DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
This action is responsive to the claims dated 3/19/2024.
Claims 1-20 are presented for examination.
Specification
The disclosure is objected to because of the following informalities: (a) paragraph [0024] refers to element 190 as the "preventative module 190," whereas the same element is identified as the "action module 190" throughout the remainder of the disclosure (see, e.g., paragraphs [0019], [0023], and [0035]); (b) paragraph [0040] recites "evaluation module 230," but element 230 is the preprocessing module and the evaluation module is element 320 (see paragraphs [0033] through [0038]); (c) paragraph [0045] contains the misspelling "tarin," which should be "train"; (d) paragraph [0039] recites "copy past data," which should read "copy and paste data" (the same omission also appears in the paragraph numbered [0036]); (e) the paragraph beginning "When the term or is used" recites "On the hand," which should read "On the other hand,"; (f) the paragraph describing transaction features recites "and may other pieces of data," which should read "and many other pieces of data"; and (g) Figure 6, element 630, recites "the previous set of synthetic communications," whereas the corresponding text of the specification and claim 1 recite "the training set of synthetic communications," rendering the figure inconsistent with the description. Appropriate correction is required.
Claim Rejections - 35 U.S.C. 112(a)
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 9, 15, and 19 are rejected under 35 U.S.C. 112(a) as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, at the time the application was filed, had possession of the claimed invention.
Each of claims 9, 15, and 19 recites automatically altering the machine learning model or machine learning classifier based on comparing output of the LLM with known labels for the synthetic communications and the existing communications. The specification does not reasonably convey possession of altering the machine learning model or classifier on the basis of a comparison of the output of the large language model against known labels.
The specification describes the large language model (model 325) as receiving the classifications output by the machine learning classifier and, in turn, outputting a summary or one or more new proposed rules, identifying dominant conditions, or identifying leakage patterns of the classifier (specification [0029]-[0033]). Separately, the specification describes the decision module (module 140), and not the large language model, comparing the classifications output by the machine learning classifier with known labels in order to determine whether to retrain the classifier (specification paragraph [0013]). The specification nowhere describes comparing the output of the large language model itself with known labels, nor altering the machine learning model or classifier based on such a comparison. The disclosure therefore does not reasonably convey to a person of ordinary skill in the art that the inventor possessed the specific claimed step of automatically altering the machine learning model or classifier based on comparing the output of the LLM with known labels.
Although claims 9, 15, and 19 are original claims, the presumption that original claims provide their own written description is rebutted here because the recited comparison of the output of the large language model to known labels finds no counterpart in the specification as a whole. See In re Koller, 613 F.2d 819 (CCPA 1980); Ariad Pharmaceuticals, Inc. v. Eli Lilly & Co., 598 F.3d 1336 (Fed. Cir. 2010) (en banc).
Applicant may overcome this rejection by amending claims 9, 15, and 19 to recite the altering relationship that is supported by the specification, namely altering the machine learning model or classifier based on the output of the large language model (for example, based on the new proposed rules or summary output by the large language model), or by pointing to specific portions of the specification as originally filed that reasonably convey possession of comparing the output of the large language model with known labels. No new matter may be added (35 U.S.C. 132).
For the purposes of examination the Examiner will interpret the said respective claim limitations of claims 9, 15, and 19 as based on the output of the LLM.
Claim Rejections - 35 U.S.C. 112(b)
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.
Claims 5 and 8 are 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 specific bases for the rejection are set forth below.
Regarding claim 5, the claim recites the limitation "categorical features that include both categorical variables and numerical variables of the set of existing communications." This limitation renders the claim indefinite because it is unclear how a feature characterized as "categorical" can simultaneously include a numerical variable. A categorical variable and a numerical variable are mutually distinct types of data, and a categorical feature that nonetheless includes numerical variables is internally contradictory. The specification describes transforming numerical variables into categorical data using discretizing or binning techniques so that the generative model may be conditioned on numerical features as categorical features, but the claim as drafted recites a categorical feature that includes the numerical variables themselves rather than categorical features derived by transforming numerical variables. As a result, a person of ordinary skill in the art would not be reasonably apprised, with reasonable certainty, of the scope of the claimed feature. See MPEP 2173.05; Nautilus, Inc. v. Biosig Instruments, Inc., 572 U.S. 898, 901, 910 (2014). For purposes of examination, the limitation is interpreted, consistent with the specification, to mean that the set of conditions includes categorical features that include both categorical features generated from categorical variables and categorical features generated by transforming numerical variables of the set of existing communications.
Regarding claim 8, the claim recites the limitation "the set of synthetic communications" (in the recitation that the output of the machine learning model includes classifications for one or more of the set of synthetic communications and the set of existing communications). There is insufficient antecedent basis for this limitation in the claim. Parent claim 1 introduces two distinct sets of synthetic communications, namely "a current set of synthetic communications" and "a training set of synthetic communications." Because two distinct sets were previously recited and no single "set of synthetic communications" was introduced, it is unclear which set is being referenced, and a person of ordinary skill in the art cannot determine, with reasonable certainty, the scope of the limitation. See MPEP 2173.05(e); Nautilus, Inc. v. Biosig Instruments, Inc., 572 U.S. 898, 901 (2014). For purposes of examination, and in light of the specification, the limitation is interpreted to refer to the current set of synthetic communications input to the machine learning model during training.
Regarding claim 9, it does not cure the deficiencies of claim 8 and thus claim 9 is rejected under 35 U.S.C. 112(b) for at least being dependent on a rejected parent claim.
Claim Rejections - 35 U.S.C. 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-7, 10-14, 16-18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Dong, US 2020/0210808 A1, in view of Xu et al. (hereinafter Xu) “Modeling Tabular Data using Conditional GAN” (2019), and further in view of KATOH et al. (hereinafter KATOH), US 2020/0242412 A1.
Xu was disclosed in an IDS dated 3/19/2024.
Regarding independent claim 1, Dong teaches a method, comprising (Dong: [0024], a method 300 for performing a training process on a neural network system, "Referring to FIG. 3, illustrated is a method 300 for performing a training process (e.g., at block 102 of FIG. 1 using a first training dataset) on a neural network system"):
generating, by a computer system using a trained generative model, a current set of synthetic communications (Dong: [0022], [0026], a fraud transaction GAN 216 having a generator (a trained generative model) executed by a neural network system 200 (a computer system) generates a second training dataset of augmented transactions (a current set of synthetic communications), "a dataset augmentation process is performed, using latent variables in the latent space of the first trained semi-supervised adversarial autoencoder model, to generate a second training dataset"), wherein the generating includes inputting into the trained generative model (Dong: [0026], a dataset augmentation process is performed to generate a second training dataset (wherein the generating includes) using latent variables in the latent space of the generative process where these latent variables act as the input conditions fed into the trained generative model of the autoencoder/decoder (inputting into the trained generative model, "a dataset augmentation process is performed, using latent variables in the latent space... to generate a second training dataset"), and wherein the trained generative model is generated by a discriminator of the generative model (Dong: [0022], a generator and a discriminator (a discriminator of the generative model) play a min-max adversarial game in which the discriminator distinguishes real from generated data and the generator is trained to increase the discriminator error rate (and wherein the trained generative model is generated by), "a GAN is a framework that establishes a min-max adversarial game between two neural networks, a generator (also referred to as a generative model) and a discriminator"):
generating, by a generator of the generative model based on a set of existing communications, a training set of synthetic communications (Dong: [0024], [0026], the generator generates the second training dataset (a training set of synthetic communications) from a first training dataset of transactions (a set of existing communications), "The method 300 starts at block 302, where a training dataset including a plurality of batches of training data is received"); and
determining, by the discriminator of the generative model, one or more differences between the set of existing communications and the training set of synthetic communications (Dong: [0022], the discriminator discriminates between instances from the true data distribution and candidates produced by the generator (determines one or more differences), "The discriminator discriminates between instances from true data distribution and candidates produced by the generator");
updating the generator based on the one or more differences (Dong: [0022], the generator is trained to fool the discriminator based on the discriminator error rate (updating the generator based on the one or more differences), "The generator's training objective is to increase the error rate of the discriminator network (i.e., 'fool' the discriminator by producing novel synthesized instances"); and
training, by the computer system using the current set of synthetic communications and the set of existing communications, a machine learning model to evaluate newly initiated electronic communications (Dong: [0026] and Figure 1 operation 106, the neural network system trains a fraud classifier (a machine learning model) using the second training dataset of augmented transactions together with the transaction data (training by the computer system using the current set of synthetic communications and the set of existing communications), "Perform a second training process, using the second training dataset, on a fraud classifier to generate a trained fraud classifier", and the trained classifier evaluates an input transaction (to evaluate newly initiated electronic communication) at [0021], "provide an output including a fraud prediction of that transaction").
Dong does not expressly teach inputting a set of conditions for the synthetic communications into the trained generative model.
However, Xu teaches inputting a set of conditions for the synthetic communications into the trained generative model (Xu: Section 4.3, Conditional vector, a conditional vector cond is input to the conditional generator (into the trained generative model) to indicate the condition (inputting a set of conditions for the synthetic communications), "We introduce the vector cond as the way for indicating the condition").
Because Dong and Xu are analogous art and within the same field of endeavor, specifically the use of generative adversarial networks to synthesize structured data records, they address the same problem-solving area of generating realistic synthetic data to improve a downstream classifier, accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Xu's conditional vector and critic-estimated distance convergence with Dong's generator-and-discriminator synthesis framework, with a reasonable expectation of success, such that the generator is driven by an input set of conditions, to teach inputting a set of conditions for the synthetic communications into the trained generative model. This modification would have been motivated by the desire to control which records are synthesized and to objectively determine when the generator has learned the real data distribution, as Xu teaches that the conditional vector lets the model evenly explore all values of a column and that the critic-estimated distance measures how closely the synthetic conditional distribution matches the real one (Xu: Section 4.3).
Dong and Xu do not expressly teach generated by iteratively performing until a discriminator of the generative model determines that synthetic communications generated by the generative model satisfy a difference threshold.
However, KATOH teaches generated by iteratively performing until a discriminator of the generative model determines that synthetic communications generated by the generative model satisfy a difference threshold (Katoh: [0005] “the generator is trained to generate fake data that is similar to real data from input data, such as noise, and the discriminator is trained to discriminate whether the data generated by the generator is real data or fake data”, [0050] “the generator training unit 31 causes the generator to generate image data… Then, the generator training unit 31 inputs the generated image data… to the discriminator and acquires a discrimination result”, [0098] “a timing to terminate the training process can be set to an arbitrarily time point, such as… a time point at which the loss of the discriminator decreases to below a threshold”).
Because Dong, in view of Xu, and KATOH are analogous art directed to the adversarial training of a generator-and-discriminator pair, it would have been further obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to terminate the iterative adversarial training of the Dong and Xu generative model when the discriminator-determined difference between the synthetic and existing communications satisfies a threshold, as taught by KATOH, with a reasonable expectation of success, by applying KATOH’s discriminator-loss-threshold termination criterion to the Dong and Xu adversarial training loop, in which Xu’s critic already estimates the distance (difference) between the synthetic and real conditional distributions. This is no more than the use of a known technique to improve a similar method in the same way, yielding the predictable result of an objective stopping criterion for the training. See KSR; MPEP 2143(D). A person of ordinary skill would have been motivated to make this combination in order to objectively determine when the generator has adequately learned the real data distribution and thereby avoid both undertraining (yielding low-fidelity synthetic communications) and unnecessary overtraining (wasting computational resources). Furthermore, the number of training iterations and the difference threshold at which training is terminated are result-effective variables, and selecting a difference threshold at which the discriminator determines that the synthetic communications are sufficiently similar to the existing communications would have been a matter of routine optimization within the level of ordinary skill. See MPEP 2144.05(II)(B).
Regarding dependent claim 2, Dong, in view of Xu and KATOH, teach the method of claim 1, further comprising: identifying, by the computer system using the trained machine learning model, one or more newly initiated communications as atypical (Dong: [0021], the trained fraud classifier outputs a fraud prediction classifying a transaction (a newly initiated communication) as belonging to a fraudulent transaction class (atypical), "The output may indicate the probability (e.g., between 0 and 1) of a particular transaction belonging to a certain class (e.g., a fraudulent transaction class or a legitimate transaction class)"); and performing, by the computer system based on identifying one or more newly initiated communications as atypical, one or more actions corresponding to the atypical newly initiated communications (Dong: [0015], Figure 9 element 904, Dong's computing device 900 includes a fraud detection engine 904 whose stated purpose is to detect fraudulent transactions because "Fraudulent transactions are major problems with internet service providers").
Regarding dependent claim 3, Dong, in view of Xu and KATOH, teach the method of claim 2, wherein the one or more actions include one or more of the following types of actions: rejecting the one or more newly initiated communications, escalating authentication for the one or more newly initiated communications, and transmitting the one or more newly initiated communications for additional review (Dong: [0015], Figure 9 element 904; this limitation is a Markush group of alternative actions for which only one member need be taught. Dong's fraud detection engine 904 detects transactions that the trained classifier predicts to belong to the fraudulent transaction class, and because "Fraudulent transactions are major problems with internet service providers" the recognized object of such detection is to prevent the fraudulent transaction from completing; it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to reject a transaction predicted to be fraudulent (rejecting the one or more newly initiated communications), which satisfies the recited group).
Regarding dependent claim 4, Dong, in view of Xu and KATOH, teach the method of claim 1, wherein synthetic communications generated by the trained generative model simulate new types of atypical communications that are different than atypical communications included in the set of existing communications (Xu: Section 4.3, the conditional generator, driven by the conditional vector, evenly explores all possible values in the discrete columns and thereby generates rows for value combinations underrepresented in the real data (new types of atypical communications that are different than atypical communications included in the set of existing communications), "Properly sample the cond vector and training data can help the model evenly explore all possible values in discrete columns"; Dong: [0026], the augmentation process generates new synthetic transactions beyond the original training data).
Regarding dependent claim 5, Dong, in view of Xu and KATOH, teach the method of claim 1, wherein the set of conditions for the synthetic communications includes conditions based on which the synthetic communications are generated (Xu: Section 4.3, the conditional vector cond conditions the generation of each synthetic row (conditions based on which the synthetic communications are generated), "the conditional generator can generate synthetic rows conditioned on one of the discrete columns"), and wherein the set of conditions includes categorical features that include both categorical features generated from categorical variables (Xu: Section 4, the discrete columns are represented as one-hot categorical features (categorical features generated from categorical variables), "Discrete values can naturally be represented as one-hot vectors") and categorical features generated by transforming numerical variables of the set of existing communications (Xu: Section 4.2, continuous (numerical) columns are processed by mode-specific normalization that represents each value by a mode indicator and a scalar, transforming the numerical variables into a categorical mode representation (categorical features generated by transforming numerical variables), "representing continuous values with arbitrary distribution" using mode-specific normalization).
Regarding dependent claim 6, Dong, in view of Xu and KATOH, teach the method of claim 1, wherein the trained generative model is further generated by: identifying, by the discriminator of the generative model based on the one or more differences, whether ones of the set of existing communications and the set of synthetic communications are synthetic (Dong: [0021], the discriminator's discrimination function distinguishes instances from the true data distribution from generator candidates, that is, identifies whether each instance is real or synthetic (identifying whether ones of the set of existing communications and the set of synthetic communications are synthetic), "The discriminator discriminates between instances from true data distribution and candidates produced by the generator").
Regarding dependent claim 7, Dong, in view of Xu and KATOH, teach the method of claim 1, wherein the generative model is a conditional tabular generative adversarial network (CTGAN) (Xu: Abstract and Section 1, the generative model is a conditional tabular GAN (a conditional tabular generative adversarial network (CTGAN)), "We design CTGAN, which uses a conditional generator").
Regarding independent claim 10, Dong teaches a non-transitory computer-readable medium having instructions stored thereon that are executable by a computing device to perform operations (Dong: claim 15, a non-transitory machine-readable medium having stored thereon machine-readable instructions executable by one or more processors of a computing device, "a non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations"). The substantive operations of claim 10 recite the counterparts of the limitations of claim 1 (generating a current set of synthetic electronic communications with a trained generative model from a set of conditions, the model trained by iterating a generator and a discriminator until a difference threshold is satisfied, and training a machine learning classifier on the synthetic and existing electronic communications to classify newly initiated electronic communications) and are taught by Dong in view of Xu, and further in view of KATOH, for the same reasons set forth above for claim 1, with the express termination of the iterative adversarial training upon the discriminator-determined difference satisfying a threshold supplied by KATOH ([0098]). The motivation to combine Dong, Xu, and KATOH is the same as set forth above for claim 1.
Regarding dependent claim 11, the rejection of claim 2 is applied in the same manner to the corresponding non-transitory computer-readable-medium-form recitation of claim 11 (classifying one or more newly initiated electronic communications as atypical and performing one or more preventative actions).
Regarding dependent claim 12, the rejection of claim 3 is applied in the same manner to the corresponding non-transitory computer-readable-medium-form recitation of claim 12.
Regarding dependent claim 13, Dong, in view of Xu and KATOH, teach the non-transitory computer-readable medium of claim 10, wherein the operations further comprise: generating one or more conditions in the set of conditions for the synthetic electronic communications, based on which the synthetic electronic communications are generated (Xu: Section 4.3, constructing the conditional vector cond that conditions generation (generating one or more conditions in the set of conditions), "Calculate the vector cond"), wherein generating the one or more conditions includes: generating a first set of categorical features from categorical variables of the existing electronic communications (Xu: Section 4, representing the discrete columns of the real data as one-hot categorical features (a first set of categorical features from categorical variables), "Discrete values can naturally be represented as one-hot vectors"); and transforming, using one or more feature transformation techniques, numerical features of the existing electronic communications to generate a second set of categorical features (Xu: Section 4.2, applying mode-specific normalization (a feature transformation technique) to the continuous (numerical) columns to produce, for each value, a one-hot mode indicator and a scalar, that is, a mode-indicator categorical representation (a second set of categorical features), "we design a mode-specific normalization to deal with columns with complicated distributions").
Regarding dependent claim 14, Dong, in view of Xu and KATOH, teach the non-transitory computer-readable medium of claim 10, wherein the set of conditions includes at least three different variables, the combination of which does not appear in existing electronic communications (Xu: Section 4.3, the conditional vector indicates a selected value across the discrete columns, and the training-by-sampling procedure lets the model evenly explore value combinations that are underrepresented or absent in the real data (a combination of variables that does not appear in existing electronic communications), "Properly sample the cond vector and training data can help the model evenly explore all possible values in discrete columns"; the recitation of at least three different variables is an arrangement of the same conditioning mechanism over plural columns, and the number of conditioned variables is a result-effective design choice that would have been obvious to a person of ordinary skill in the art).
Regarding independent claim 16, Dong teaches a method, comprising: training, by a computer system based on existing communications, a generative model, wherein the training includes iteratively performing, until a discriminator of the generative model determines that synthetic communications generated by the generative model during training satisfy a difference threshold (Dong: [0022], [0024], a neural network system (a computer system) trains a generator and a discriminator of a GAN (a generative model) on a first training dataset of transactions (existing communications), "a GAN is a framework that establishes a min-max adversarial game between two neural networks, a generator (also referred to as a generative model) and a discriminator"; the express termination of the iterative adversarial training upon the discriminator-determined difference satisfying a threshold is supplied by Katoh ([0098]) as set forth above for claim 1): generating, by a generator of the generative model based on the existing communications, a training set of synthetic communications (Dong: [0026], the generator generates the second training dataset (a training set of synthetic communications) from the transaction data); and determining, by the discriminator of the generative model, one or more differences between the existing communications and the training set of synthetic communications (Dong: [0022], "The discriminator discriminates between instances from true data distribution and candidates produced by the generator"); updating the generator based on the one or more differences (Dong: [0022], "The generator's training objective is to increase the error rate of the discriminator network"); and generating, by the computer system using the trained generative model, a current set of synthetic communications, wherein the generating includes inputting a set of conditions for the synthetic communications into the trained generative model (Dong: [0026] generating the augmented dataset (a current set of synthetic communications); Xu: Section 4.3, inputting the conditional vector cond (a set of conditions) to the conditional generator, "We introduce the vector cond as the way for indicating the condition"); training, by the computer system using the current set of synthetic communications generated by the trained generative model and the existing communications, a machine learning classifier to classify newly initiated communications (Dong: [0026] and Figure 1 operation 106, "Perform a second training process, using the second training dataset, on a fraud classifier to generate a trained fraud classifier"); and classifying, by the computer system using the trained machine learning classifier, one or more newly initiated electronic communications (Dong: [0021], the trained classifier classifies an input transaction (a newly initiated electronic communication), "provide an output including a fraud prediction of that transaction"). The differences between claim 16 and Dong, namely the input set of conditions (supplied by Xu) and the express termination of the iterative adversarial training upon the discriminator-determined difference satisfying a threshold (supplied by Katoh), are obvious to combine with Dong for the same reasons and with the same motivation set forth above for claim 1.
Regarding dependent claim 17, the rejection of claim 2 is applied in the same manner to the corresponding recitation of claim 17 (performing, based on classifying one or more newly initiated communications as atypical, one or more actions corresponding to the atypical newly initiated communications).
Regarding dependent claim 18, Dong, in view of Xu and KATOH, teach the method of claim 17, wherein the synthetic communications generated by the generative model simulate new types of atypical communications that are different than atypical communications included in the existing communications (Xu: Section 4.3, the conditional generator evenly explores value combinations underrepresented in the real data (new types of atypical communications different than those in the existing communications), "help the model evenly explore all possible values in discrete columns"), and wherein the one or more actions include one or more of the following types of actions: rejecting the one or more newly initiated communications, escalating authentication for the one or more newly initiated communications, and transmitting the one or more newly initiated communications for additional review (Dong: [0015], Figure 9 element 904; this is a Markush group of alternative actions for which only one member need be taught, and because Dong's fraud detection engine 904 detects transactions predicted to belong to the fraudulent transaction class and "Fraudulent transactions are major problems with internet service providers," it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention, as a matter of common sense, to reject a transaction predicted to be fraudulent (rejecting the one or more newly initiated communications), see MPEP 2143.01).
Regarding dependent claim 20, Dong, in view of Xu and KATOH, teach the method of claim 16, wherein the generative model is a conditional tabular generative adversarial network (CTGAN) (Xu: Abstract and Section 1, "We design CTGAN, which uses a conditional generator").
Allowable Subject Matter
Claims 8, 9, 15, and 19 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims and overcoming the 35 U.S.C. 112(a) and 112(b) rejections.
The following is a statement of reasons for the indication of allowable subject matter:
The closest prior arts found when taken individually or in combination do not expressly teach or render obvious the limitations recited in dependent claims 8, 9, 15 and 19 when taken in the context of the claims as a whole.
At best the closest prior arts uncovered, specifically, Dong disclose: Systems and methods for data augmentation in a neural network system includes performing a first training process, using a first training dataset on a neural network system including an autoencoder including an encoder and a decoder to generate a trained autoencoder. A trained encoder is configured to receive a first plurality of input data in an N-dimensional data space and generate a first plurality of latent variables in an M-dimensional latent space, wherein M is an integer less than N. A sampling process is performed on the first plurality of latent variables to generate a first plurality of latent variable samples. A trained decoder is used to generate a second training dataset using the first plurality of latent variable samples. The second training dataset is used to train a first classifier including a first classifier neural network model to generate a trained classifier for providing transaction classification (Abstract); Xu disclose: Modeling the probability distribution of rows in tabular data and generating realistic synthetic data is a non-trivial task. Tabular data usually contains a mix of discrete and continuous columns. Continuous columns may have multiple modes whereas discrete columns are sometimes imbalanced making the modeling difficult. Existing statistical and deep neural network models fail to properly model this type of data. We design CTGAN, which uses a conditional generator to address these challenges. To aid in a fair and thorough comparison, we design a benchmark with 7 simulated and 8 real datasets and several Bayesian network baselines. CTGAN outperforms Bayesian methods on most of the real datasets whereas other deep learning methods could not (Abstract); and KATOH disclose: An anomaly detection apparatus performs training for the generator and the discriminator such that the generator maximizes a discrimination error of the discriminator and the discriminator minimizes the discrimination error The anomaly detection apparatus stores, while the training is being performed, a state of the generator that is half-trained and satisfies a pre-set condition, and retrains the discriminator by using an image generated by the half-trained generator that has the stored state (Abstract).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KUANG FU CHEN whose telephone number is (571)272-1393. The examiner can normally be reached M-F 9:00-5:30pm ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jennifer Welch can be reached on (571) 272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KC CHEN/Primary Patent Examiner, Art Unit 2143
Prosecution Insights