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 .
Amendments
This action is in response to amendments filed February 12th, 2026, in which Claims 1, 6, 10, 12, 17, and 19 have been amended. No claims have been cancelled nor added, and the amendments have been entered. Claims 1-20 are currently pending.
Claim Rejections - 35 USC § 112
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.
Claim 12 is rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, 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, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
In the specification as originally filed, the only written description support for determining that the score (currently second output) is within a predetermined threshold from a cutoff point comes from the claim language of Claim 12 itself (which is part of the disclosure) – note that that no threshold from a cutoff point is in the text of the description, only a cutoff threshold where the cutoff is itself a threshold for the decision making. Therefore, there is no support at all in the specification as originally filed for amended Claim 12, including both altering a decision (i.e. making the replaced value fall on the other side of a decision threshold) and the altering the plurality of values be responsive to determining the output is within some predetermined threshold.
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 towards an abstract idea, without significantly more.
Claim 1 recites a system, comprising: a non-transitory memory and one or more hardware processors, thus a machine, one of the four statutory categories of patentable subject matter. However, Claim 1 further recites the steps of determining, for each input feature in the plurality of input features, a corresponding contribution of the input feature to a correct classification or an incorrect classification in classifying one or more transactions by the machine learning model based on analyzing changes to model outputs of the machine learning model (a mental process that consists purely of analyzing data); determining a plurality of impact scores for the plurality of input features based on the corresponding contribution determined for each input feature in the plurality of input features (a mental process); determining that one or more particular input features in the plurality of input features have an adverse effect on performing a correct transaction by a machine learning model based on the plurality of impact scores indicating that the one or more particular input features positively contributes to the incorrect classification and/or negatively contributes to the correct classification (a mental process); determining replacement values for one or more first values in the plurality of values that correspond to the one or more particular input features, wherein the replacement values, when input to the machine learning model, change a category output produced by the machine learning model for the first transaction (a mental process); modifying a plurality of values associated with the first transaction based on the determining …, wherein the modifying comprises replacing the one or more first values from the plurality of values with the replacement values (a mental process); to determine based on the modified plurality of values, a first category for the first transaction that is different from a second category (a mental process); and processing the first transaction according to the first category (a particular method of organizing human activity). Thus, the claim recites an abstract idea of determine which input values lead to an adverse determination, changing them, and making a new determination.
Further, the claim does not recite any additional element which could integrate this abstract idea into a practical application, because the additional elements recited of consist of:
a non-transitory memory; and one or more hardware processors coupled with the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to, that is, generic computer components on which to implement the abstract idea (see MPEP 2106.05(f));
receiving a request for categorizing a first transaction; obtaining a plurality of values associated with the first transaction and corresponding to a plurality of data features, that is, insignificant extra-solution activity of data gathering (see MPEP 2106.05(g))
providing the modified plurality of values and different feature values corresponding to the input feature, both as input values to a machine learning model configured to generate an output based on the input, which is merely “using a computer or other machinery” as a tool to perform the abstract idea step of generating an output (see MPEP 2106.05(f))
none of which can integrate the abstract idea into a practical application. Thus the claim is directed towards the abstract idea.
Finally, none of the additional elements, taken alone or in combination, provide significantly more than the abstract idea itself, because performance of an abstract idea on a computer or other machinery (“apply it”) cannot provide an inventive concept (see MPEP 2106.05(f)) and collecting data is well-understood, routine, and conventional (MPEP 2106.05(d), “transmitting or receiving data over a network). The claim is ineligible.
Claim 2, dependent upon Claim 1, only provides further definition of the abstract idea step of replacing, i.e. without modifying remaining values, but recites no additional elements, thus no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claim 3, dependent upon Claim 1, only recites an additional mathematical process step of calculating, for each input feature in the plurality of input features, a Shapley value and provides further definition of the abstract idea step of determining the plurality of impact scores, i.e. based on the Shapely values, but recites no additional elements, thus no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claims 4 and 5, dependent upon Claim 1, merely recite to determine a second transaction that has been correctly/incorrectly categorize by the machine learning model (a mental process) and then repeats steps of Claim 1 on the second transaction, and thus recites no new additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claims 6-8, dependent upon Claim 1, merely recite additional mental process steps (Claim 6: determining, replacing; Claim 7: determining based on …; Claim 8: determining a value distribution based on …) and repeats the use of the machine learning model to perform a mental process (MPEP 2106.05(f)), but no new additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claim 9, dependent upon Claim 8, only provides further definition of the abstract idea step of replacing, i.e. wherein the first replacement value is one of …, but recites no additional elements, thus no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claim 10 recites a method, one of the four statutory categories of patentable subject matter. However, the claim further recites the steps of determining, for each input feature in the plurality of input features, a corresponding contribution of the input feature to a correct classification or an incorrect classification in classifying one or more transactions by the machine learning model based on analyzing changes to model outputs of the machine learning model (a mental process that consists purely of analyzing data); determining a plurality of impact scores for the plurality of input features based on the corresponding contribution determined for each input feature in the plurality of input features (a mental process); determining, from the plurality of input features, one or more particular input features that have an adverse effect on performing a correct event classification by [a] machine learning model based on the plurality of impact scores indicating that the one or more particular input features positively contributes to the correct classification (a mental process); identifying, from [a] plurality of values associated with [a] first event, one or more first values corresponding to one or more particular input features (a mental process); determining replacement values for the one or more first values that change a classification output (a mental process); altering the plurality of values based on replacing the one or more first values from the plurality of values with one or more second values (a mental process); to generate an output based on to generate an output based on the input values, i.e. the altered plurality of values (a mental process) and determine a category for the first transaction based on the output (a mental process). Thus, the claim recites an abstract idea of determine which input values lead to an adverse determination, changing them, and making a new determination.
Further, the claim does not recite any additional element which could integrate this abstract idea into a practical application, because the additional elements recited of consist of:
receiving a request for categorizing a first event; obtaining a plurality of values associated with the first event and corresponding to a plurality of data features, that is, insignificant extra-solution activity of data gathering (see MPEP 2106.05(g))
providing the altered plurality of values as input values to a machine learning model configured to generate an output based on the input, which is merely “using a computer or other machinery” as a tool to perform the abstract idea step of generating an output (see MPEP 2106.05(f))
the fact that this is all performed by a computer system, which is merely “using a computer or other machinery” as a tool to perform the abstract idea (see MPEP 2106.05(f)(2))
none of which can integrate the abstract idea into a practical application. Thus the claim is directed towards the abstract idea.
Finally, none of the additional elements, taken alone or in combination, provide significantly more than the abstract idea itself, because performance of an abstract idea on a computer or other machinery (“apply it”) cannot provide an inventive concept (see MPEP 2106.05(f)) and collecting data is well-understood, routine, and conventional (MPEP 2106.05(d), “transmitting or receiving data over a network). The claim is ineligible.
Claim 11, dependent upon Claim 10, recites training the machine learning model to perform the mental process of to generate an output based on the input, and thus recites an additional element which is no more than part of the use of a computer or other machinery as a tool (i.e. “apply it”) of the independent claim, and thus (by MPEP 2106.05(f)) cannot integrate the abstract idea into a practical application nor provide significant more than the abstract idea itself.
Claim 12, dependent upon Claim 10, only provides further definition of the abstract idea step of altering the plurality of values, i.e. by the mental process of determining a second output, determining that the second output is within a threshold, and altering responsive to the determining that the second output is within the threshold, but recites no additional elements, thus no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claim 13, dependent upon Claim 10, only specifies the particular technological environment in which the abstract idea takes place (i.e. the machine learning model comprises a neural network or a gradient boosting tree) which by MPEP 2106.05(h) cannot integrate the abstract idea into a practical application nor provide significantly more than the abstract idea itself.
Claim 14, dependent upon Claim 10, only provides further definition of the abstract idea step of replacing the one or more first values, i.e. by replacing without modifying remaining values, but no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claims 15 and 16, dependent upon Claim 10, merely recite to determine a second event that has been correctly/incorrectly categorize by the machine learning model (a mental process) and then repeats steps of Claim 10 on the second event, and thus recites no new additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claim 17 recites a non-transitory machine readable medium, thus an article of manufacture, one of the four statutory categories of patentable subject matter. However, the claim further recites the steps of determining, for each data feature in the plurality of data features, a corresponding contribution of the data feature to a correct classification or an incorrect classification in classifying one or more transactions by the machine learning model based on analyzing changes to model outputs of the machine learning model (a mental process that consists purely of analyzing data); determining a plurality of impact scores for the plurality of data features based on the corresponding contribution determined for each data feature in the plurality of input features (a mental process); determining that one or more particular data features in a plurality of data features have an adverse effect on performing a risk assessment by a machine learning model based on the plurality of impact scores indicating that the one or more particular input features positively contributes to the incorrect classification and/or negatively contributes to the correct classification (a mental process); determining replacement values for one or more first values in the plurality of values that correspond to the one or more particular input features, wherein the replacement values, when input to the machine learning model, change a category output produced by the machine learning model for the first transaction (a mental process); altering the plurality of values based on the determining that one or more particular data features in a plurality of data features have an adverse effect on performing a risk assessment by a machine learning model (a mental process); replacing one or more first values associated with a first transaction with the replacement values (a mental process); to generate an output based on to generate an output based on the input values, i.e. the altered plurality of values (a mental process) and determining a risk score for the first transaction based on the output (a mental process). Thus, the claim recites an abstract idea of determine which input values lead to an adverse determination, changing them, and making a new determination.
Further, the claim does not recite any additional element which could integrate this abstract idea into a practical application, because the additional elements recited of consist of:
a non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations, that is, generic computer components on which to implement the abstract idea (see MPEP 2106.05(f));
receiving a request for a risk associated with a first transaction; obtaining a plurality of values associated with the first transaction and corresponding to a plurality of data features, that is, insignificant extra-solution activity of data gathering (see MPEP 2106.05(g))
providing the altered plurality of values as input values to a machine learning model configured to generate an output based on the input, which is merely “using a computer or other machinery” as a tool to perform the abstract idea step of generating an output (see MPEP 2106.05(f))
none of which can integrate the abstract idea into a practical application. Thus the claim is directed towards the abstract idea.
Finally, none of the additional elements, taken alone or in combination, provide significantly more than the abstract idea itself, because performance of an abstract idea on a computer or other machinery (“apply it”) cannot provide an inventive concept (see MPEP 2106.05(f)) and collecting data is well-understood, routine, and conventional (MPEP 2106.05(d), “transmitting or receiving data over a network). The claim is ineligible.
Claim 18, dependent upon Claim 17, recites the additional mental process step classifying the first transaction based on the risk score but no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself, because the only new additional element is an insignificant application of processing the first transaction based on the classification, which includes an approval or a denial of the first transaction (see MPEP 2106.05(f)(3), “the particularity or generality of the application”).
Claim 19, dependent upon Claim 17, recites the additional mental process steps of determining, for a first data feature from the plurality of data features, a first replacement value (mental process) and replacing a first one of the one or more first values with the first replacement value (mental process) but no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claim 20, dependent upon Claim 19, only provides further definition of the abstract idea step of replacing a first one of the one or more first values, i.e. by replacing with a replacement value determine based on first feature values associated with a plurality of previously processed transactions and corresponding to the first data feature, but no additional elements which could integrate the abstract idea into a practical application or provide significantly more than the abstract idea itself.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-9; 10, 11, 13-16; and 17-20 are rejected under 35 U.S.C. 102(2) as being anticipated by Miroshnikov, US PG Pub 2021/0383268.
Regarding Claim 1, Miroshnikov teaches a system comprising: a non-transitory memory; and one or more hardware processors coupled with the non-transitory memory and configured to read instructions from the non-transitory memory (Miroshnikov, [0011], “the system includes a memory and one or more processors coupled to the memory”) to cause the system to perform operations comprising: receiving a request for categorizing a first transaction (Miroshnikov, [0081], “to receive a request to determine whether to extend credit to an applicant”); obtaining a plurality of values associated with the first transaction, wherein the plurality of values corresponds to a plurality of input features associated with a machine learning model, wherein the machine learning model is configured to classify transactions based on the plurality of input features (Miroshnikov, [0081-0082], “The request may include information that identifies the applicant … The input vector generator can then access various databases to collect information about the applicant … the various information from one or more sources are collected by the input vector generator and compiled into an input vector. The input vector is transmitted to the AI engine, which processes the input vector via an ML model … the output vector can be a binary value that indicates whether the application to extend credit to the applicant is denied or accepted” ); determining, for each input feature in the plurality of input features, a corresponding contribution of the input feature to a correct classification or an incorrect classification in classifying one or more transactions by the machine learning model based on analyzing changes to model outputs of the machine learning model when different feature values corresponding to the input features are provided to the machine learning model (Miroshnikov, [0045], “Given a subvector” i.e. input feature in the plurality of input features, “then an explainer function E can be denoted … where the explainer function E provides a measure of the subvector to the score value” i.e. classification, “typical implementations may utilize … Shapley Additive Explanations (SHAP) values”); determining a plurality of impact scores for the plurality of input features based on the corresponding contribution determined for each input feature in the plurality of input features (Miroshnikov, [0070, “a bias contribution value is calculated for each group of input values … the bias contribution value is based on a cumulative distribution function (CDF) for the score explainer function”); determining that one or more particular input features in the plurality of input features have an adverse effect on performing a correct transaction categorization by the machine learning model based on the impact scores indicating that the one or more particular input features positively contributes to the incorrect classification … by the machine learning model (Miroshnikov, [0023], “The bias contribution level for each grouped predictor can be separated into to values: a positive bias and a negative bias … For example, where a model exhibits a bias disfavoring Hispanic populations over White populations, the bias explanation can produce a rank of all model attributes by contributions to the differences in outcome of the model” where a biased classification is an incorrect classification); determining replacement values for one or more first values in the plurality of values that correspond to the one or more particular input features, wherein the replacement values, when input to the machine learning model, change a category output produced by the machine learning model for the first transaction; modifying the plurality of values associated with the first transaction based on the determining that the one or more particular input features have the adverse effect on performing the correct transaction categorization by the machine learning model, wherein the modifying comprises replacing the one or more first values from the plurality of values with the replacement values (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model. Neutralization refers to replacing certain predictions in the input for a sample with neural values (e.g. median or mean values for a population). Partial neutralization refers to replacing certain predictors with a scaled value for the predictor, as defined by a transformation” where “mitigation” of bias is changing an output from a biased output to an unbiased output); providing the modified plurality of values as input values to the machine learning model, wherein the machine learning model is configured to determine, based on the modified plurality of values, a first category for the first transaction that is different from a second category determined by the machine learning model based on the plurality of values without any modification (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model” where “processing the predictors for a new sample with the model” denotes determine a category by the machine learning model and “mitigating” the bias of the model denotes changing the output in some given situation); and processing the first transaction according to the first category (Miroshnikov, [0081], “to extend credit to an applicant” using the fair unbiased decision & [0083], “to send a response to a client device … that indicates whether the credit application was accepted or denied”).
Regarding Claim 2, Miroshnikov teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated) Miroshnikov further teaches wherein the replacing the one or more first values is performed without modifying remaining values from the plurality of values (Miroshnikov, [0024], “Neutralization refers to replacing certain predictors” i.e. not all of the predictors, also see [0051], “the values of the input variables in subvector
X
S
are replaced by new values” i.e. only a sub-portion of the predictors are replaced).
Regarding Claim 3, Miroshnikov teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated) Miroshnikov has already been shown to teach calculating, for each input feature of the plurality of input features, a Shapley value representing a contribution of the input feature to an output value of the machine learning model, wherein the determining the plurality of impact scores is further based on the Shapley value calculated for each input feature in the plurality of input features (Miroshnikov, [0045], “then an explainer function E can be denoted … where the explainer function E provides a measure of the subvector to the score value … typical implementations may utilize … Shapley Additive Explanations (SHAP) values” where the impact score/“bias contribution value” is based on the SHAP values via the “score explainer function”).
Regarding Claim 4, Miroshnikov teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated) Miroshnikov further teaches determining a second transaction that has been incorrectly categorized by the machine learning model; accessing a second plurality of values associated with the second transaction and corresponding to the plurality of input features; and determining, from the second plurality of values, one or more values that contributed to an incorrect categorization of the second transaction by iteratively: modifying each value in the second plurality of values; and categorizing, using the machine learning, the second transaction based on unaltered values in the second plurality of values in addition to the modified value ([0046], “over a certain collection of N samples from the trained data set” indicates that Miroshnikov performs their bias mitigation learning and performance over multiple, not just one, example, see also [0023], “The bias contribution level for each grouped predictor can be separated into to values: a positive bias and a negative bias … For example, where a model exhibits a bias disfavoring Hispanic populations over White populations, the bias explanation can produce a rank of all model attributes by contributions to the differences in outcome of the model” indicates that many transactions are analyzed in the same way as the first transaction, some of which have a biased/incorrect classification and some of which have correct categorizations).
Regarding Claim 5, Miroshnikov teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated) Miroshnikov further teaches determining a second transaction that has been correctly categorized by the machine learning model; accessing a second plurality of values associated with the second transaction and corresponding to the plurality of input features; and determining, from the second plurality of values, one or more values that contributed to an incorrect categorization of the second transaction by iteratively: modifying each value in the second plurality of values; and categorizing, using the machine learning, the second transaction based on unaltered values in the second plurality of values in addition to the modified value ([0046], “over a certain collection of N samples from the trained data set” indicates that Miroshnikov performs their bias mitigation learning and performance over multiple, not just one, example, see also [0023], “The bias contribution level for each grouped predictor can be separated into to values: a positive bias and a negative bias … For example, where a model exhibits a bias disfavoring Hispanic populations over White populations, the bias explanation can produce a rank of all model attributes by contributions to the differences in outcome of the model” indicates that many transactions are analyzed in the same way as the first transaction, some of which have a biased/incorrect classification and some of which have correct categorizations).
Regarding Claim 6, Miroshnikov teaches the system of Claim 1 (and thus the rejection of Claim 1 is incorporated). Miroshnikov further determining, for a first input feature from the plurality of input features, a first contribution of the input feature to a first output of the machine learning model based on providing a first plurality of feature values corresponding to the input feature to the machine learning model, wherein the first input feature is one of the one or more particular input features (Miroshnikov, [0045], “then an explainer function E can be denoted … where the explainer function E provides a measure of the subvector to the score value … typical implementations may utilize … Shapley Additive Explanations (SHAP) values” where the bias contribution value is based on the SHAP values via the first plurality of feature values/training set, see [0046], “over a certain collection of N samples from the trained data set”; determining a first replacement feature based on the first plurality of feature values; replacing a first one of the one or more first values with the first replacement value (Miroshnikov, [0050-0051], “the values of the input variables in subvector
X
S
are replaced by new values that are equal to the median value for those input variables taken from the training data set”).
Regarding Claim 7, Miroshnikov teaches the system of Claim 6 (and thus the rejection of Claim 6 is incorporated). Miroshnikov further teaches wherein the first replacement value is determined based on selecting a feature value from the first plurality of feature values (Miroshnikov, [0050-0051], “the values of the input variables in subvector
X
S
are replaced by new values that are equal to the median value for those input variables taken from the training data set”).
Regarding Claim 8, Miroshnikov teaches the system of Claim 6 (and thus the rejection of Claim 6 is incorporated). Miroshnikov further teaches determining a value distribution based on the first plurality of feature values, wherein the first replacement value is determined based further on the value distribution (Miroshnikov, [0050-0051], “the values of the input variables in subvector
X
S
are replaced by new values that are equal to the median value for those input variables taken from the training data set … any other quantile value can be used … in lieu of the median value”).
Regarding Claim 9, Miroshnikov teaches the system of Claim 8 (and thus the rejection of Claim 8 is incorporated). Miroshnikov further teaches wherein the first replacement value is one of a mean, a maximum, or a minimum associated with the value distribution (Miroshnikov, [0051], “a mean value could be used instead of the median value”).
Regarding Claim 10, Miroshnikov teaches a method comprising … by a computer system (Miroshnikov, [0011], “the system includes a memory and one or more processors coupled to the memory”) receiving a request for classifying a first event (Miroshnikov, [0081], “to receive a request to determine whether to extend credit to an applicant”); obtaining a plurality of values associated with the first event, wherein the plurality of values corresponds to a plurality of input features for a machine learning model (Miroshnikov, [0081-0082], “The request may include information that identifies the applicant … The input vector generator can then access various databases to collect information about the applicant … the various information from one or more sources are collected by the input vector generator and compiled into an input vector. The input vector is transmitted to the AI engine, which processes the input vector via an ML model … the output vector can be a binary value that indicates whether the application to extend credit to the applicant is denied or accepted” ); determining, for each input feature in the plurality of input features, a corresponding contribution of the input feature to a correct classification or an incorrect classification in classifying one or more events by the machine learning model based on analyzing changes to model outputs of the machine learning model when different feature values corresponding to the input features are provided to the machine learning model (Miroshnikov, [0045], “Given a subvector” i.e. input feature in the plurality of input features, “then an explainer function E can be denoted … where the explainer function E provides a measure of the subvector to the score value” i.e. classification, “typical implementations may utilize … Shapley Additive Explanations (SHAP) values”); determining a plurality of impact scores for the plurality of input features based on the corresponding contribution determined for each input feature in the plurality of input features (Miroshnikov, [0070, “a bias contribution value is calculated for each group of input values … the bias contribution value is based on a cumulative distribution function (CDF) for the score explainer function”); determining from the plurality of input features, one or more particular input features that have an adverse effect on performing a correct event classification by the machine learning model based on the plurality of impact scores indicating that the one or more particular input features positively contributes to the incorrect classification … by the machine learning model (Miroshnikov, [0023], “The bias contribution level for each grouped predictor can be separated into to values: a positive bias and a negative bias … For example, where a model exhibits a bias disfavoring Hispanic populations over White populations, the bias explanation can produce a rank of all model attributes by contributions to the differences in outcome of the model” where a biased classification is an incorrect classification); identifying from the plurality of values associated with the first event, one or more first values corresponding to the one or more particular input features; determining replacement values for the one or more first values, that, when input to the machine learning model, change a classification output produced by the machine learning model; altering the plurality of values based on replacing the one or more first values from the plurality of values with the replacement values (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model. Neutralization refers to replacing certain predictions in the input for a sample with neural values (e.g. median or mean values for a population). Partial neutralization refers to replacing certain predictors with a scaled value for the predictor, as defined by a transformation” where “mitigation” of bias is changing an output from a biased output to an unbiased output); providing the altered plurality of values as input values to the machine learning model, wherein the machine learning model is configured to generate a first output based on the altered plurality of values; and determining a classification for the first event based on the output from the machine learning model (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model” where “processing the predictors for a new sample with the model” denotes determine a classification by the machine learning model and “mitigating” the bias of the model denotes changing the output in some given situation).
Regarding Claim 11, Miroshnikov teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated). Miroshnikov further teaches training the machine learning model based on data values corresponding to the plurality of input features and associated with a plurality of events (Miroshnikov, Abstract, “training the model based on at training data set”).
Regarding Claim 13, Miroshnikov teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated). Miroshnikov further teaches wherein the machine learning model comprises a neural network or a gradient boosting tree (Miroshnikov, [0027], “the ML algorithm can comprise a neural network … as well as tree-based methods such as … gradient boosting machines”).
Regarding Claim 14, Miroshnikov teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated) Miroshnikov further teaches wherein the replacing the one or more first values is performed without modifying remaining values from the plurality of values (Miroshnikov, [0024], “Neutralization refers to replacing certain predictors” i.e. not all of the predictors, also see [0051], “the values of the input variables in subvector
X
S
are replaced by new values” i.e. only a sub-portion of the predictors are replaced).
Regarding Claim 15, Miroshnikov teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated) Miroshnikov further teaches determining a second event that has been incorrectly categorized by the machine learning model; accessing a second plurality of values associated with the second event corresponding to the plurality of input features; and determining, from the second plurality of values, one or more values that contributed to an incorrect categorization of the second event ([0046], “over a certain collection of N samples from the trained data set” indicates that Miroshnikov performs their bias mitigation learning and performance over multiple, not just one, example, see also [0023], “The bias contribution level for each grouped predictor can be separated into to values: a positive bias and a negative bias … For example, where a model exhibits a bias disfavoring Hispanic populations over White populations, the bias explanation can produce a rank of all model attributes by contributions to the differences in outcome of the model” indicates that many transactions are analyzed in the same way as the first transaction, some of which have a biased/incorrect classification and some of which have correct categorizations).
Regarding Claim 16, Miroshnikov teaches the method of Claim 10 (and thus the rejection of Claim 10 is incorporated) Miroshnikov further teaches Miroshnikov further teaches determining a second event that has been correctly categorized by the machine learning model; accessing a second plurality of values associated with the second event corresponding to the plurality of input features; and determining, from the second plurality of values, one or more values that failed to contribute to a categorization of the second event ([0046], “over a certain collection of N samples from the trained data set” indicates that Miroshnikov performs their bias mitigation learning and performance over multiple, not just one, example, see also [0023], “The bias contribution level for each grouped predictor can be separated into to values: a positive bias and a negative bias … For example, where a model exhibits a bias disfavoring Hispanic populations over White populations, the bias explanation can produce a rank of all model attributes by contributions to the differences in outcome of the model” indicates that many transactions are analyzed in the same way as the first transaction, some of which have a biased/incorrect classification and some of which have correct categorizations).
Regarding Claim 17, Miroshnikov teaches a non-transitory computer readable medium having stored thereon machine readable instructions executable to cause a machine to perform operations (Miroshnikov, [0011], “the system includes a memory and one or more processors coupled to the memory”) comprising: receiving a request for determining a risk associated with a first transaction (Miroshnikov, [0081], “to receive a request to determine whether to extend credit to an applicant”); obtaining a plurality of values associated with the first transaction, wherein the plurality of values corresponds to a plurality of data features associated with a machine learning model (Miroshnikov, [0081-0082], “The request may include information that identifies the applicant … The input vector generator can then access various databases to collect information about the applicant … the various information from one or more sources are collected by the input vector generator and compiled into an input vector. The input vector is transmitted to the AI engine, which processes the input vector via an ML model … the output vector can be a binary value that indicates whether the application to extend credit to the applicant is denied or accepted” ); determining, for each data feature in the plurality of data features, a corresponding contribution of the data feature to a correct classification or an incorrect classification in classifying one or more transactions by the machine learning model based on analyzing changes to model outputs of the machine learning model when different feature values corresponding to the data features are provided to the machine learning model (Miroshnikov, [0045], “Given a subvector” i.e. input feature in the plurality of input features, “then an explainer function E can be denoted … where the explainer function E provides a measure of the subvector to the score value” i.e. classification, “typical implementations may utilize … Shapley Additive Explanations (SHAP) values”); determining a plurality of impact scores for the plurality of data features based on the corresponding contribution determined for each data feature in the plurality of data features (Miroshnikov, [0070, “a bias contribution value is calculated for each group of input values … the bias contribution value is based on a cumulative distribution function (CDF) for the score explainer function”); determining that one or more particular input features in the plurality of data features have an adverse effect on performing a correct transaction categorization by the machine learning model based on the impact scores indicating that the one or more particular input features positively contributes to the incorrect classification … by the machine learning model (Miroshnikov, [0023], “The bias contribution level for each grouped predictor can be separated into to values: a positive bias and a negative bias … For example, where a model exhibits a bias disfavoring Hispanic populations over White populations, the bias explanation can produce a rank of all model attributes by contributions to the differences in outcome of the model” where a biased classification is an incorrect classification); determining replacement values for one or more first values in the plurality of values that correspond to the one or more particular input features, wherein the replacement values, when input to the machine learning model, change a category output produced by the machine learning model for the first transaction; altering the plurality of values associated with the first transaction based on the determining that the one or more particular data features have the adverse effect on performing the risk assessment by the machine learning model, wherein the altering comprises replacing one or more first values from the plurality of values with the replacement values (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model. Neutralization refers to replacing certain predictions in the input for a sample with neural values (e.g. median or mean values for a population). Partial neutralization refers to replacing certain predictors with a scaled value for the predictor, as defined by a transformation” where “mitigation” of bias is changing an output from a biased output to an unbiased output); providing the altered plurality of values as input values to the machine learning model, wherein the machine learning model is configured to generate an output based on the input values; and determining a risk score the first transaction based on the output from the machine learning model (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model” where “processing the predictors for a new sample with the model” to determine whether to extend credit or not denotes determine a risk score by the machine learning model).
Regarding Claim 18, Miroshnikov teaches the non-transitory machine-readable medium of Claim 17 (and thus the rejection of Claim 17 is incorporated). Miroshnikov further teaches classifying the first transaction based on the risk score, and processing the first transaction according to the first category (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model” where “processing the predictors for a new sample with the model” to determine whether to extend credit or not denotes classifying the first transaction based on the risk score & [0081], “to extend credit to an applicant” using the fair unbiased decision & [0083], “to send a response to a client device … that indicates whether the credit application was accepted or denied” denotes processing the first transaction).
Regarding Claim 19, Miroshnikov teaches the non-transitory machine-readable medium of Claim 17 (and thus the rejection of Claim 17 is incorporated). Miroshnikov further teaches wherein determining the replacement values comprises: determining, for a first data feature from the one or more particular data features, a first replacement value; and replacing a first one of the first one or more values with the first replacement values (Miroshnikov, [0024], “the bias of the model can be mitigated by neutralizing or partially neutralizing certain predictors before processing the predictors for a new sample with the model. Neutralization refers to replacing certain predictions in the input for a sample with neural values (e.g. median or mean values for a population)”).
Regarding Claim 20, Miroshnikov teaches the non-transitory machine-readable medium of Claim 19 (and thus the rejection of Claim 19 is incorporated). Miroshnikov further teaches wherein the first replacement value is determined based on first feature values associated with a plurality of previously processed transactions and corresponding to the first data feature (Miroshnikov, [0050-0051], “the values of the input variables in subvector
X
S
are replaced by new values that are equal to the median value for those input variables taken from the training data set”).
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over Miroshnikov, in view of Nourian, US PG Pub 2021/0049503.
Miroshnikov does not teach, but Nourian teaches prior to the altering the plurality of values, determining, using the machine learning model, a second output based on unaltered values in the plurality of values associated with the first event (Nourian, [0066], “Referring to FIG. 18, correction example searches may be provided to indicate possible changes to an instances feature values that would shift the projected score to the other side of the decision threshold”); and determining that the second output is within a predetermined threshold from a cutoff point (Nourian, [0066], “Among such shifts, those that are minimal and plausible may be selected”), wherein the first event is classified based on the score with respect to the cutoff point, and wherein the altering the plurality of values is responsive to determining that the second output is within the predetermined threshold from the cutoff point (Nourian, [0066], “Referring to FIG. 18, correction example searches may be provided to indicate possible changes to an instances feature values that would shift the projected score to the other side of the decision threshold .. Among such shifts, those that are minimal and plausible may be selected” where “minimal” denotes a predetermined threshold). It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to incorporate the feature-set reduction of Nourian into the bias mitigation scheme of Miroshnikov. The motivation to do so is to remove features that do not effect the final decision.
Response to Arguments
Applicant’s arguments filed February 12th, 2026 have been fully considered, but are not fully persuasive.
Applicant’s arguments regarding the 35 U.S.C. 101 rejections of the previous office action have been fully considered, but are unpersuasive.
Applicant first argues that “when viewed as a whole, amended Claim 1 integrates the abstract idea into a practical application” because “the above highlighted additional elements recited in the claim integrate the abstract idea into a practical application of improving the performance of a machine learning model based on selectively modifying a subset of the input values provided to the machine learning model.” First, note that the highlighted features of pg. 11 of the response consist not of additional elements, but of mental process steps of determining and of modifying values. Further, the claims as a whole merely recite determining that one or more particular features have an adverse effect and determining impact scores and then modifying the features – the independent claims fail to recite any details as to how the features are identified or modified, i.e. fails to recite the steps that achieve the improvement. By MPEP 2106.05(f)(1), “the claim recites only the idea of a solution or outcome, i.e. the claim fails to recite details of how a solution to a problem is accomplished”. The independent claims are no more detailed than “determine that some input values cause problems, and then replace them” which is not a technical solution to a technical problem.
Further, even if particular solutions to a problem of “data that provides an adverse effect” were recited, it would fail to be a solution to a technical problem. The machine learning problem is merely a computer or other machinery, used as a tool to perform the process to determine an output based on an input. The solution to misclassified outputs recited in the claims (“alter the inputs”) is a same solution that would be appropriate if a human being were making the classification decision (i.e. if a human makes a wrong decision based on faulty data, then provide modified non-faulty data). Thus, the independent claims only recite an improvement in the abstract idea of determining an output based on an input by providing better input data, which is not a technical improvement in a computer or other technology.
Applicant further argues that “similar to the claim in Example 39, Claim 1 as amended herein does not recite any mathematical relationships, formulas, or calculations.” The rejection does not assert that Claim 1 recites any mathematical calculations. Applicant continues to argue “Claim 1 as amended herein is directed to improving prediction performance of a machine learning model, which ‘are not practically performed in the human mind’”. However, this assertion does not properly follow the subject matter eligibility guidance, which first asks whether the claim recites any abstract ideas, which Claim 1 clearly does: determining that features have an adverse effect; modifying values; generating an output based on an input; determining a category. Example 39 only states that a claim that recites no abstract ideas at all cannot be rejected under 35 U.S.C. 101 as directed to an abstract idea without significantly more; Example 39 provides no guidance in a case where abstract ideas/mental processes are recited in the claim language.
Applicant’s arguments with respect the prior art rejections of the previous office action have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Conclusion
Applicant’s amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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 BRIAN M SMITH whose telephone number is (469)295-9104. The examiner can normally be reached Monday - Friday, 8:00am - 4pm Pacific.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Kakali Chaki can be reached on (571) 272-3719. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BRIAN M SMITH/ Primary Examiner, Art Unit 2122
Prosecution Insights