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 office action is in response to submission of application on 11/20/2023.
Claims 1-20 are presented for examination.
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 1-20 are rejected under 35 U.S.C. § 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Is the claim directed to a process, machine, manufacture or composition of matter?
Claims 1-7 are directed do a method (i.e., a process); claims 8-14 are directed to a system (i.e., machine/apparatus); and claims 15-20 are directed to a non-transitory computer readable medium (i.e., a product); therefore, all pending claims are directed to one of the four statutory categories of invention.
Step 2A: Prong 1: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Claim 1 recites limitations of:
implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al mode – mental process (observation, evaluation, judgement, opinion) as a human mind can generate counterfactuals and corresponding target data samples.
computing an average counterfactual distance between closest counterfactuals and corresponding target data samples on the predefined time period – mathematical concepts (relationships, formulas or equations, calculations) – as computing an average distance is a calculation.
comparing the average counterfactual distance to a predefined threshold value – mental process (observation, evaluation, judgement, opinion) as a human mind can compare an average distance to a predefined threshold.
identifying that the Al model needs to be retrained when output data from comparing indicates that the average counterfactual distance is less than the predefined threshold value – mental process (observation, evaluation, judgement, opinion) as a human mind can determine if an AI model should be retrained based on a comparison of an output distance to a threshold.
which are mental processes and therefore abstract ideas.
Step 2A, Prong 2: Does the claim recite additional elements that integrate the judicial exception into a practical application?
The claim recites the additional elements of:
A method for identifying when to retrain an artificial intelligence (Al) model by utilizing one or more processors along with allocated memory - computer components recited at a high level are construed as generic computer components used to implement the abstract idea. See MPEP 2106.05(f)(2).
receiving an Al model and a set of target data samples on a predefined time period for identifying when to retrain the Al model – inputting data is insignificant, extra-solution activity. See MPEP 2106.05(g).
automatically retraining the Al model when it is determined that the average counterfactual distance is less than the predefined threshold value – training a model without a description of the model or the training is mere instructions to feed data to a black box. See MPEP 2106.05(f)(3).
The limitations do not integrate the abstract idea into a practical application.
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception?
The additional elements of:
A method for identifying when to retrain an artificial intelligence (Al) model by utilizing one or more processors along with allocated memory - computer components recited at a high level are construed as generic computer components used to implement the abstract idea. See MPEP 2106.05(f)(2).
receiving an Al model and a set of target data samples on a predefined time period for identifying when to retrain the Al model – inputting data is insignificant, extra-solution activity. See MPEP 2106.05(g). Transmitting data is well-understood, routine, and conventional. See MPEP 2106.05(d)(II)(i).
automatically retraining the Al model when it is determined that the average counterfactual distance is less than the predefined threshold value – training a model without a description of the model or the training is mere instructions to feed data to a black box. See MPEP 2106.05(f)(3).
The additional elements do not amount to significantly more than the abstract idea. Therefore, claim 1 is not patent eligible.
Independent claims 8 and 15 recite the same relevant limitations and a similar analysis applies. Claim 8 recites the additional elements of “A system for identifying when to retrain an artificial intelligence (Al) model, the system comprising: a processor; and a memory operatively connected to the processor via a communication interface, the memory storing computer readable instructions” – computer components recited at a high level are construed as generic computer components used to implement the abstract idea. See MPEP 2106.05(f)(2). Claim 15 recites the additional elements of “A non-transitory computer readable medium” – computer components recited at a high level are construed as generic computer components used to implement the abstract idea. See MPEP 2106.05(f)(2). The additional elements do not integrate the judicial exception into a practical application. Nor do they amount to significantly more. Therefore, the independent claims are not patent eligible.
The above analysis similarly applies to the dependent claims.
Claims 2, 9, and 16 recite the additional elements of “the plurality of counterfactuals are plausible samples on a desired side of a decision boundary generated by perturbing a query sample.” – description of the result of performing the mental process. See MPEP 2106.05(f)(3).
Claims 3, 10, and 17 recite the additional elements of “the predefined threshold value is use case or dataset specific and is used to identify whether retraining of the Al model is needed or not.” – description of a condition in performing the mental process, as such a mental process (observation, evaluation, judgment, opinion).
Claims 4, 11, and 18 recite the additional elements of “displaying how counterfactual distance evolves over time on a two-dimensional graph and wherein x-axis of the graph represents days and y-axis of the graph represents average counterfactual distance.” – description of the output is insignificant, extra-solution activity. See MPEP 2106.05(g). Transmitting data is well-understood, routine, and conventional. See MPEP 2106.05(d)(II)(i).
Claims 5 and 12 recite the additional elements of “analyze how the average counterfactual distance evolves over time by monitoring the graph.” – mental process (observation, evaluation, judgment, opinion) as a human mind can analyze how an average distance evolves over time by monitoring a graph.
Claims 6 and 13 recite the additional elements of “determine that the Al model is less confident in outputting a prediction data based on determining that the target data samples are getting close to the decision boundary over time.” – mental process (observation, evaluation, judgement, opinion) as a human mind can determine that data samples are close to a decision boundary.
Claim 19 is a combination of the relevant limitations of claims 5 and 6 and is rejected for the same reasons as claims 5 and 6.
Claims 7, 14, and 20 recited the additional elements of “identifying that the AI model does not need to be retrained when output data from comparing indicates that the average counterfactual distance is equal to or more than the predefined threshold value.” – mental process (observation, evaluation, judgement, opinion) as a human mind can determine that an AI model does not need to be retrained based on comparing outputs to a threshold value.
The dependent claim limitations do not integrate the judicial exception into a practical application. Nor do they amount to significantly more. Therefore, the claims are not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-3, 7, 8-10, 14, 15-17, and 20 are rejected under 35 U.S.C. § 103 as being unpatentable over Tirupathi, et al (US 2024/0281722 A1, Forecasting and Mitigating Concept Drift Using Natural Language Processing, herein Tirupathi), Carlevaro, et al (Counterfactual Building and Evaluation via eXplainable Support Vector Data Description, herein Carlevaro), and Bodria, et al (Transparent Latent Space Counterfactual Explanations for Tabular Data, herein Bodria).
Regarding claim 1,
Tirupathi teaches a method for identifying when to retrain an artificial intelligence (Al) model by utilizing one or more processors along with allocated memory (Tirupathi, Figure 2, and, paragraph [0003], line 1 “According to one embodiment, a method, computer system, and computer program product for automatically forecasting and mitigating concept drift in machine learning models using natural language processing is provided.”
PNG
media_image1.png
666
468
media_image1.png
Greyscale
In other words, method is method, computer system is one or more processors and memory, and retrain… the target model is retrain an artificial intelligence model.), the method comprising:
receiving an Al model and a set of target data samples on a predefined time period for identifying when to retrain the Al model (Tirupathi, Abstract, line 1 “An embodiment for automatically forecasting and mitigating concept drift in machine learning models using natural language processing. The embodiment may automatically detect features and variables considered by a target machine learning model. The embodiment may automatically store and update a configurable corpus of relevant documents pertaining to a domain of the target machine learning model. The embodiment may automatically extract event types and corresponding time stamps from the configurable corpus of relevant documents.” In other words, models and a target model is receive a target model, corpus of relevant documents is target data samples, corresponding timestamps is on a predefined time period, and from prior mapping in response to detection automatically retraining is identifying when to retrain the AI model.) ;
[implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al model];
[computing an average counterfactual distance between closest counterfactuals and corresponding target data samples on the predefined time period];
[comparing the average counterfactual distance to a predefined threshold value];
identifying that the Al model needs to be retrained when output data from comparing indicates that the [average counterfactual distance] is less than the predefined threshold value; and automatically retraining the Al model (Tirupathi, Figure 2, and, paragraph [0038], line 12 “In embodiments, concept drift detection program 150 may be configured to perform further suitable known model interpretability techniques to gather additional information about a target model's domain, weight or importance assigned to the detected features and variables using known model agnostic interpretability methods such as Local Interpretable Model Agnostic Explanations (LIME), Shapley Additive explanations (SHAP), Anchors, counterfactuals, etc.,”, and, paragraph [0041], line 25 “In embodiments, the threshold for drift forecasting or detection may be manually set at a pre-selected percentage or probability based on a user's desired level of sensitivity and can be adjusted over time as the system adapts to additional data.” In other words, detecting drift is identifying that the AI model needs to be retrained, threshold is predefined threshold, and automatically retrain is automatically retraining the AI model. Examiner notes that average counterfactual distance is previous mapped to Bodria.) when it is determined that
[the average counterfactual distance is less than the predefined threshold value].
Thus far, Tirupathi does not explicitly teach implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al model.
Carlevaro teaches implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al model (Carlevaro, Algorithm 1, and, page 60849, column 1, paragraph 1, line 7 “Specifically, a change of a certain delta in the features describing the observation x, belonging to class C, leads to the generation of an observation x’ (i.e., the counterfactual of x) that will be classified as belonging to class Cl.” and, page 60850, column 2, paragraph 3, line 1 “The aim of this paper is to introduce a novel methodology for counterfactual generation and validation. The counterfactuals generation method uses regions defined by Two Class-Support Vector Data Descriptors (TC-SVDDs) and is here introduced in both analytical (II-A) and numerical (II-B) form.” And, page 60852, column 2, paragraph 3, line 1 “Algorithm 1 returns the set C of counterfactuals of points belonging to S1. Of course, the same procedure can be applied to find the counterfactuals of the points belonging to S2 simply by reversing the roles of S1 and S2.”
PNG
media_image2.png
648
516
media_image2.png
Greyscale
In other words, x is target sample, x’ is counterfactual, and counterfactual generation is generate a plurality of counterfactuals.)
Carlevaro teaches comparing the (Carlevaro, page 60857, column 2, paragraph 1, line 3 “The factuals (i.e., points of the collision class, which are mapped into the corresponding counterfactuals) are mapped into two classes; the classes label CD values under and above the 0.03 threshold. Values larger
than the threshold represent over dimensioned and almost over dimensioned points, as evidenced in Figure 6a.” In other words, counterfactuals is counterfactuals, .03 is threshold distance, and under and above…threshold is comparing the distance to a predefined threshold. Examiner notes that average counterfactual distance is previously mapped to Bodria.)
Carlevaro teaches the (Carlevaro, page 60857, column 2, paragraph 1, line 3. See above mapping. In other words, under… threshold is counterfactual distance is less than the predefined threshold.)
Both Tirupathi and Carlevaro are directed to artificial intelligence and counterfactuals, among other things. Tirupathi teaches a method for identifying when to retrain an artificial intelligence (Al) model by utilizing one or more processors along with allocated memory, the method comprising: receiving an Al model and a set of target data samples on a predefined time period for identifying when to retrain the Al model, and identifying that the Al model needs to be retrained; but does not explicitly teach implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al model, or comparing the
In view of the teaching of Tirupathi, it would be obvious to one of ordinary skill in the art before the effective date of the claimed invention to combine the teaching of Carlevaro into Tirupathi. This would result in a method for identifying when to retrain an artificial intelligence (Al) model by utilizing one or more processors along with allocated memory, the method comprising: receiving an Al model and a set of target data samples on a predefined time period for identifying when to retrain the Al model, implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al model, comparing the
One of ordinary skill in the art would do this to achieve better control of the event being predicted by the machine learning model (Carlevaro, abstract, line 1 “Increasingly in recent times, the mere prediction of a machine learning algorithm is considered insufficient to gain complete control over the event being predicted. A machine learning algorithm should be considered reliable in the way it allows to extract more knowledge and information than just having a prediction at hand. In this perspective, the counterfactual theory plays a central role. By definition, a counterfactual is the smallest variation of the input such that it changes the predicted behaviour.”)
Thus far, the combination of Tirupathi and Carlevaro does not explicitly teach average counter factual distance, or computing an average counterfactual distance between closest counterfactuals and corresponding target data samples.
Bodria teaches average counterfactual distance ( Bodria, equation 2,
PNG
media_image3.png
66
266
media_image3.png
Greyscale
In other words, ddist is average counterfactual distance.)
Bodria teaches computing an average counterfactual distance between closest counterfactuals and corresponding target data samples (Bodria, equation 2, and page 6, column 2, paragraph 2, line 1 “Metrics. We measure the proximity of a counterfactual x’ with its original sample x, its robustness, the diversity of the counterfactuals found. We express all the evaluation metrics as distances. For proximity and robustness measures, the lower the values, the better are the counterfactuals returned; for diversity measures, the higher, the better. For methods returning more than one counterfactual we selected the best counterfactual among the set returned. We measure proximity in two different fashions [10], [13]. The first one, named ddist, is the average distance between x and the counterfactual x’, the distance is computed using a mixed distance defined in Equation 2:
PNG
media_image3.png
66
266
media_image3.png
Greyscale
In other words, counterfactual is x’, target data sample is x, and equation 2 shows computing an average counterfactual distance between the counterfactuals and target data samples.)
Both Bodria and the combination of Tirupathi and Carlevaro are directed to artificial intelligence and counterfactuals, among other things. The combination of Tirupathi and Carlevaro teaches a method for identifying when to retrain an artificial intelligence (Al) model by utilizing one or more processors along with allocated memory, the method comprising: receiving an Al model and a set of target data samples on a predefined time period for identifying when to retrain the Al model, implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al model, comparing the
In view of the teaching of the combination of Tirupathi and Carlevaro, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Bodria into the combination of Tirupathi and Carlevaro.
This would result in a method for identifying when to retrain an artificial intelligence (Al) model by utilizing one or more processors along with allocated memory, the method comprising: receiving an Al model and a set of target data samples on a predefined time period for identifying when to retrain the Al model, implementing an artificial intelligence technique to generate a plurality of counterfactuals and corresponding target data samples among the set of target data samples for analyzing performance of the received Al model, computing an average counterfactual distance between closest counterfactuals and corresponding target data samples on the predefined time period, comparing the average counterfactual distance to a predefined threshold value, identifying that the Al model needs to be retrained, and automatically retraining the AI model when it is determined the average counterfactual distance is less than the predefined threshold value.
One or ordinary skill in the art would be motivated to do this to better evaluate the predictions of AI models. (Bodria, abstract, line 1 “Artificial Intelligence decision-making systems have dramatically increased their predictive performance in recent years, beating humans in many different specific tasks. However, with increased performance has come an increase in the complexity of the black-box models adopted by the AI systems, making them entirely obscure for the decision process adopted.”)
Regarding claim 2,
The combination of Tirupathi and Bodria teaches the method according to claim 1, wherein
the plurality of counterfactuals are plausible samples on a desired side of a decision boundary generated by perturbing a query sample (Carlevaro, Algorithm 1, 60850, column 2, paragraph 1, line 3 “By contrast, in this study, the search for counterfactuals is guided by directly perturbing only controllable features.” And, page 60852, column 2, paragraph 3, line 1 “Algorithm 1 returns the set C of counterfactuals of points belonging to S1. Of course, the same procedure can be applied to find the counterfactuals of the points belonging to S2 simply by reversing the roles of S1 and S2.” Examiner notes that the specification of the instant application recites “According to exemplary embodiments, the plurality of counterfactuals may
be plausible samples (i.e., y and z as illustrated in FIG. 5) on a desired side of a decision
boundary 502 generated by perturbing a query sample 504 (i.e., x as illustrated in FIG. 5).” (Specification, paragraph [00102], line 1.) Therefore, examiner is interpreting that a plausible sample is any sample on the desired side of a decision boundary. In other words, perturbing controllable features is perturbing a query sample, points belonging to S1 are plausible, and S1 is the set of counterfactuals on the desired side of the decision boundary.)
Regarding claim 3,
The combination of Tirupathi and Bodria teaches the method according to claim 2, wherein
the predefined threshold value is use case or dataset specific and is used to identify whether retraining of the Al model is needed or not (Tirupathi, Figure 2, and, paragraph [0038], line 12 “In embodiments, concept drift detection program 150 may be configured to perform further suitable known model interpretability techniques to gather additional information about a target model's domain, weight or importance assigned to the detected features and variables using known model agnostic interpretability methods such as Local Interpretable Model Agnostic Explanations (LIME), Shapley Additive explanations (SHAP), Anchors, counterfactuals, etc.,” and, paragraph [0041], line 25 “In embodiments, the threshold for drift forecasting or detection may be manually set at a pre-selected percentage or probability based on a user's desired level of sensitivity and can be adjusted over time as the system adapts to additional data.” And, paragraph [0041], line 17 “It also allows concept drift detection program 150 to forecast drift based upon individual features rather than a whole dataset and can be useful to identify which specific features may cause drift.” In other words, features of a whole dataset is dataset specific, threshold is predefined threshold, and detecting drift is identifying that the AI model needs to be retrained. Note, not detecting a drift is identifying that retraining the AI model is not needed.)
Regarding claim 7,
The combination of Tirupathi, Bodria, and Hamman teaches the method according to claim 1, further comprising:
identifying that the Al model does not need to be retrained (Tirupathi, Figure 2, and, paragraph [0038], line 12 “In embodiments, concept drift detection program 150 may be configured to perform further suitable known model interpretability techniques to gather additional information about a target model's domain, weight or importance assigned to the detected features and variables using known model agnostic interpretability methods such as Local Interpretable Model Agnostic Explanations (LIME), Shapley Additive explanations (SHAP), Anchors, counterfactuals, etc.,”, and, paragraph [0041], line 25 “In embodiments, the threshold for drift forecasting or detection may be manually set at a pre-selected percentage or probability based on a user's desired level of sensitivity and can be adjusted over time as the system adapts to additional data.” In other words, detecting drift is identifying that the AI model needs to be retrained, and not detecting drift is determining the AI model does not need to be retrained.) when output data from comparing
indicates that the average counterfactual distance is equal to or more than the predefined threshold value (Carlevaro, Figure 2, and page 60854, column 1, paragraph 1, line 1 “From Figure 2 it is easy to see that the lower the q, the better the counterfactual and if q < 0 then the counterfactual determined is incorrect.”
PNG
media_image4.png
362
606
media_image4.png
Greyscale
In other words, counterfactuals is counterfactuals, from prior mapping, .03 is threshold distance, if q < 0 is correct is the average counterfactual distance is equal to or more than the predefined threshold.).
Claims 8-10 and 14, are system claims that correspond to method claims 1-3, and 7, respectively. Otherwise, they are not patentably distinct. The combination of Tirupathi, Carlevaro, and Bodria teaches a system (Tirupathi, paragraph [0003], line 1 “According to one embodiment, a method, computer system, and computer program product for automatically forecasting and mitigating concept drift in machine learning models using natural language processing is provided.” In other words, computer system is system.) Therefore, claims 8-10, and 14 are rejected for the same reasons as claims 1-3, and 7, respectively.
Claims 15-18, and 20, are non-transitory, computer readable medium claims that correspond to method claims 1-4, and 7, respectively. Otherwise, they are not patentably distinct. The combination of Tirupathi, Carlevaro, and Bodria teaches a non-transitory, computer readable medium (Tirupathi, claim 15, line 1 “A computer program product, the computer program product comprising: one or more computer-readable tangible storage medium and program instructions stored on at least one of the one or more computer-readable tangible storage medium…” In other words, computer-readable tangible storage medium is a non-transitory computer readable medium.). Therefore, claims 15-18, and 20 are rejected for the same reasons as claims 1-4, and 7, respectively.
Claims 4, 11, and 18 are rejected under 35 U.S.C. § 103 as being unpatentable over Tirupathi, Carlevaro, Bodria, and Freiburg, et al (Machine Learning Professorship Freiburg, herein Freiburg).
Regarding claim 4,
The combination of Tirupathi, Carlevaro, and Bodria teaches the method according to claim 2, further comprising:
Thus far, the combination of Tirupathi, Carlevaro and Bodria does not teach displaying how counterfactual distance evolves over time on a two-dimensional graph wherein x-axis of the graph represents days and y-axis of the graph represents average counterfactual distance.
Freiburg teaches displaying how counterfactual distance evolves over time on a two-dimensional graph wherein x-axis of the graph represents days and y-axis of the graph represents average counterfactual distance (Freiburg, Toy Example, and paragraph 1, line 1
“Plot the Performance over Time
Auto-Pytorch uses SMAC to fit individual machine learning algorithms and then ensembles them together using Ensemble Selection (https://www.cs.cornell.edu/~caruana/ctp/ct.papers/caruana.icml04.icdm06long.pdf)
The following examples show how to plot both the performance of the individual models and their respective ensemble.
Additionally, as we are compatible with matplotlib, you can input any args or kwargs that are compatible with ax.plot. In the case when you would like to create multipanel visualization, please input plt.Axes obtained from matplotlib.pyplot.subplots.”
PNG
media_image5.png
506
638
media_image5.png
Greyscale
Examiner notes that it is known in the art to be able to present accuracy or performance over time in a two dimensional graph. Freiburg provides generic code that is applied to “machine learning algorithms” that demonstrates this. The particular values to be displayed, such as counterfactual distance and time does not change the fact that displaying the data on a two-dimensional graph is known in the art. In other words, plot performance over time is displaying how counterfactual distance evolves over time on a two-dimensional graph wherein x-axis of the graph represents days and y-axis of the graph represents average counterfactual distance.)
Both Freiburg and the combination of Tirupathi, Carlevaro and Bodria are dedicated to machine learning algorithms and displaying the output, among other things. The combination of Tirupathi, Carlevaro and Bodria teaches the method of claim 2, but does not explicitly teach displaying how counterfactual distance evolves over time on a two-dimensional graph wherein
x-axis of the graph represents days and y-axis of the graph represents average counterfactual distance. Freiburg teaches displaying how counterfactual distance evolves over time on a two-dimensional graph wherein x-axis of the graph represents days and y-axis of the graph represents average counterfactual distance.
In view of the teaching of Tirupathi, Carlevaro and Bodria, it would be obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teaching of Freiburg into the combination of Tirupathi, Carlevaro, and Bodria. This would result in the method of claim 2, and displaying how counterfactual distance evolves over time on a two-dimensional graph wherein the x-axis of the graph represents days and the y-axis of the graph represents average counterfactual distance.
One of ordinary skill in the art would be motivated to do this to graphically present the results of the machine learning model which makes it easier for users to evaluate performance.
Claim 11 is a system claim that corresponds to method claim 4. Otherwise, they are not patentably distinct. Therefore, claim 11 is rejected for the same reasons as claim 4.
Claim 18 is a non-transitory computer readable medium claim that corresponds to method claim 4. Otherwise, they are not patentably distinct. Therefore, claim 18 is rejected for the same reasons as claim 4.
The prior art made of record and not used is considered pertinent to applicant’s disclosure.
Ge, et al “Counterfactual Evaluation for Explainable AI” discloses two algorithms to find the proper counterfactuals in both discrete and continuous scenarios and then use the acquired counterfactuals to measure faithfulness.
Geada, et al “Trusty AI Explainability Toolkit” discloses a Java and Python library that provides XAI explanations of decision services and predictive models for both enterprise and data science use-cases.
Huang, et al “Accurate, Explainable, and Private Models: Providing Recourse While Minimizing Training Data Leakage” discloses two novel methods to generate differentially private recourse: Differentially Private Model (DPM) and Laplace Recourse (LR), and using logistic regression classifiers and real world and synthetic datasets, it is demonstrated that DPM and LR perform well in reducing what an adversary can infer, especially at low FPR.
Rasouli, et al “Analyzing and Improving the Robustness of Tabular Classifiers using Counterfactual Explanations” discloses a decision boundary modification approach using customized counterfactual data points to improve the robustness of the models without compromising their accuracy, and a framework that addresses the robustness of black-box classifiers in the tabular setting, which is considered an under-explored research area.
Allowable Subject Matter
The claims have been searched but no prior art which anticipates or renders the following claims obvious have been found.
Claims 5, 12, and 19 are objected to as being dependent upon a rejected base claim, but would be allowable if found eligible and rewritten in independent form including all of the limitations of the base claim and any intervening claims. Dependent claims 6, and 13 inherit the particular limitations from claims 5 and 12 and are similarly objected to.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BART RYLANDER whose telephone number is (571)272-8359. The examiner can normally be reached Monday - Thursday 8:00 to 5:30.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Miranda Huang can be reached at 571-270-7092. 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.
/Bart I Rylander/Examiner, Art Unit 2124
Prosecution Insights