph, and update a parameter value by using the expected graph.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites:
"wherein the link prediction model is a model in which a parameter of the link prediction model is updated (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
Step 2B:
"wherein the link prediction model is a model in which a parameter of the link prediction model is updated (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 3
Step 1: A process, as above.
Step 2A Prong 1: See the rejection of Claim 2 above, which claim 3 depends on. Claim 3 further recites:
"(...) the edge-incomplete graph is converted into a line graph in which two adjacent edges in the structure of the edge-incomplete graph are represented by two connected nodes...": This limitation is a mental process because a person mentally or with a pen and paper can convert a graph into a line graph by drawing a node for each edge of the original graph and connecting two drawn nodes whenever the corresponding two edges are adjacent in the original graph.
"...and an expectation for the random variable is computed using a Markov network obtained by modeling a joint probability distribution of nodes of the resulting line graph.": This limitation falls within the mathematical concepts grouping because it involves computing an expectation, i.e., performing a mathematical calculation, using a joint probability distribution that is modeled over the nodes of the line graph.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites:
"wherein the link prediction model is a model in which (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
Step 2B:
"wherein the link prediction model is a model in which (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 4
Step 1: A process, as above.
Step 2A Prong 1: See the rejection of Claim 2 above, which claim 4 depends on. Claim 4 further recites:
"(...) a structure of the expected edge-incomplete graph is approximated in such a manner as to set a number of edges to be maintained within the structure of the expected edge-incomplete graph and not connect remaining node pairs except those having a higher probability of being connected than a reference value.": This limitation is a mental process because a person mentally or with a pen and paper can approximate a graph by setting a number of edges to be kept in the graph, comparing the probability of being connected of each remaining node pair with a reference value, and not connecting the remaining node pairs whose probability does not exceed the reference value.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites:
"wherein the link prediction model is a model in which (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
Step 2B:
"wherein the link prediction model is a model in which (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 5
Step 1: A process, as above.
Step 2A Prong 1: See the rejection of Claim 2 above, which claim 5 depends on.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites:
"wherein the link prediction model is a model that is trained by propagating information in a graph convolutional network of the link prediction model using the expected edge-incomplete graph.": This additional element recites a mere instruction to apply an exception with a recitation of the words "apply it" (or an equivalent) as identified in MPEP 2106.05(f), and does not provide integration into a practical application. -- Examiner's Note (EN): The claim denotes generic training and a generic graph convolutional network with no additional details or limitations beyond a generic, off the shelf graph convolutional network; the limitation amounts to no more than mere instructions to apply the abstract ideas recited in the rejection of claim 2 above using a generic computer component.
Step 2B:
"wherein the link prediction model is a model that is trained by propagating information in a graph convolutional network of the link prediction model using the expected edge-incomplete graph.": This additional element recites a mere instruction to apply an exception with a recitation of the words "apply it" (or an equivalent) as identified in MPEP 2106.05(f), and does not provide integration into a practical application. -- Examiner's Note (EN): The claim denotes generic training and a generic graph convolutional network with no additional details or limitations beyond a generic, off the shelf graph convolutional network; the limitation amounts to no more than mere instructions to apply the abstract ideas recited in the rejection of claim 2 above using a generic computer component.
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 6
Step 1: A process, as above.
Step 2A Prong 1: See the rejection of Claim 2 above, which claim 6 depends on. Claim 6 further recites:
"(...) the random variable of the expected edge-incomplete graph is updated by using a prediction probability (...)": This limitation is a mental process. A person mentally or with a pen and paper can update the value of the random variable of the expected graph by using a prediction probability, i.e., write down a new value for the random variable based on the probability.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites:
"wherein the link prediction model is a model in which (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
"...a prediction probability output by the link prediction model.": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): This limitation amounts to using a generic link prediction model to output data (a prediction probability), which is merely an instruction to apply the abstract idea using a generic computer component.
Step 2B:
"wherein the link prediction model is a model in which (...)": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model; the limitation amounts to no more than an instruction to apply the abstract ideas recited in step 2A prong 1 using a generic model.
"...a prediction probability output by the link prediction model.": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): This limitation amounts to using a generic link prediction model to output data (a prediction probability), which is merely an instruction to apply the abstract idea using a generic computer component.
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 7
Step 1: A process, as above.
Step 2A Prong 1: See the rejection of Claim 2 above, which claim 7 depends on. Claim 7 further recites:
"(...) (i) a dual loss function to which one or more randomly sampled edges are applied in order to strike a balance between a number of connected edges and a number of unconnected edges in the structure of the edge-incomplete graph by taking into consideration one or more added edges in the expected edge-incomplete graph...": This limitation falls within the mathematical concepts grouping because the dual loss function is a mathematical equation (shown as Equation 3 of the specification) and the limitation involves evaluating the mathematical equation to which the randomly sampled edges, the numbers of connected and unconnected edges, and the added edges are applied as variables.
"...and (ii) a correction loss function which prevents excessive self-reinforcement based on the randomly sampled edges by taking into consideration one or more added edges in the expected edge-incomplete graph.": This limitation falls within the mathematical concepts grouping because the correction loss function is a mathematical equation that measures binary cross-entropy for edges (shown as Equation 4 of the specification) and the limitation involves evaluating the mathematical equation to which the randomly sampled edges and the added edges are applied as variables.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites:
"wherein the link prediction model is a model that is trained according to (...)": This additional element recites a mere instruction to apply an exception with a recitation of the words "apply it" (or an equivalent) as identified in MPEP 2106.05(f), and does not provide integration into a practical application. -- Examiner's Note (EN): This limitation amounts to performing the abstract ideas (as recited in step 2A prong 1) and applying them to do generic training of a generic, off the shelf link prediction model.
Step 2B:
"wherein the link prediction model is a model that is trained according to (...)": This additional element recites a mere instruction to apply an exception with a recitation of the words "apply it" (or an equivalent) as identified in MPEP 2106.05(f), and does not provide integration into a practical application. -- Examiner's Note (EN): This limitation amounts to performing the abstract ideas (as recited in step 2A prong 1) and applying them to do generic training of a generic, off the shelf link prediction model.
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 8
Step 1: The claim recites an apparatus; therefore, it is directed to the statutory category of machine.
Step 2A Prong 1: The claim recites the following abstract ideas:
"(...) predict one or more edges having a probability of being connected in a structure of the edge-incomplete graph (...)": This limitation is a mental process because a person mentally or with a pen and paper can observe the structure of a graph in which one or more edges are missing and predict which node pairs have a probability of being connected.
"(...) performs binary classification by processing at least one edge observed in the structure of the edge-incomplete graph as positive data and processing at least one node pair unconnected in the structure of the edge-incomplete graph as unlabeled data.": This limitation is a mental process because a person mentally or with a pen and paper can perform binary classification by designating (labeling) at least one edge observed in the structure of the graph as positive data and designating (labeling) at least one node pair unconnected in the structure of the graph as unlabeled data.
Step 2A Prong 2: This judicial exception is not integrated into a practical application because the additional elements are as follows:
"A link prediction apparatus comprising: memory configured to store an edge-incomplete graph and a link prediction model; and a controller configured to...": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim only recites generic, off the shelf memory and a generic, off the shelf controller performing the generic computer functions of storing data and executing the recited abstract ideas; this amounts to no more than mere instructions to apply the exception using generic computer components.
"...by entering the edge-incomplete graph into the link prediction model;": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model. The "link prediction model" is interpreted as applying a generic model on the abstract idea of "predict one or more edges having a probability of being connected."
"wherein the link prediction model is a model that...": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The "link prediction model" in the claim is interpreted as applying a generic, off the shelf model on the abstract idea of "(...) performs binary classification."
Step 2B:
"A link prediction apparatus comprising: memory configured to store an edge-incomplete graph and a link prediction model; and a controller configured to...": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim only recites generic, off the shelf memory and a generic, off the shelf controller performing the generic computer functions of storing data and executing the recited abstract ideas; this amounts to no more than mere instructions to apply the exception using generic computer components.
"...by entering the edge-incomplete graph into the link prediction model;": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim denotes a generic link prediction model with no additional details or limitations beyond a generic, off the shelf model. The "link prediction model" is interpreted as applying a generic model on the abstract idea of "predict one or more edges having a probability of being connected."
"wherein the link prediction model is a model that...": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The "link prediction model" in the claim is interpreted as applying a generic, off the shelf model on the abstract idea of "(...) performs binary classification."
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim 9
Step 1: The claim recites a non-transitory computer-readable storage medium; therefore, it is directed to the statutory category of manufacture.
Step 2A Prong 1: Claim 9 recites a program that, when executed by a processor, causes the processor to execute the method set forth in claim 1; therefore, claim 9 recites the same abstract ideas as claim 1. See the rejection of Claim 1 above.
Step 2A Prong 2: This judicial exception is not integrated into a practical application. The claim further recites:
"A non-transitory computer-readable storage medium having stored thereon a program that, when executed by a processor, causes the processor to execute the method set forth in claim 1.": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim recites a generic, off the shelf storage medium, a generic program, and a generic, off the shelf processor as tools to perform the abstract ideas recited in the rejection of claim 1 above.
Step 2B:
"A non-transitory computer-readable storage medium having stored thereon a program that, when executed by a processor, causes the processor to execute the method set forth in claim 1.": Adding the words "apply it" (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (MPEP 2106.05(f)). -- Examiner's Note (EN): The claim recites a generic, off the shelf storage medium, a generic program, and a generic, off the shelf processor as tools to perform the abstract ideas recited in the rejection of claim 1 above.
The additional elements considered individually or in combination do not amount to significantly more than the judicial exception. Therefore, the claim is not patent eligible.
Claim Rejections - 35 USC § 102
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 the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Examiner’s Note: Some rejections will include an Examiner’s Note (labeled ‘EN’) to provide additional context or rationale explaining the basis for the rejection.
Claims 1, 8 and 9 are rejected as being anticipated by Gan et al., (“Positive-Unlabeled Learning for Network Link Prediction,”), hereinafter “Gan”
Claim 1
Gan teaches,
A link prediction method, the link prediction method being performed by a link prediction apparatus, the link prediction method comprising: (Abstract, p. 1, "Link prediction is an important problem in network data mining, which is dedicated to predicting the potential relationship between nodes in the network... To address this problem, we propose a positive-unlabeled learning framework with network representation for network link prediction only using positive samples and unlabeled samples." Section 3.1, p. 8, "The source code and data are available at https://github.com/naodandandan/PU-for-link-prediction (accessed on 10 September 2022)." - EN: Gan's framework is a link prediction method performed in software; the computer that executes that software is the "link prediction apparatus" that performs the method.)
predicting one or more edges having a probability of being connected in a structure of an edge-incomplete graph by entering the edge-incomplete graph into a link prediction model; (Section 2.2, p. 4, "We first input the network data into the network representation module (Figure 2a) to obtain the representation vector for each node. Then, we concatenate representation vectors of node pairs and feed them into different classifiers to make binary classification (i.e., to predict whether the link between a node pair exist or not)." Section 2.4.2, p. 6, "We present a bagging positive-unlabeled learning, whose goal is to obtain a function that can give the probability that an instance in the data set belongs to a positive instance." Section 1, p. 2, "A satisfying representation of a network is expected to have a good ability to capture inherent structures of the network for predicting possible but unobserved links" - EN: Gan's input network contains only observed edges while real links between its nodes may be unobserved, so it is a graph with missing edges, which reads on "an edge-incomplete graph". The "link prediction model" corresponds to Gan's framework of the network representation module and the positive-unlabeled classifier; Gan enters the graph into this model by inputting the network data into the framework, and the classifier's output probability that a node pair is a positive instance constitutes predicting "one or more edges having a probability of being connected".)
wherein the link prediction model is a model that performs binary classification by processing at least one edge observed in the structure of the edge-incomplete graph as positive data and processing at least one node pair unconnected in the structure of the edge-incomplete graph as unlabeled data. (Section 1, p. 1, "Generally, node pairs with links in the network are considered as positive samples, while ones without links are negative samples. However, these negative samples are not always accurate because the links between nodes may not have been observed. Thus, the unobserved links need to be regarded as unlabeled samples, and their labels may be positive or negative." Section 1, p. 2, "In order to solve the above problem, positive-unlabeled (PU) learning is proposed to learn a binary classifier based on positive data and unlabeled data" Section 2.2, p. 4, "Then, we concatenate representation vectors of node pairs and feed them into different classifiers to make binary classification (i.e., to predict whether the link between a node pair exist or not)." - EN: Gan's classifiers perform the recited "binary classification". Gan's positive samples are the node pairs with links, i.e., the edges observed in the network, which reads on processing "at least one edge observed in the structure of the edge-incomplete graph as positive data"; Gan's unlabeled samples are the node pairs without observed links, which reads on processing "at least one node pair unconnected in the structure of the edge-incomplete graph as unlabeled data".)
Claim 8
Gan teaches,
A link prediction apparatus comprising: (Abstract, p. 1, "To address this problem, we propose a positive-unlabeled learning framework with network representation for network link prediction only using positive samples and unlabeled samples." - EN: the computer that runs Gan's software framework is a "link prediction apparatus".)
memory configured to store an edge-incomplete graph and a link prediction model; and (Section 2.1, p. 3, "The input of our framework is a homogeneous network G = (V, E)" Section 2.1, p. 3, "The network G can be represented by an adjacency matrix" Section 3.1, p. 7, "We select three different types of network data sets shown in Table 1, including the bioinformatic network DrugBank, the social network Karate and the citation network Cora." - EN: Gan's network data sets are stored data representing the graph, which Gan represents as an adjacency matrix, and Gan's source code, which implements the framework, is stored software; the computer memory that holds this data and code reads on "memory configured to store an edge-incomplete graph and a link prediction model".)
a controller configured to predict one or more edges having a probability of being connected in a structure of the edge-incomplete graph by entering the edge-incomplete graph into the link prediction model; (Section 2.2, p. 4, "We first input the network data into the network representation module (Figure 2a) to obtain the representation vector for each node. Then, we concatenate representation vectors of node pairs and feed them into different classifiers to make binary classification (i.e., to predict whether the link between a node pair exist or not)." - EN: the processor that executes Gan's framework to make these predictions reads on "a controller" configured to perform the recited predicting, which is addressed at claim 1.)
Additional Examiner's Note: Gan discloses an executed software implementation of the framework: Gan publishes the source code and data of the framework and reports experimental results obtained by training classifiers on the Table 1 network data sets and evaluating them under k-fold cross-validation. Performing those disclosed experiments necessarily uses a computing apparatus having memory that stores the network data and the framework and a processor that performs the prediction; the recited "memory" and "controller" are therefore necessarily present in Gan's disclosure, not merely probable. See the Abstract and Sections 2.2, 3.1, and 3.2 of Gan.
The remaining limitations of claim 8 are substantially the same as method claim 1, therefore claim 8 is rejected under the same rationale as claim 1.
Claim 9
Gan teaches,
A non-transitory computer-readable storage medium having stored thereon a program that, when executed by a processor, causes the processor to execute the method set forth in claim 1. (Section 3.1, p. 8, "The source code and data are available at https://github.com/naodandandan/PU-for-link-prediction (accessed on 10 September 2022)." - EN: Gan's published source code is "a program" that implements the link prediction method mapped at claim 1; the repository storage that holds the retrievable code files is "a non-transitory computer-readable storage medium"; and Gan's reported results are produced by a processor executing that program.)
Additional Examiner's Note: Code and data that are stored in and retrievable from a repository are necessarily held on a non-transitory storage medium, and Gan's experimental results are necessarily produced by executing the program on a processor. These features are necessarily present in Gan's disclosure, not merely probable. See Sections 3.1 and 3.2 of Gan.
The method set forth in claim 1 is addressed at claim 1 above, therefore claim 9 is rejected under the same rationale as claim 1.
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.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Examiner’s Note: Some rejections will include an Examiner’s Note (labeled ‘EN’) to provide additional context or rationale explaining the basis for the rejection.
Claims 2 and 4-6 are rejected under 35 U.S.C. 103 as being unpatentable over Gan et al. (“Positive-Unlabeled Learning for Network Link Prediction,”), hereinafter “Gan”, as applied to claim 1 above, and further in view of Zhao et al. (“Data Augmentation for Graph Neural Networks,”), hereinafter “Zhao”.
Claim 2
Regarding claim 2, Gan teaches the method of claim 1 as shown above, including the recited link prediction model. Zhao further teaches:
The method of claim 1, wherein the link prediction model is a model in which a parameter of the link prediction model is updated by using an expected edge-incomplete graph to which a random variable representing a connection state of the unconnected node pair in the structure of the edge-incomplete graph is applied. (Section 4.1, “we use an edge predictor function to obtain edge probabilities for all possible and existing edges in G” Section 4.2, “To prevent the edge predictor from arbitrarily deviating from original graph adjacency, we interpolate the predicted M with the original A to derive an adjacency P.” Section 4.2, “In the edge sampling phase, we sparsify P with Bernoulli sampling on each edge to get the graph variant adjacency A.” Fig. 3, “GAug-O is comprised of three main components: (1) a differentiable edge predictor which produces edge probability estimates, (2) an interpolation and sampling step which produces sparse graph variants, and (3) a GNN which learns embeddings for node classification using these variants. The model is trained end-to-end with both classification and edge prediction losses.” - EN: Zhao’s interpolated adjacency P assigns a probability of being connected to every node pair, including the node pairs that have no edge in the given graph, and Zhao’s Bernoulli sampling on each edge applies to each such node pair a binary random variable whose sampled value is that pair’s connection state. A graph whose entries are connection probabilities is an expectation of the graph structure, so P with its per-edge Bernoulli variables reads on the “expected edge-incomplete graph to which a random variable representing a connection state of the unconnected node pair… is applied”. Zhao trains the edge predictor and the GNN end-to-end on the graph variants sampled from P, so the parameters of the model are updated by using the expected edge-incomplete graph. In the combination, Zhao’s training scheme is applied to the link prediction model of Gan, and the given graph is Gan’s edge-incomplete network as mapped at claim 1. See the motivation for combining below.)
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to combine the positive-unlabeled link prediction framework of Gan with the interpolated edge probabilities and sampled graph variants of Zhao during the training step. The motivation for doing so would be to train the model over the likely missing edges of the incomplete network rather than over the observed edges alone, so that information from node pairs whose links were not observed also reaches the model. Gan itself recognizes that node pairs without observed links may hide true links, so supplying the probable missing edges reduces the contamination of Gan’s unlabeled training data. As Zhao elaborates regarding the benefit of this augmentation methodology in Section 3.2, “strategically adding edges between nodes of the same group (intra-class) and removing edges between those in different groups (inter-class) substantially improves node classification test performance”.
Claim 4
Regarding claim 4, Gan in view of Zhao teaches the method of claim 2 as shown above, including the link prediction model taught by Gan and the expected edge-incomplete graph taught by Zhao. Zhao further teaches:
The method of claim 2, wherein the link prediction model is a model in which a structure of the expected edge-incomplete graph is approximated in such a manner as to set a number of edges to be maintained within the structure of the expected edge-incomplete graph and not connect remaining node pairs except those having a higher probability of being connected than a reference value. (Section 4.1, “GAug-M deterministically adds the top i|E| non-edges with highest edge probabilities, and removes the j|E| existing edges with least edge probabilities from G to produce G_m, where i,j ∈ [0,1].” Section 3.3, “we can either (1) apply one or multiple graph transformation operation f: G → G_m, such that G_m replaces G for both training and inference” - EN: Zhao’s G_m is a single deterministic graph constructed from the edge probabilities, so G_m approximates the structure of the probability-weighted expected edge-incomplete graph mapped at claim 2. The budgets i|E| and j|E| set the number of edges that are added to and maintained within G_m, and only the top-ranked non-edges become connected. Under the broadest reasonable interpretation (BRI) of the claim language, the edge probability of the lowest-ranked added non-edge is the recited “reference value”: every remaining node pair, whose probability does not exceed that value, is not connected.)
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to combine the expected edge-incomplete graph training of Gan in view of Zhao with the deterministic edge selection of Zhao, in which set numbers of the highest-probability non-edges are added and the lowest-probability edges are removed, when constructing the training graph. The motivation for doing so would be to admit only the most probable predicted edges into the training graph so that low-confidence predictions do not degrade it. As Zhao elaborates regarding the efficiency of this deterministic modification methodology in Section 4.1, “GAug-M shares the same time and space complexity as its associated GNN architecture during training/inference”.
Claim 5
Regarding claim 5, Gan in view of Zhao teaches the method of claim 2 as shown above, including the link prediction model, which is taught by Gan as set forth in the rejection of claim 1, and the expected edge-incomplete graph taught by Zhao. Zhao further teaches:
The method of claim 2, wherein the link prediction model is a model that is trained by propagating information in a graph convolutional network of the link prediction model using the expected edge-incomplete graph. (Section 4.2, “The graph variant adjacency A’ is passed along with node features X to the GNN node classifier.” Section 4.2, “each training iteration exposes the node-classifier to a new augmented graph variant” Section 3.1, “In this work, we use the well-known graph convolutional network (GCN)… as an example when explaining GNNs” Section 4.1, “M = σ(ZZ^T), where Z = f_GCL^(1)(A, f_GCL^(0)(A, X)).” - EN: Zhao’s node classifier is a GNN whose named implementation is the graph convolutional network, and Zhao’s f_GCL are graph convolution layers. Passing the sampled variant adjacency A’ to the GNN trains the model by propagating information through the graph that contains the added edges of the expected edge-incomplete graph rather than through the given graph alone. In the combination, this graph convolutional network is part of the link prediction model that is trained as set forth at claim 2.)
It would have been obvious to one of ordinary skill in the art (POSITA) prior to the effective filing date of the claimed invention to implement the link prediction model of Gan using the graph convolutional network (GCN) training framework of Zhao, wherein information is propagated through an augmented/expected graph containing probable added edges. A POSITA would have been motivated to combine these teachings because, as recognized by Zhao, real-world graphs suffer from partial observation and missing links, creating a gap between the observed graph and the ideal underlying connectivity. By propagating information through an expected graph structure with added plausible edges rather than the incomplete observed graph alone, the GCN allows neighborhood message passing across likely connections, thereby improving parameter inference, preventing overfitting to missing data, and producing more accurate node representations for downstream link prediction. See Zhao Section 3.2, p. 3, “In reality, a processed or observed graph may not exactly align with the process it intended to model... Noise can also be induced by partial observation: e.g. a friend recommendation system which never suggests certain friends to an end-user, thus preventing link formation... All these scenarios can produce a gap between the 'observed graph' and the so-called 'ideal graph'... Enabling an inference engine to bridge this gap suggests the promise of data augmentation via edge manipulation. In the best case, we can produce a graph
G
i
(ideal connectivity), where supposed (but missing) links are added…”
Claim 6
Regarding claim 6, Gan in view of Zhao teaches the method of claim 2 as shown above, including the link prediction model taught by Gan and the random variable of the expected edge-incomplete graph taught by Zhao. Zhao further teaches:
The method of claim 2, wherein the link prediction model is a model in which the random variable of the expected edge-incomplete graph is updated by using a prediction probability output by the link prediction model. (Section 4.2, p. 5, “GAug-O ... utilizes both edge prediction and node-classification losses to iteratively improve augmentation capacity of the edge predictor and classification capacity of the node classifier GNN. Figure 3 shows the overall architecture: each training iteration exposes the node-classifier to a new augmented graph variant... To prevent the edge predictor from arbitrarily deviating from original graph adjacency, we interpolate the predicted M with the original A to derive an adjacency P. In the edge sampling phase, we sparsify P with Bernoulli sampling on each edge to get the graph variant adjacency A’...
P
i
j
=
α
M
i
j
+
(
1
-
α
)
A
i
j
”; Fig. 3, p. 5, “GAug-O is comprised of three main components: (1) a differentiable edge predictor which produces edge probability estimates, (2) an interpolation and sampling step which produces sparse graph variants... The model is trained end-to-end with both classification and edge prediction losses.” - EN: In Zhao's training iterations, the edge predictor outputs edge probabilities
M
that update the interpolated distribution
P
used for Bernoulli sampling. Thus, the random variables of the expected graph variants are updated using the prediction probabilities output by the model.)
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to modify the link prediction method of Gan to update the random variables of the expected graph using the model's current prediction probabilities as taught by Zhao in order to progressively refine edge probability estimates throughout training and conduct more stable training, thus improving the model. (Zhao, Section 4.2; Section C.1, “we pretrain both components of GAug-O to achieve more stable joint training”).
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Gan in view of Zhao as applied to claim 2 above, and further in view of Cai et al. (“Line Graph Neural Networks for Link Prediction”), hereinafter “Cai”, and Taskar et al. (“Link Prediction in Relational Data”), hereinafter “Taskar”.
Claim 3
Regarding claim 3, Gan in view of Zhao teaches the method of claim 2 as shown above, including the link prediction model taught by Gan and the random variable of the expected edge-incomplete graph taught by Zhao. Cai further teaches:
The method of claim 2, wherein the link prediction model is a model in which the edge-incomplete graph is converted into a line graph in which two adjacent edges in the structure of the edge-incomplete graph are represented by two connected nodes (…) (Section II-C, “To overcome this challenge, we propose to convert the enclosing subgraph to the line graph, which represents the adjacencies between edges of the original graph.” Section II-C, Definition 1, “The edges in the original graph G are considered as nodes in the line graph L(G). Two nodes in L(G) are connected if and only if the two corresponding links share the same node.” Introduction, p. 1, “Therefore, the link prediction task can be regarded as the node classification problem in our proposed framework.” - EN: Cai’s conversion of graph structures into a line graph where original edges become nodes and adjacent edges sharing a vertex become connected nodes corresponds to the claims converting an edge-incomplete graph into the recited line graph.)
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to combine the link prediction model of Gan in view of Zhao with the line graph conversion of Cai when representing the candidate edges of the graph. The motivation for doing so would be to learn a representation for each candidate edge directly as a node representation of the line graph, which needs fewer parameters and converges faster than pooling node representations into edge scores. As Cai elaborates regarding the benefit of this conversion methodology in the Introduction, p. 2, “Our proposed method can achieve promising performance only with graph convolution layers, and thus requires fewer parameters. In addition, the neural network consisting of graph convolution layers converges significantly faster…”
Gan, Zhao, and Cai do not explicitly teach:
and an expectation for the random variable is computed using a Markov network obtained by modeling a joint probability distribution of nodes of the resulting line graph.
However, Taskar teaches:
and an expectation for the random variable is computed using a Markov network obtained by modeling a joint probability distribution of nodes of the resulting line graph. (Section 2, “Each potential link is associated with a tuple of entity objects, but it may or may not actually exist. We denote this event using a binary existence attribute Exists, which is true if the link between the associated entities exists and false otherwise.” Abstract, “We apply the relational Markov network framework of Taskar et al. to define a joint probabilistic model over the entire link graph” Section 3, “Given a particular instantiation I of the schema, the RMN M produces an unrolled Markov network over the attributes of entities in I, in the obvious way. The cliques in the unrolled network are determined by the clique templates C.” Section 4, “Another useful type of subgraph template involves transitivity patterns, where the presence of an A-B link and of a B-C link increases (or decreases) the likelihood of an A-C link... By introducing cliques over triples of relations, we can capture such patterns as well.” Section 3, “Taskar et al. therefore propose the use of belief propagation.” - EN: Taskar models the joint probability distribution of binary link variables over adjacent connections via an unrolled Markov network and computes each link variable's expectation using belief propagation, which corresponds to the claim’s computing expectations for random variables on the line graph nodes that represent those links.)
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to combine the connection state random variables of Gan in view of Zhao and Cai with the joint probabilistic link model of Taskar, in which a Markov network over the candidate links computes their marginal probabilities collectively. The motivation for doing so would be to predict correlated candidate edges collectively rather than scoring each node pair in isolation, which improves prediction accuracy. As Taskar elaborates regarding the benefit of this collective methodology in the Abstract, “We show that the collective classification approach of RMNs, and the introduction of subgraph patterns over link labels, provide significant improvements in accuracy over flat classification, which attempts to predict each link in isolation.”
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Gan in view of Zhao as applied to claim 2 above, and further in view of Wang et al. (“Noisy Positive-Unlabeled Learning with Self-Training for Speculative Knowledge Graph Reasoning,”), hereinafter “Wang”.
Claim 7
Regarding claim 7, Gan in view of Zhao teaches the method of claim 2 as shown above, including the link prediction model taught by Gan, which is trained by using the expected edge-incomplete graph taught by Zhao. Wang further teaches:
The method of claim 2, wherein the link prediction model is a model that is trained according to (i) a dual loss function to which one or more randomly sampled edges are applied in order to strike a balance between a number of connected edges and a number of unconnected edges in the structure of the edge-incomplete graph by taking into consideration one or more added edges in the expected edge-incomplete graph (Section 2.2, p. 3, "For each labeled triple s^l_i = (e_h, r, e_t), we construct K unlabeled triples s^u_ik by replacing the head and tail respectively with other entities: s^u_ik = (e_h, r, e^−_k) or (e^−_k, r, e_t), where e^−_k is the selected entity that ensures s^u_ik ∉ S^L. Initially, the construction can be randomized." Section 3.2, Equation 6, "We denote the loss function measuring the probability as L_triple: L_triple = −1/(K|S^L|) Σ_i Σ_k (w^l_i log[φ^l_1(s^l_i)] + (1−w^l_i) log[φ^l_0(s^l_i)] + w^u_ik log[φ^⋆_1(s^l_i, s^u_ik)] + (1−w^u_ik) log[φ^⋆_0(s^l_i, s^u_ik)])." Section 3.4, p. 6, "the latest estimation W̃^U for uncollected facts enables us to continuously sample unlabeled triplets with high label posterior to cover positive samples in the unlabeled set to the greatest extent." - EN: Wang's L_triple is a loss function that jointly processes collected triples (connected edges) through the w^l_i-weighted terms and uncollected triples (unconnected edges) through the w^u_ik-weighted terms, thereby striking a balance between the number of connected edges and the number of unconnected edges. The uncollected triples are constructed by randomly replacing entities, which reads on "one or more randomly sampled edges are applied." Wang's self-training strategy selects uncollected triples with high label posterior, i.e., triples likely to be true but missing from the observed graph, and adds them to the training set, which reads on "taking into consideration one or more added edges in the expected edge-incomplete graph." In the combination, the L_triple of Wang is applied to train the link prediction model of Gan as modified by Zhao's expected edge-incomplete graph.)
and (ii) a correction loss function which prevents excessive self-reinforcement based on the randomly sampled edges by taking into consideration one or more added edges in the expected edge-incomplete graph. (Section 3.2, Equation 9, "min_Θ L = min_Θ L_triple + L_KL + L_reg." Section 3.2, p. 5, "we set W^L and W^U as free parameters and utilize the term L_KL = KL(W^L ∥ W̃^L) + KL(W^U ∥ W̃^U) to regularize the difference." Section 3.2, p. 5, "where L_reg = ∥W^L∥_1 + ∥W^U∥_1 can be viewed as a normalization term. Considering the sparsity property of real-world graphs, L_reg penalizes the posterior estimation that there are too many true positive facts on KG." Section 3.4, p. 5, "the latest posterior estimation W̃^L for collected links updates neighbor sets to gradually prevent the encoder effects by false positive links." - EN: Wang's L_KL and L_reg together constitute a correction loss function. L_KL constrains the free label-posterior parameters W^L and W^U against the model-estimated posteriors W̃^L and W̃^U, preventing the model from arbitrarily assigning label posteriors that reinforce its own predictions without external constraint; this reads on "prevents excessive self-reinforcement." L_reg further corrects by penalizing the model when it estimates too many true positive facts, preventing a feedback loop in which the model adds more and more edges and then trains on them. Both L_KL and L_reg are computed over the randomly sampled uncollected triples (through W^U and W̃^U) and over collected triples (through W^L and W̃^L). The self-training strategy uses the latest posteriors to update which edges are considered added (high-posterior uncollected triples) and which collected edges may be false positives, which reads on "taking into consideration one or more added edges in the expected edge-incomplete graph." In the combination, the correction loss of Wang is applied alongside the dual loss to train the link prediction model of Gan in view of Zhao.)
Before the effective filing date of the claimed invention, it would have been obvious to one of ordinary skill in the art to combine the positive-unlabeled link prediction framework of Gan, as modified by the interpolated edge probabilities and sampled graph variants of Zhao, further with the dual loss and correction loss training methodology of Wang. The motivation for doing so would be to provide a principled training objective that accounts for the fact that some observed edges may be incorrect (false positives) and some unobserved edges may be truly connected (false negatives) in the edge-incomplete graph. Gan itself recognizes that unobserved links should be treated as unlabeled rather than negative, and Zhao's augmentation introduces probable missing edges; Wang's dual loss with balancing between connected and unconnected edges, together with the correction loss that prevents posterior estimates from self-reinforcing, would improve the accuracy of the combined model's training. As Wang explains regarding its ablation study in Section 4.4, "training the encoder without the proposed noisy Positive-Unlabeled framework will cause the performance drop, as this variant ignores the false negative/positive issues" (Table 3).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NAYMUR RAHMAN ALI whose telephone number is (571)272-0007. The examiner can normally be reached Mon-Fri. 9:30-6:30 pm.
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, Alexey Shmatov can be reached at (571)270-3428. 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.
/NAYMUR RAHMAN ALI/Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights