DETAILED ACTION
This action is in response to the filing on 06/08/2026. Claims 1, 3, 6, 8-9, 11, 14, and 16-26 are pending and have been considered below.
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 .
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 U.S.C. 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claims 19-20 rejected under 35 U.S.C. 112(d) or pre-AIA 35 U.S.C. 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends.
Claims 19 and 20 fail to further the limit of the subject matter of claims 1 and 9 upon which they depend, the claims merely change the phrase “a first equation” from claims 1 and 9 to “the first equation” in claims 19-20 and restate a portion of claims 1 and 9 verbatim.
Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
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, 3, 6, 8-9, 11, 14, 16-18, and 21-26 are rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more.
Independent Claims 1, 8, and 9
Step 1:
Claims 1, 8, and 9 recite a method, manufacture, and system respectively; therefore, they are directed to one of the four categories of statutory subject matter (process/method, machine/product/apparatus, manufacture, or composition of matter).
Step 2A Prong 1:
Claim 1 recites a method comprising:
A method of classifying data, the method comprising: — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
training a classification model for classifying input data into at least one class, such that a first output value is generated — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)).
training a classification model for classifying input data into at least one class, such that a first output value is generated, wherein a first equation comprises an equation corresponding to a Bayes’ rule representing a probability of the input data being classified as each of the at least one class, and wherein training the classification model comprises replacing a label distribution component in the first equation with a uniform prior, wherein the uniform prior pu(y=c) = 1/C, where C denotes a total number of classes, wherein the training the classification model comprises training the classification model using only a distribution of samples x from source data and a conditional distribution of the samples x given labels y — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically mathematical formulas or equations (see MPEP § 2106.04(a)(2)(I)(B)).
generating a second output value by applying, to the first output value, information indicating a label distribution of target data — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
classifying facial images into the at least one class by using the second output value — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 8 recites a system comprising:
executing the method of claim 1 — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 9 recites a system comprising:
training a classification model for classifying input data into at least one class, such that a first output value is generated — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)).
training a classification model for classifying input data into at least one class, such that a first output value is generated, wherein a first equation comprises an equation corresponding to a Bayes’ rule representing a probability of the input data being classified as each of the at least one class, and wherein training the classification model comprises replacing a label distribution component in the first equation with a uniform prior, wherein the uniform prior pu(y=c) = 1/C, where C denotes a total number of classes, wherein the training the classification model comprises training the classification model using only a distribution of samples x from source data and a conditional distribution of the samples x given labels y — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically mathematical formulas or equations (see MPEP § 2106.04(a)(2)(I)(B)).
generating a second output value by applying, to the first output value, information indicating a label distribution of target data — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
classifying facial images into the at least one class by using the second output value — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Step 2A Prong 2:
This judicial exception is not integrated into a practical application.
Claim 8 recites the additional element of:
A non-transitory, computer-readable recording medium having recorded thereon a program for executing the method of claim 1 on a computer — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to a generic computer component.
Claim 9 recites the additional elements of:
A device for classifying data, the device comprising: a memory storing at least one program; and a processor configured to execute the at least one program to — This element amounts to no more than a generic device comprising generic computer components.
Step 2B:
The claims do not contain significantly more than the judicial exception.
Claim 8 recites the additional element of:
A non-transitory, computer-readable recording medium having recorded thereon a program for executing the method of claim 1 on a computer — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to a generic computer component.
Claim 9 recites the additional elements of:
A device for classifying data, the device comprising: a memory storing at least one program; and a processor configured to execute the at least one program to — This element amounts to no more than a generic device comprising generic computer components.
As such claims 1, 8, and 9 are not patent eligible.
Dependent Claims 3, 6, 11, 14, and 16-18
Step 1:
Claims 3, 6, 11, 14, and 16-18; therefore, they are directed to one of the four categories of statutory subject matter (process/method, machine/product/apparatus, manufacture, or composition of matter).
Step 2A Prong 1:
Claims 3, 6, 11, 14, and 16-18 merely narrow the previously cited abstract idea limitations. For the reasons described above with respect to independent claim 1 and 9 this judicial exception is not meaningfully integrated into a practical application, or significantly more than the abstract idea. The claim(s) disclose similar limitations described for the independent claim(s) above and do not provide anything more than the abstract idea.
Claim 3 recites a method comprising:
wherein, in the generating of the second output value, the information indicating the label distribution of the target data is applied to the first output value by performing a multiplication operation — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 11 recites a system comprising:
the processor is further configured to execute the at least one program to apply, to the first output value, the information indicating the label distribution of the target data by performing a multiplication operation — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mental process, or a concept that can be performed in the human mind with the use of a physical aid (e.g. pen and paper), including observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)). Or a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 16 recites a method comprising:
wherein the classification model is trained by using a regularized Donsker-Varadhan (DV) representation represented by the following formula:
PNG
media_image1.png
64
538
media_image1.png
Greyscale
where P and Q denote arbitrary distributions that satisfy supp(P) supp(Q) and for every function T: Ω ➔ R some domain Ω, the function T that minimizes the regularized DV representation is the log-likelihood ratio of P and Q — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 17 recites a method comprising:
wherein the classification model is trained by plugging
PNG
media_image2.png
100
422
media_image2.png
Greyscale
and
PNG
media_image3.png
70
652
media_image3.png
Greyscale
into
PNG
media_image4.png
114
756
media_image4.png
Greyscale
— Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 18 recites a method comprising:
wherein the classification model is trained by using information indicating the label distribution of the source data
PNG
media_image5.png
84
294
media_image5.png
Greyscale
and
PNG
media_image6.png
198
650
media_image6.png
Greyscale
, where LLADER is a loss, λ, a1, ..., ac denotes hyperparameters, and C denotes a total number of classes — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 21 recites a method comprising:
wherein, in the training of the classification model, the classification model is trained by using at least one approximation formula with respect to a second equation in which a component corresponding to a label distribution of source data is disentangled in the first equation, and information indicating the label distribution of the source data — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 22 recites a method comprising:
wherein, in the training of the classification model, the classification model is trained by using at least one approximation formula with respect to a second equation in which a component corresponding to a label distribution of source data is disentangled in the first equation, and information indicating the label distribution of the source data — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 23 recites a method comprising:
wherein the at least one approximation formula comprises at least one selected from the group consisting of regularized Donsker-Varadhan (DV) representation and a Monte Carlo approximation formula — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Claim 24 recites a method comprising:
wherein the at least one approximation formula comprises at least one selected from the group consisting of regularized Donsker-Varadhan (DV) representation and a Monte Carlo approximation formula — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept (see MPEP § 2106.04(a)(2)(I)), specifically organizing information and manipulating information through mathematical correlations.
Step 2A Prong 2:
This judicial exception is not integrated into a practical application.
Claim 6 recites the additional element of:
wherein, in the training of the classification model, the classification model is trained by using information indicating regularization with respect to the label distribution of the source data — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
Claim 14 recites the additional element of:
the processor is further configured to execute the at least one program to train the classification model by using information indicating regularization with respect to the label distribution of the source data — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
Claim 25 recites the additional element of:
wherein the at least one class comprises classed based on genders or ages — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
Claim 26 recites the additional element of:
wherein the at least one class comprises classed based on genders or ages — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
Step 2B:
The claims do not contain significantly more than the judicial exception.
Claim 6 recites the additional element of:
wherein, in the training of the classification model, the classification model is trained by using information indicating regularization with respect to the label distribution of the source data — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
Claim 14 recites the additional element of:
the processor is further configured to execute the at least one program to train the classification model by using information indicating regularization with respect to the label distribution of the source data — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
Claim 25 recites the additional element of:
wherein the at least one class comprises classed based on genders or ages — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
Claim 26 recites the additional element of:
wherein the at least one class comprises classed based on genders or ages — This element amounts to no more than generally linking the use of a judicial exception to a particular technological environment or field of use (see MPEP § 2106.05(h)). This element merely limits the use of the abstract idea to training a classification model with particular information.
As such claims 3, 6, 11, 14, and 16-18 and 21-26 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, 8, 9, 11, 17-18, and 21-24 are rejected under 35 U.S.C. 103 as being unpatentable over Southern et al. (US 2022/0095974 A1, first cited in office action mailed 10/08/2025), hereinafter Southern, in view of Huber et al. (US 2015/0363643 A1), hereinafter Huber, and further in view of Richard et al. (Neural Network Classifiers Estimate Bayesian a posteriori Probabilities, published 1991. first cited in Office Action mailed 04/03/2025), hereinafter Richard, and further in view of Langseth, H., Nielsen, T.D. (“Classification using Hierarchical Naïve Bayes models”. Mach Learn 63, 135–159 (2006). doi:10.1007/s10994-006-6136-2. first cited in Office Action mailed 04/03/2026), hereinafter Langseth.
Regarding claim 1, Southern teaches a method of classifying data, the method comprising (The present invention relates to the determination or classification of the mental state of a user and associated confidence values for the determined mental state. [see Southern, Abstract]):
training a classification model for classifying input data into at least one class, such that a first output value is generated, wherein a first equation comprises an equation corresponding to a Bayes’ rule representing a probability of the input data being classified as each of the at least one class (Southern discloses that their classification model is trained prior to deployment [see Southern, para. 93] to classify input data into one or more classes [see Southern, para. 101, and 103], in party by calculating a posterior probability according to Bayes' rule in which a label distribution component is used [see Southern, para. 77 and Equation 1]);
classifying facial images into the at least one class by using the output value (Southern discloses using facial images as input data [see Southern, para. 46] for classifying data into one or more classes [see Southern, para. 93, 101, and 103] according to the second output value [see Southern, para. 87]).
However, Southern fails to teach wherein training the classification model comprises replacing a label distribution component in the first equation with a uniform prior, wherein the uniform prior pu(y=c) = 1/C, where C denotes a total number of classes, wherein the training the classification model comprises training the classification model using only a distribution of samples x from source data and a conditional distribution of the samples x given labels y; generating a second output value by applying, to the first output value, information indicating a label distribution of target data; and classifying facial images into the at least one class by using the second output value.
In the same field of endeavor Huber teaches:
wherein training the classification model comprises replacing a label distribution component in the first equation with a uniform prior, wherein the uniform prior pu(y=c) = 1/C, where C denotes a total number of classes ().
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate wherein training the classification model comprises replacing a label distribution component in the first equation with a uniform prior, wherein the uniform prior pu(y=c) = 1/C, where C denotes a total number of classes as suggested in Huber into Southern because both methods perform classification based on facial data (see Southern, Abstract and para. 92, para. 1; see Huber, para. 1 and 3). Incorporating the teaching of Huber into Southern would be a simple substitution of a prior label distribution as used in Southern for a uniform prior as suggested in Huber, such that the Bayes probability of classification can be calculated. Huber discloses that “Often the prior probability is uniform” (see Huber, para. 39), thus, the substitution would replace the prior probability with the uniform prior.
However, the combination of Southern and Huber fails to teach wherein the training the classification model comprises training the classification model using only a distribution of samples x from source data and a conditional distribution of the samples x given labels y; generating a second output value by applying, to the first output value, information indicating a label distribution of target data; and classifying facial images into the at least one class by using the second output value.
In the same field of endeavor, Richard teaches:
generating a second output value by applying, to the first output value, information indicating a label distribution of target data (However, the output y.sub.i(X) is implicitly the corresponding a priori class probability p(C.sub.i) times the class likelihood p(X | C.sub.i) divided by the unconditional input probability p(X). It is possible to vary a priori class probabilities during classification without retraining, since these probabilities occur only as multiplicative terms in producing the network outputs. As a result, class probabilities can be adjusted during use of a classifier to compensate for training data with class probabilities that are not representative of actual use or test conditions. Correct class probabilities can be used during classification by first dividing network outputs by training-data class probabilities and then multiplying by the correct class probabilities. Training-data class probabilities can be estimated as the frequency of occurrence of patterns from different classes in the training data. Correct class probabilities required for testing can be obtained from an independent set of training data that needs to contain only class labels and not input patterns. Such data are often readily available. The second output value would be the resulting value from multiplying the first output value by the correct class probabilities. [see Richard, Subsection 4.1 Compensating for Varying a priori Class Probabilities, para. 1, lines 3-18]);
classifying into the at least one class by using the second output value (Minimum-error Bayesian classifiers perform this task by calculating the Bayesian probability, p(C.sub.i |X), for each class, and assigning the input to the class with the highest Bayesian probability. The Bayesian probability p(C.sub.i | X) represents the conditional probability of class C.sub.i given the input X. Use of Bayes rule allows it to be expressed as follows: p(C.sub.i | X) = p(X | C.sub.i)p(C.sub.i)/p(x) (2.1) In this equation, p(X I C.sub.i) is the likelihood or conditional probability of producing the input if the class is C.sub.i, is the probability of p(C.sub.i) a priori class C.sub.i, and p(X) is the unconditional probability of the input. Conventional Bayesian classifiers estimate the Bayesian probability for each class by separately estimating the factors in the above equation. Since p(X) is common to all classes, it is usually omitted and instead p(X | C.sub.i,)p(C.sub.i) is used for classification. In addition, conventional classifiers estimate the likelihoods, p(X | C.sub.i), by assuming they can be well-modeled by specific parametric distributions, such as gaussian or gaussian mixture distributions. Training involves estimating the parameters of the assumed likelihood distributions and estimating the class probabilities from a priori training data. [see Richard, Subsection 2.1 Pattern Classification and Bayesian Probabilities, para. 1, lines 6—23]).
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate generating a second output value by applying, to the first output value, information indicating a label distribution of target data and classifying into the at least one class by using the second output value as suggested in Richard into the combination of Southern and Huber because both methods classify data with classification models (see Southern, Abstract; see Richard, Subsection 2.1 Pattern Classification and Bayesian Probabilities, para. 1, lines 6—23). Incorporating the teaching of Richard into the combination of Southern and Huber would make it possible to compensate for differences in pattern class probabilities between test and training data, to combine outputs of multiple classifiers for higher level decision making, to use alternative risk functions different from minimum-error risk, to implement conventional optimal rules for pattern rejection, and to compute alternative measures of network performance (see Richard, pg. 462, lines 4-9).
It would have been further obvious to one of ordinary skill in the art before the effective filing date to incorporate classifying facial images into the at least one class by using the second output value because classifying into at least one class using the second output value as suggested in Richard (see Richard, Subsection 2.1 Pattern Classification and Bayesian Probabilities, para. 1, lines 6—23) could be used to classify the facial images of Southern.
However, the combination of Southern, Huber, and Richard fails to teach wherein the training the classification model comprises training the classification model using only a distribution of samples x from source data and a conditional distribution of the samples x given labels y.
In the same field of endeavor, Langseth teaches:
wherein the training the classification model comprises training the classification model using only a distribution of samples x from source data and a conditional distribution of the samples x given labels y (Langseth discloses that Bayesian classifiers learn from a set of labeled training samples denoted by Dn, such that P(C=c|A=a, Dn) is the a posteriori conditional probability that C=c given A=a after observing Dn [see Langseth, Section 2, pg. 137-138 para. 1]).
It would have been obvious to one of ordinary skill, in the art at the time before the effective filing date of the invention to incorporate wherein the training the classification model comprises training the classification model using only a distribution of samples x from source data and a conditional distribution of the samples x given labels y as suggested in Langseth into the combination of Southern, Huber, and Richard because both methods use Bayesian classifiers [see Southern, para. 22; see Langseth, Abstract]. Incorporating the teaching of Langseth into Southern would aid in extending the utility of training the classification model to classify input data into at least one class [see Southern, para. 89-93 and 101].
Regarding claim 8, claim 8 contains substantially similar limitations to those found in claim 1. Therefore it is rejected for the same reason as claim 1 above. Additionally, it would have been obvious to one of ordinary skill in the art to implement the method of Southern on a computer readable medium to execute on a computer to further teach a non-transitory, computer-readable recording medium having recorded thereon a program for executing the method of claim 1 on a computer.
Regarding claim 9, claim 9 contains substantially similar limitations to those found in claim 1. Therefore it is rejected for the same reason as claim 1 above. Additionally, it would have been obvious to one of ordinary skill in the art to implement the method of Southern on a general purpose computer with a processor and memory storing a program to implement the method to teach A device for classifying data, the device comprising: a memory storing at least one program; and a processor configured to execute the at least one program to.
Regarding claim 3, the combination of Southern, Huber, Richard, and Langseth as applied in claim 1 above teaches all the limitations of claim 1 and further teaches:
wherein, in the generating of the second output value, the information indicating the label distribution of the target data is applied to the first output value by performing a multiplication operation (However, the output y.sub.i(X) is implicitly the corresponding a priori class probability p(C.sub.i) times the class likelihood p(X | C.sub.i) divided by the unconditional input probability p(X). It is possible to vary a priori class probabilities during classification without retraining, since these probabilities occur only as multiplicative terms in producing the network outputs. As a result, class probabilities can be adjusted during use of a classifier to compensate for training data with class probabilities that are not representative of actual use or test conditions. Correct class probabilities can be used during classification by first dividing network outputs by training-data class probabilities and then multiplying by the correct class probabilities. Training-data class probabilities can be estimated as the frequency of occurrence of patterns from different classes in the training data. Correct class probabilities required for testing can be obtained from an independent set of training data that needs to contain only class labels and not input patterns. Such data are often readily available. [see Richard, Subsection 4.1 Compensating for Varying a priori Class Probabilities, para. 1, lines 3-18]).
Regarding claim 11, claim 11 contains substantially similar limitations to those found in claim 3 above. Consequently, claim 11 is rejected for the same reasons.
Regarding claim 17, the combination of Southern, Huber, Richard, and Langseth as applied in claim 1 above teaches all the limitations of claim 1 and further teaches:
wherein the classification model is trained by plugging
PNG
media_image2.png
100
422
media_image2.png
Greyscale
and
PNG
media_image3.png
70
652
media_image3.png
Greyscale
into
PNG
media_image4.png
114
756
media_image4.png
Greyscale
(Southern discloses that the machine learning model uses a Monte Carlo dropout method during training to approximate posterior distributions [see Southern, para. 78]).
Regarding claim 18, the combination of Southern, Huber, Richard, and Langseth as applied in claim 1 above teaches all the limitations of claim 1 and further teaches:
wherein the classification model is trained by using information indicating the label distribution of the source data
PNG
media_image5.png
84
294
media_image5.png
Greyscale
and
PNG
media_image6.png
198
650
media_image6.png
Greyscale
, where LLADER is a loss, λ, a1, ..., ac denotes hyperparameters, and C denotes a total number of classes (Southern discloses a posterior probability according to Bayes' rule in which a label distribution component is used [see Southern, para. 77-78 and Equation 1]).
Regarding claim 21, the combination of Southern, Huber, Richard, and Langseth as applied in claim 1 above teaches all the limitations of claim 1 and further teaches:
Wherein, in the training of the classification model, the classification model is trained by using at least one approximation formula with respect to a second equation in which a component corresponding to a label distribution of source data is disentangled in the first equation, and information indicating the label distribution of the source data (Southern discloses approximation the equation used for calculating the posterior probability which includes the label distribution component [see Southern, para. 77 and Equation 1], using a Monte Carlo approximation [see Southern, para. 78]).
Regarding claim 22, claim 22 contains substantially similar limitations to those found in claim 21 above. Consequently, claim 22 is rejected for the same reasons.
Regarding claim 23, the combination of Southern, Huber, Richard, and Langseth as applied in claim 21 teaches all the limitations of claim 21 and further teaches:
wherein the at least one approximation formula comprises at least selected from the group consisting of a regularized Donsker-Varadhan (DV) representation and a Monte Carlo approximation formula (Southern discloses approximation the equation used for calculating the posterior probability which includes the label distribution component [see Southern, para. 77 and Equation 1], using a Monte Carlo approximation [see Southern, para. 78]).
Regarding claim 24, claim 24 contains substantially similar limitations to those found in claim 23 above. Consequently, claim 24 is rejected for the same reasons.
Claims 6, 14, and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Southern et al. (US 2022/0095974 A1, first cited in office action mailed 10/08/2025), hereinafter Southern, in view of Huber et al. (US 2015/0363643 A1), hereinafter Huber, and further in view of Richard et al. (Neural Network Classifiers Estimate Bayesian a posteriori Probabilities, published 1991. first cited in Office Action mailed 04/03/2025), hereinafter Richard, and further in view of Langseth, H., Nielsen, T.D. (“Classification using Hierarchical Naïve Bayes models”. Mach Learn 63, 135–159 (2006). doi:10.1007/s10994-006-6136-2. first cited in Office Action mailed 04/03/2026), hereinafter Langseth, as applied in claim 1 above, and further in view of Mroueh et al. (US 11,630,989 B2, first cited in office action filed 04/03/2025), hereinafter Mroueh.
Regarding claim 6, the combination of Southern, Huber, Richard, and Langseth as applied in claim 1 above teaches all the limitations of claim 1.
However, the combination of Southern, Huber, Richard, and Langseth fails to teach wherein, in the training of the classification model, the classification model is trained by using information indicating regularization with respect to the label distribution of the source data.
In the same field of endeavor, Mroueh teaches:
wherein, in the training of the classification model, the classification model is trained by using information indicating regularization with respect to the label distribution of the source data (What is provided herein is a new estimator of MI that can be used in direct MI maximization or as a regularizer, thanks to its unbiased gradients. Our starting point is the DV lower bound of the KL divergence that is represented equivalently via a joint optimization that is referred to herein as η-DV on a witness function ƒ and an auxiliary variable η. [see Mroueh, Col. 3, lines 62-67]; The η-DV bound restricted to an RKHS, amounts to the following regularized convex minimization: P=(min.sub.w,ηL(w,η)+Ω(w)) [see Mroueh, Col. 7, lines 54-56]; Mutual Information (MI) is information indicating similar label distribution between the source data and the target data. Thus, using regularized DV representation to approximate the MI, would be using a regularized DV representation with respect to information indicating label distribution of training data).
It would have been obvious to one of ordinary skill, in the art at the time before the effective filing date of the invention to incorporate wherein, in the training of the classification model, the classification model is trained by using information indicating regularization with respect to the label distribution of the source data as suggested in Mroueh into the combination of Southern, Huber, Richard, and Langseth because both systems are computing devices that perform machine learning (see Southern, Abstract; see Mroueh, Col. 4, lines 16-36). Incorporating the teaching of Mroueh into the combination of Southern, Huber, Richard, and Langseth would improve computational efficiency and accuracy in computing devices that perform machine learning and artificial intelligence (see Col. 3, lines 45-47).
Regarding claim 14, claim 14 contains substantially similar limitations to those found in claim 6 above. Consequently, claim 14 is rejected for the same reasons.
Regarding claim 16, the combination of Southern, Huber, Richard, and Langseth as applied in claim 1 above teaches all the limitations of claim 1.
However, the combination of Southern, Huber, Richard, and Langseth fails to teach wherein the classification model is trained by using a regularized Donsker-Varadhan (DV) representation represented by the following formula:
PNG
media_image1.png
64
538
media_image1.png
Greyscale
where P and Q denote arbitrary distributions that satisfy supp(P) supp(Q) and for every function T: Ω ➔ R some domain Ω, the function T that minimizes the regularized DV representation is the log-likelihood ratio of P and Q.
In the same field of endeavor, Mroueh teaches:
wherein the classification model is trained by using a regularized Donsker-Varadhan (DV) representation represented by the following formula:
PNG
media_image1.png
64
538
media_image1.png
Greyscale
where P and Q denote arbitrary distributions that satisfy supp(P) supp(Q) and for every function T: Ω ➔ R some domain Ω, the function T that minimizes the regularized DV representation is the log-likelihood ratio of P and Q (Mroueh discloses that the regularized neural network using DV-MINE converges after 140k steps [see Mroueh, Col. 12, lines 54-59], thus, it was trained using a regularized DV representation).
It would have been obvious to one of ordinary skill, in the art at the time before the effective filing date of the invention to incorporate wherein the classification model is trained by using a regularized Donsker-Varadhan (DV) representation represented by the following formula:
PNG
media_image1.png
64
538
media_image1.png
Greyscale
where P and Q denote arbitrary distributions that satisfy supp(P) supp(Q) and for every function T: Ω ➔ R some domain Ω, the function T that minimizes the regularized DV representation is the log-likelihood ratio of P and Q as suggested in Mroueh into the combination of Southern, Huber, Richard, and Langseth because both systems are computing devices that perform machine learning (see Southern, Abstract; see Mroueh, Col. 4, lines 16-36). Incorporating the teaching of Mroueh into the combination of Southern, Huber, Richard, and Langseth would improve computational efficiency and accuracy in computing devices that perform machine learning and artificial intelligence (see Col. 3, lines 45-47).
Response to Arguments
Applicant’s arguments, filed 06/08/2026, traversing the rejection of claims 1, 3, 6, 8-9, 11, 14, and 16-26 under 35 U.S.C. 101 have been fully considered and are not persuasive.
Applicant’s arguments, filed 06/08/2026, traversing the rejection of claims 1, 3, 6, 8-9, 11, 14, and 16-26 under 35 U.S.C. 103 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.
Claims 25-26 are rejected under 35 U.S.C. 103 as being unpatentable over Southern et al. (US 2022/0095974 A1, first cited in office action mailed 10/08/2025), hereinafter Southern, in view of Huber et al. (US 2015/0363643 A1), hereinafter Huber, and further in view of Richard et al. (Neural Network Classifiers Estimate Bayesian a posteriori Probabilities, published 1991. first cited in Office Action mailed 04/03/2025), hereinafter Richard, and further in view of Langseth, H., Nielsen, T.D. (“Classification using Hierarchical Naïve Bayes models”. Mach Learn 63, 135–159 (2006). doi:10.1007/s10994-006-6136-2. first cited in Office Action mailed 04/03/2026), hereinafter Langseth, as applied in claim 1 above, and further in view of Hwang et al. (US 2007/0104362 A1), hereinafter Hwang.
Regarding claim 25, the combination of Southern, Huber, Richard, and Langseth as applied in claim 1 above teaches all the limitations of claim 1.
However, the combination of Southern, Huber, Richard, and Langseth fails to teach wherein the at least one class comprises classes based on genders or ages.
In the same field of endeavor, Hwang teaches:
wherein the at least one class comprises classes based on genders or ages (An embodiment of the present invention relates to a face recognition method, medium, and system using gender information, and more particularly, to a method, medium, and system determining the gender of a query facial image and recognizing a face using the determined gender. [see Hwang, para. 3]).
It would have been obvious to one of ordinary skill, in the art at the time before the effective filing date of the invention to incorporate wherein the at least one class comprises classes based on genders or ages as suggested in Hwang into the combination of Southern, Huber, Richard, and Langseth because both methods perform classification of facial data (see Southern, Abstract and para. 92; see Hwang, Abstract) Incorporating the teaching of Hwang into the combination of Southern, Huber, Richard, and Langseth would aid in overcoming the problems with recognition performance and reliability for female faces (see Hwang, para. 6-7 and 10).
Regarding claim 26, claim 26 contains substantially similar limitations to those found in claim 25 above. Consequently, claim 26 is rejected for the same reasons.
Response to Arguments
Applicant's arguments, filed 06/08/2026, traversing the rejection of claims 1, 3, 6, 8-9, 11, 14, and 16-18 under 35 U.S.C 103 have been fully considered but they are not persuasive. Applicant argues that the claims are not directed to an abstract idea because no human could practically perform the claimed process mentally or with the aid of pen and paper, and that the claims are integrated into a practical application even if they did recite a judicial exception, specifically, that the claims solve the technical problem of degradation of classification model accuracy when the label distribution of training data differs from the label distribution of inference data through replacing the label distribution component with a uniform prior, Examiner respectfully disagrees.
With respect to Applicant’s argument that the claims do not recite a judicial exception: as identified above in the 35 U.S.C. 101 section above and in the previous office action mailed 04/03/2026, the independent claims recite both mental processes and mathematical concepts. Applicant argues that the claims do not recite mental processes and cites to the limitation reciting “wherein a first equation comprises an equation corresponding to a Bayes’ rule representing a probability of the input data being classified as each of the at least one class, and wherein training the classification model comprises replacing a label distribution component in the first equation with a uniform prior, wherein the uniform prior pu(y=c) = 1/C, where C denotes a total number of classes”. However, it was never stated that this limitation recites a mental process in the previously mailed office action, or in the 35 U.S.C 101 section above. Generally, the steps reciting classifying data recite mental processes as the process of classification are reasonably performable in the mind or with the aid of pen and paper as the steps are directed to observation, evaluation, judgement or opinion (see MPEP § 2106.04(a)(2)(III)), and are also mathematical concepts as the classifications are direct mathematical probabilities and calculations. While the steps of training the classification model recite mathematical concepts as the claim recites direct and distinct mathematical formulas and operations for how the model mathematically calculates outputs, trains the models with mathematical equations and formulas, and recites direct mathematical components of the model. Thus, the argument that the mathematical operations are not simple operations that a human mind could perform is not pertinent to the eligibility of the mathematical concepts recited in the claims, as mathematical concepts and specifically mathematical formulas are examples of judicial exceptions and are not eligible subject matter regardless of whether the mathematical formulas and calculations complexity or ability to be reasonably performable in the human mind.
With respect to Applicant’s argument that the claims are integrated into a practical application, specifically, the improvement of classification models regardless of whether the distribution of the source data and target data are different from each other. Applicant states, consistent with the specification, that the technical problem is found by replacing the label distribution component in Bayes’ rule with a uniform prior (pu(y=c) = 1/C) and then applying the target label distribution at inference time. However, a judicial exception cannot be integrated into a practical application by an improvement founded solely by the abstract idea itself: “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements. See the discussion of Diamond v. Diehr, 450 U.S. 175, 187 and 191-92, 209 USPQ 1, 10 (1981))“ (see MPEP § 2106.05(a)). Thus, the purported improvement cannot integrate the judicial exception as the improvement is founded by the judicial exception of the mathematical concepts recited.
For at least the aforementioned reasons, the rejection of claims 1, 3, 6, 8-9, 11, 14, and 16-26 under 35 U.S.C. 101 are respectfully maintained.
Applicant’s arguments, filed 06/08/2026, traversing the rejection of claims 1, 3, 6, 8-9, 11, 14, and 16-18 under 35 U.S.C. 103, with respect to amended independent claims 1 and 8-9 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.
For at least the aforementioned reasons, the rejection of claims 1, 3, 6, 8-9, 11, 14, and 16-26 under 35 U.S.C. 103 are respectfully maintained.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Wang et al. (US 2018/0114056 A1) discloses performing a classification with a naïve Bayesian classifier where the uniform prior is used in place of the prior probability of the label distribution.
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 JAKE BREEN whose telephone number is (571)272-0456. The examiner can normally be reached Monday - Friday, 7:00 AM - 3:00 PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jennifer Welch can be reached at (571) 272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/J.T.B./Examiner, Art Unit 2143
/JENNIFER N WELCH/Supervisory Patent Examiner, Art Unit 2143
Prosecution Insights