Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 08-DECEMBER-2025 has been entered.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
The amendment filed on 08-DECEMBER-2025 in response to the RCE mailed 14-AUGUST-2025 has been entered. Claims 1-7, 9-20 remain pending in the application.
With regards to the 101 rejection, the applicant’s amendments to claim 1 have been found sufficient to overcome the 101 rejection set forth in the final office action.
With regards to the 103 rejections, the applicant’s amendments to the claims have not overcome the rejections to claims 1-20. The newly added amendments have been further rejected in light of the newly added prior art ABEYKOON.
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.
Claim(s) 1-2, 11-12, and 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over HUSAIN et al. (U.S. Pub. No. US 20190122096 A1) in view of SHEN et al. (U.S. Pub. No. US 2020/0193217 A1) in view of NOURIAN et al. (U.S. Pub. No. US 11568286 B2) in view of ABEYKOON et al. (U.S. Pub. No. US 20220180291 A1)
Regarding claim 1, HUSAIN teaches the invention substantially as claimed, including:
A system comprising: a computing device including one or more processors configured to: obtain, over a network, first training data associated with a first machine learning model related to a first system, the first training data including one or more first features; receive, over the network, second training data associated with a second machine learning model related to a second system different from the first system, the second training data including one or more second features; ([0049] The method 300 also include, at 304, generating an output indicative of the classification result. For example, the processors 206 may send a signal to the display device 261 indicating the classification result 260. Additionally or in the alternative, the processors 206 may send a signal to the classifier generation and training instructions 262 to cause the neural network to be further trained. Additionally or in the alternative, the classification result 260 may be used to generate a training data entry in the training data 270. The training data entry may be used to update or further train the trained classifier. Claim 5: The computer system of claim 1, wherein the updated trained classifier is further based on training data entries associated with one or more other neural networks include first training data associated with a first neural network that is configured to analyze a first type of data and second training data associated with a second neural network that is configured to analyze a second type of data, wherein the first type of data is different from the second type of data. (it can be seen that the training data entries are made by the neural network itself, thus the training data is transmitted to and from the network, making it obtained and received by a network.));
While HUSAIN does teach obtaining training data associated with different systems, it does not explicitly teach:
iteratively apply the one or more first features of the first training data and the one or more second features of the second training data set to an overlap analysis, and in response generate a third training data having a plurality of data samples by extracting an overlap portion of the second training data based on an overlap between the one or more first features and the one or more second features;
However, in analogous art that similarly generates training data, SHEN teaches:
iteratively apply the one or more first features of the first training data and the one or more second features of the second training data set to an overlap analysis, and in response generate a third training data having a plurality of data samples by extracting an overlap portion of the second training data based on an overlap between the one or more first features and the one or more second features; ([0010] step S2, respectively extracting overlapped features in each sentence sample to form an overlapped feature matrix (i.e., one or more “third features”) and combining the corresponding character/word vector matrix with the overlapped feature matrix (i.e. third training data) for each sentence sample to serve as input data of the first neural network model; (inputting training data to reach a result is training. Training a model is fundamentally iterative, thus, the overlap of the training data for the current step of training would also be iterative.));
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with SHEN’s teaching of identifying third features and training using third training data and, with HUSAIN’s, teaching of a method to obtaining training data related to different systems, to realize, with a reasonable expectation of success, a method that trains a system associated with the first training data with training data based on an overlap, as in SHEN, where the training data is associated with different systems, as in HUSAIN. A person of ordinary skill would have been motivated to improve the measurement of the overlapping data(SHEN [0005])
While HUSAIN does teach obtaining training data associated with different systems and finding an overlap in training data, it does not explicitly teach:
[[and]] execute the first machine learning model to iteratively apply each of the data samples of the third training data over the network to generate a plurality of predictions, wherein each prediction of the plurality of predictions corresponds to the respective data sample of the third training data iteratively apply each prediction of the plurality of predictions to a cost-sensitive loss function to generate a weighted misclassification cost corresponding to each prediction of the plurality of predictions; and execute the first machine learning model to iteratively apply each of the data samples of the third training data and the weighted misclassification cost corresponding to each prediction of the plurality of predictions over the network to train the first machine learning model.
However, in analogous art that similarly generates training data, NOURIAN teaches:
[[and]] execute the first machine learning model to iteratively apply each of the data samples of the third training data over the network to generate a plurality of predictions, wherein each prediction of the plurality of predictions corresponds to the respective data sample of the third training data ((Col 3, lines 58-67) In accordance with one or more embodiments, learning software 112 may process the training data x associated with certain features without taking the labels t into consideration (i.e., blindly) and may categorize the training data according to an initial set of weights (w) and biases (b). The generated output y may indicate that training data x is classified as belonging to a certain class by learning software 112. In one aspect, the result y may be checked against the associated label (i.e., tag t) to determine how accurately learning software 112 is classifying the training data. (here we see that the prediction correlates with the specific data ‘x’, which in combination with HUSAIN can be referred to as a ‘third training data’. Again, training a model is inherently iterative.))
iteratively apply each prediction of the plurality of predictions to a cost-sensitive loss function to generate a weighted misclassification cost corresponding to each prediction of the plurality of predictions, ((Col 4, lines 1-13) In the initial stages of the learning phase, the categorization may be based on randomly assigned weights and biases (i.e. costs), and therefore highly inaccurate. However, learning software 112 may be trained based on certain incentives or disincentives (e.g., a calculated loss function) to adjust the manner in which the provided input is classified. The adjustment may be implemented by way of updating weights and biases over and over again. Through multiple iterations and adjustments, the internal state of learning software 112 may be continually updated to a point where a satisfactory predictive state is reached (i.e., until learning software 112 starts to more accurately classify the training data). (the loss function is influenced by weights and biases, which are known as ‘costs’. Thus, the function can be said to be sensitive to cost, i.e. a cost-sensitive loss function.))
and execute the first machine learning model to iteratively apply each of the data samples of the third training data and the weighted misclassification cost corresponding to each prediction of the plurality of predictions over the network to train the first machine learning model. ((Col 3, lines 58-67) In accordance with one or more embodiments, learning software 112 may process the training data x associated with certain features without taking the labels t into consideration (i.e., blindly) and may categorize the training data according to an initial set of weights (w) and biases (b). The generated output y may indicate that training data x is classified as belonging to a certain class by learning software 112. In one aspect, the result y may be checked against the associated label (i.e., tag t) to determine how accurately learning software 112 is classifying the training data.
(Col 4, lines 1-13) In the initial stages of the learning phase, the categorization may be based on randomly assigned weights and biases, and therefore highly inaccurate. However, learning software 112 may be trained based on certain incentives or disincentives (e.g., a calculated loss function) to adjust the manner in which the provided input is classified. The adjustment may be implemented by way of updating weights and biases over and over again. Through multiple iterations and adjustments, the internal state of learning software 112 may be continually updated to a point where a satisfactory predictive state is reached (i.e., until learning software 112 starts to more accurately classify the training data).)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with NOURIAN’s teaching of prediction using the third data with the weight classification and, with HUSAIN’s, as modified by SHEN, teaching of a method to obtaining training data related to different systems, to realize, with a reasonable expectation of success, a method that trains a system associated with the first training data with training data based on an overlap, as in NOURIAN, where the training data is associated with different systems, as in HUSAIN, as modified by SHEN. A person of ordinary skill would have been motivated to reduce errors (NOURIAN Col 1, lines 50-59)
While HUSAIN, as modified by SHEN and NOURIAN, does teach obtaining training data associated with different systems and finding an overlap in training data and finding a misclassification cost, it does not explicitly teach:
wherein the weighted misclassification cost is based on an operational cost, a customer friction cost, a chargeback cost, or a lost revenue cost;
However, in analogous art that similarly handles loss costs, ABEYKOON teaches:
wherein the weighted misclassification cost is based on an operational cost, a customer friction cost, a chargeback cost, or a lost revenue cost; ([0091] The cost/loss function elements may include various kinds of matrices that represent quality, operational cost, time, KPI or the like related to event response.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with ABEYKOON’s teaching of a loss function based on a retailer cost and, with HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a misclassification cost, to realize, with a reasonable expectation of success, a loss function based on a retailer cost, as in ABEYKOON, where the function finds the misclassification, as in HUSAIN, as modified by SHEN and NOURIAN. A person of ordinary skill would have been motivated to improve efficiency (ABEYKOON [0004])
Regarding claim 2, Shen further teaches:
The system of claim 1, wherein the computing device is further configured to: compare the one or more second features to the one or more first features to match at least a subset of the one or more second features to at least a subset of the one or more first features; and based on comparing the one or more second features to the one or more first features, identify the one or more third features, the one or more third features including the subset of the one or more second features that match the subset of the one or more first features. ([0070] Specifically, in the present embodiment, in the above steps, in order to facilitate the processing by a computer, the first character can be 1 and the second character can be 0, and a binary overlapped feature vector associated with each sentence sample can be formed. For example, for two sentence samples “ 我想听歌(I want to listen to songs)” and “ 給我放首歌(Play a song for me)”, the mutually overlapped parts (i.e., the overlapped features) are “我 (me)” and “歌 (song)” respectively, then the overlapped feature sequence for “我想听歌” is 1001, the overlapped feature sequence for “給我放首歌” is 01001, then the above two overlapped feature sequences 1001 and 01001 are respectively mapped to form an overlapped feature matrix)
Regarding claims 11 and 17, they comprise of limitations similar to claim 1, and are therefore rejected for similar rationale. Regarding claims 12 and 18, they comprise of limitations similar to claim 2, and are therefore rejected for similar rationale.
Claim(s) 3-6, 13-16, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over HUSAIN et al. (U.S. Pub. No. US 20190122096 A1) in view of SHEN et al. (U.S. Pub. No. US 2020/0193217 A1), NOURIAN et al. (U.S. Pub. No. US 11568286 B2), ABEYKOON et al. (U.S. Pub. No. US 20220180291 A1) in further view of LI (U.S. Pub. No. US 2020/0380524 A1).
Regarding claim 3 SHEN further teaches:
The system of claim 2, wherein the computing device is further configured to: train a first system, associated with the first training data, using the third training data ([0043] Specifically, the first neural network model and the second neural network model are integrally formed by unified training, that is, the sentence similarity determining model including the first neural network model and the second neural network model is established at first (the output of the first neural network model is used as the input of the second neural network model), and then the carrying out training by inputting training samples to the first neural network model to form the complete sentence similarity determining model.;)
wherein the one or more third features include the subset of the one or more second features ([0010] step S2, respectively extracting overlapped features in each sentence sample to form an overlapped feature matrix, and combining the corresponding character/word vector matrix with the overlapped feature matrix for each sentence sample to serve as input data of the first neural network model)
While HUSAIN, as modified by SHEN and NOURIAN, does teach training a first system, that is related to the first training data, with the third training data, which has a subset of second features, it does not explicitly teach:
the subset of the one or more second features are related to fraud detection.
However, in analogous art that similarly uses training data and features to train systems, LI teaches features that are related to fraud detection.
the subset of the one or more second features are related to fraud detection. ([0009] An apparatus for generating a transaction feature is provided, where the transaction feature is used to identify an illegal transaction; [0023] It should be understood that although terms “first”, “second”, “third”, etc. may be used in the present specification to describe various types of information, the information should not be limited by these terms.; [0079] In the present implementation, a machine learning model under supervision can be used as the transaction feature generation model, such as a neural network model, which is not specially limited in the present specification.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined with LI’s teaching of features that are related to fraud detection and, with HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to train a system using the generated overlap training data and the first training data associated with the first system through training and generation by using SHEN’s training data to train HUSAIN’s neural network, to realize, with a reasonable expectation of success, a method that trains a system associated with the first training data with training data based on an overlap which would have a subset of the second features, as in HUSAIN, as modified by SHEN and NOURIAN, where the second features are related to fraud detection, as in LI. A person of ordinary skill would have been motivated to make this combination to aid in the detection of fraud (LI [0007])
Regarding claim 4, while HUSAIN, as modified by SHEN and NOURIAN, does teach claim 1, which claim 4 is dependent on, it does not teach:
The system of claim 1, wherein the second training data includes data samples based on historical transaction data associated with a plurality of customers, each data sample includes observed labels for at least one of the one or more second features, each observed label corresponding to a second feature of the one or more second features.
However, in analogous art that similarly uses a cost-sensitive loss function for identification, LI teaches:
The system of claim 1, wherein the second training data includes data samples based on historical transaction data associated with a plurality of customers ([0035] Step 104: Obtain some original features of the sample transaction data and determine one or more combination methods for the original features.[0036] In the present implementation, the original features are features of the sample transaction data, such as a transaction amount, the number of transactions, a distance between a transaction location and a merchant, a merchant category, a user category, etc.)
each data sample includes observed labels for at least one of the one or more second features, each observed label corresponding to a second feature of the one or more second features. ([0129] For example, the feature generation model can be trained by using sample dataset 1 in a first transaction scenario and sample dataset 2 in a second transaction scenario. Each piece of user data in sample dataset 1 and sample dataset 2 has a user label.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined LI’s teaching of data samples related to a customer’s transaction data and labels assigned to said data, and with HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to generate data using the overlap between two training data sets, to realize, with a reasonable expectation of success, a system that generates training data, as in HUSAIN, as modified by SHEN and NOURIAN, that includes data samples related to labeled transactional data related to customers, as in LI. A person of ordinary skill would have been motivated to make this combination to aid in the detection of fraud (LI [0007])
Regarding claim 5, LI further teaches:
The system of claim 4, wherein the third training data includes each data sample of the plurality of data samples with a subset of the observed labels corresponding to the one or more third features ([0007] A method for training a transaction feature generation model includes the following: obtaining a sample dataset, where the sample dataset includes some pieces of sample transaction data with a transaction label)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined LI’s teaching of a method which shows transaction data associated with a transaction label, and with HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to generate data using the overlap between two training data sets, to realize, with a reasonable expectation of success, a system that generates training data, as in HUSAIN, as modified by SHEN and NOURIAN, that includes data samples corresponding to the features of the training data, as in LI. A person of ordinary skill would have been motivated to make this combination to aid in the detection of fraud (LI [0007])
Regarding claim 6, LI further teaches:
The system of claim 4, wherein the third training data includes a subset of the plurality of data samples that include at least one label corresponding to one of the one or more third features. ([0072] In the present implementation, a smaller difference between the new feature obtained through the combination and the transaction label of each piece of sample transaction data can indicate a more reliable new feature obtained. Therefore, in the present example, the difference between the new feature and the transaction label is used as the feature label of the new feature.; [0023] It should be understood that although terms “first”, “second”, “third”, etc. may be used in the present specification to describe various types of information, the information should not be limited by these terms.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined LI’s teaching of a method which shows transaction data associated with a transaction label, and with HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to generate data using the overlap between two training data sets, to realize, with a reasonable expectation of success, a system that generates training data, as in HUSAIN, as modified by SHEN and NOURIAN, that includes data samples corresponding to the features of the training data, as in LI. A person of ordinary skill would have been motivated to make this combination to aid in the detection of fraud (LI [0007])
Regarding claim 13, it comprises of limitations similar to claim 3, and is therefore rejected for similar rationale. Regarding claims 14 and 19, the comprise of limitation similar to claim 4, and are therefore rejected for similar rationale. Regarding claim 15, they comprise of limitations similar to claim 5, and are therefore rejected for similar rationale. Regarding claim 16, it comprises of limitation similar to claim 6, and is therefore rejected for similar rationale.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over HUSAIN et al. (U.S. Pub. No. US 20190122096 A1), SHEN et al. (U.S. Pub. No. US 2020/0193217 A1), and NOURIAN et al. (U.S. Pub. No. US 11568286 B2), ABEYKOON et al. (U.S. Pub. No. US 20220180291 A1) in further view of BUCKINGHAM (U.S. Pub. No. US 11120450 B1).
Regarding claim 7, while HUSAIN, as modified by SHEN and NOURIAN, does teach claim 1, which claim 7 is dependent on, it does not teach:
The system of claim 1, wherein the one or more first features include a set of multi- dimensional customer-device pairs, each multi-dimensional customer-device pair associated with a set of transaction specific features that indicate riskiness of a corresponding historical transaction within the first training data.
However, in analogous art that similarly terms, BUCKINGHAM teaches:
wherein the one or more first features include a set of multi- dimensional customer-device pairs, each multi-dimensional customer-device pair associated with a set of transaction specific features ((4)… the first set of features further includes one or more of a geolocation feature, a user device registration feature, an authentication token feature, a payment token feature, a direct deposit feature, a check-free account feature, a P2P transfer feature, a geographic account limitation feature, a single-use account number feature, a no-push transfers feature, and a federated identity management feature; the second set of features includes one or more of a no-hold deposit feature, a no-limit withdrawal feature, a password-free account feature, a PIN-free account feature, and a federated identity management feature.)
that indicate riskiness of a corresponding historical transaction within the first training data. ((14) The user's selection of the feature(s) 110A may be communicated to one or more risk analysis computing devices 112. The risk analysis computing device(s) 112 may include any appropriate number and type of computing device. For example, the risk analysis computing device(s) 112 may include one or more server computers, distributed computing devices (e.g., cloud computing devices), and so forth. In some implementations, the risk analysis computing device(s) 112 may execute a risk engine 114. The risk engine 114 may analyze the feature set 110A and determine a risk metric 116 that indicates a level of fraud risk associated with the feature set 110A. In some instances, the risk metric 116 may be evaluated for the account(s) 120 that are associated with a particular user 102. Accordingly, the risk metric 116 may also be described as being associated with the user 102. In some implementations, a separate risk metric 116 may be determined for each of the user's accounts 120, in instances where a particular user 102 has multiple accounts 120 managed by the service module(s) 108. In some implementations, the risk engine 114 may determine an overall risk for the user 102 who has multiple financial and/or non-financial accounts 120. The risk metric 116 may be based on one or more of the feature(s) enabled for the account(s) 120. In some implementations, the risk metric 116 may be further based on the user's past behavior, the user's location, the user's transaction history, the user's credit history, and/or other characteristics of the user 102.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined BUCKINGHAM’s teaching of a set of customer-device pairs and features related to fraud riskiness, and with HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to generate data using the overlap between two training data sets, to realize, with a reasonable expectation of success, a system that generates training data, as in HUSAIN, as modified by SHEN and NOURIAN, that includes data samples corresponding risk of fraud corresponding to customer transaction history, as in BUCKINGHAM. A person of ordinary skill would have been motivated to make this combination to aid in the detection and prevention of fraud (BUCKINGHAM (1))
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over HUSAIN et al. (U.S. Pub. No. US 20190122096 A1), SHEN et al. (U.S. Pub. No. US 2020/0193217 A1) and NOURIAN et al. (U.S. Pub. No. US 11568286 B2), ABEYKOON et al. (U.S. Pub. No. US 20220180291 A1) in view of HUMPHREY (E.P. Pub. No. EP 3528462 A1).
Regarding claim 9, while HUSAIN, as modified by SHEN and NOURIAN, does teach claim 1, which claim 9 is dependent on, it does not teach:
the second training data is more densely populated than the first training data such that the second training data has more data samples than the first training data.
However, in analogous art that similarly uses a model to identify threats, HUMPHREY teaches training data and features associated to a first system and second system that are different from one another.
the second training data is more densely populated than the first training data such that the second training data has more data samples than the first training data. ([0097] Many of these time series data sets are extremely sparse, with most data points equal to 0. Examples would be employee's using swipe cards to access a building or part of a building, or user's logging into their workstation, authenticated by Microsoft Windows Active Directory Server, which is typically performed a small number of times per day. Other time series data sets are much more populated)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined HUMPHREY’s teaching of data sets that are less populated than others, and, with HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to generate data using the overlap between two training data sets, to realize, with a reasonable expectation of success, a system that generates training data from a first and second training data, as in HUSAIN, as modified by SHEN and NOURIAN, that has more data samples in the second as it does in the first, as in HUMPHREY. A person of ordinary skill would have been motivated to make this combination to aid in increasing system security (HUMPHREY [0005])
Claim 10 and 21 are rejected under 35 U.S.C. 103 as being unpatentable over HUSAIN et al. (U.S. Pub. No. US 20190122096 A1), SHEN et al. (U.S. Pub. No. US 2020/0193217 A1) and NOURIAN et al. (U.S. Pub. No. US 11568286 B2), ABEYKOON et al. (U.S. Pub. No. US 20220180291 A1) in view of CHUN (U.S. Pub. No. US 20220150583 A1).
Regarding claim 10, while HUSAIN, as modified by SHEN and NOURIAN, does teach claim 1, which claim 10 is dependent on, it does not teach:
the third training data further includes a cost-sensitive loss that varies based on a type of misclassification, such that the cost-sensitive loss is lower for a false positive misclassification than for a false negative misclassification detected when training a third system using the third training data.
However, in analogous art that similarly uses a cost-sensitive loss function for identification, CHUN teaches a cost-sensitive loss function that weighs false negatives heavier than a false negative.
the third training data further includes a cost-sensitive loss that varies based on a type of misclassification, such that the cost-sensitive loss is lower for a false positive misclassification than for a false negative misclassification detected when training a third system using the third training data. ([0181] For example, in case of proceeding with binary classification supposing that the minority class is one, the cost-sensitive learning may guide the model to learn the minority class by increasing loss weight regarding false negative during the learning of the model.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined CHUN’s teaching of a cost-sensitive loss function that weighs false-negatives highly, and HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to generate data using the overlap between two training data sets, to realize, with a reasonable expectation of success, a system that generates training data from a first and second training data, as in HUSAIN, as modified by SHEN and NOURIAN, which has a cost-sensitive loss function that weighs false-negatives higher than false-positives, as in CHUN. A person of ordinary skill would have been motivated to make this combination to aid identifying important features from the third training data for training the third system.(CHUN [0005])
ABEYKOON further teaches:
wherein the cost-sensitive loss for the false positive misclassification isbased on the customer friction cost, the operational cost, or the lost revenue cost, and the cost-sensitive loss for the false negative misclassification is based on a retailer cost. ([0091] The cost/loss function elements may include various kinds of matrices that represent quality, operational cost, time, KPI or the like related to event response.)
Regarding claim 21, it comprises of limitations similar to those of claim 10 and is therefore rejected for similar rationale.
Claim(s) 20 is rejected under 35 U.S.C. 103 as being unpatentable over HUSAIN et al. (U.S. Pub. No. US 20190122096 A1) in view of SHEN et al. (U.S. Pub. No. US 2020/0193217 A1), NOURIAN et al. (U.S. Pub. No. US 11568286 B2), ABEYKOON et al. (U.S. Pub. No. US 20220180291 A1), LI (U.S. Pub. No. US 2020/0380524 A1) in view of CHUN (U.S. Pub. No. US 20220150583 A1).
Regarding claim 20, while HUSAIN, as modified by SHEN and NOURIAN, does teach claim 11, which claim 20 is dependent on, it does not teach:
the third training data further includes a cost-sensitive loss that varies based on a type of misclassification, such that the cost-sensitive loss is lower for a false positive misclassification than for a false negative misclassification detected when training a third system using the third training data.
However, in analogous art that similarly uses a cost-sensitive loss function for identification, CHUN teaches a cost-sensitive loss function that weighs false negatives heavier than a false negative.
the third training data further includes a cost-sensitive loss that varies based on a type of misclassification, such that the cost-sensitive loss is lower for a false positive misclassification than for a false negative misclassification detected when training a third system using the third training data. ([0181] For example, in case of proceeding with binary classification supposing that the minority class is one, the cost-sensitive learning may guide the model to learn the minority class by increasing loss weight regarding false negative during the learning of the model.)
It would have been obvious to a person skilled in the art before the effective filing date of the invention to have combined CHUN’s teaching of a cost-sensitive loss function that weighs false-negatives highly, and HUSAIN’s, as modified by SHEN and NOURIAN, teaching of a method to generate data using the overlap between two training data sets, to realize, with a reasonable expectation of success, a system that generates training data from a first and second training data, as in HUSAIN, as modified by SHEN and NOURIAN, which has a cost-sensitive loss function that weighs false-negatives higher than false-positives, as in CHUN. A person of ordinary skill would have been motivated to make this combination to aid identifying important features from the third training data for training the third system.(CHUN [0005])
ABEYKOON further teaches:
wherein the cost-sensitive loss for the false positive misclassification isbased on the customer friction cost, the operational cost, or the lost revenue cost, and the cost-sensitive loss for the false negative misclassification is based on a retailer cost. ([0091] The cost/loss function elements may include various kinds of matrices that represent quality, operational cost, time, KPI or the like related to event response.)
Response to Arguments
Applicant’s arguments filed 08-DECEMBER-2025 have been fully considered, but they are found to be non-persuasive
With regards to the applicant’s remarks regarding the 103 rejections in the final action, the applicant argues that the prior art does not teach the newly amended claims 1, 10-11, 17, and 20-21. The examiner acknowledges this argument and has added newly found prior art ABEYKOON. Further, the examiner has adjusted all dependent claims accordingly.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SKIELER A KOWALIK whose telephone number is (571)272-1850. The examiner can normally be reached 8-5.
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, Mariela D Reyes can be reached at (571)270-1006. 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.
/SKIELER ALEXANDER KOWALIK/Examiner, Art Unit 2142 /Mariela Reyes/Supervisory Patent Examiner, Art Unit 2142
Prosecution Insights