DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
This Office Action is in response to applicant’s communication filed 9 April 2026, in response to the Office Action mailed 12 January 2026. The applicant’s remarks and any amendments to the claims or specification have been considered, with the results that follow.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 9 April 2026 has been entered.
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.
Claim(s) 1-29 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The claim(s) recite(s) mental processes and/or mathematical concepts. This judicial exception is not integrated into a practical application and does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception, as described below.
Step 1 for all claims:
Under the first part of the analysis, claims 1-19 recite a method and claims 20-29 recite a device. Accordingly, these claims fall within the four statutory categories of invention and the analysis proceeds to Step 2A, prongs 1 and 2, and Step 2B, as described below.
As per claim 1:
Under step 2A, prong 1, the claim recites an abstract idea including the following mental process and/or mathematical concept elements:
wherein the training comprises: generating one or more target features using a target data set – a data scientist generates one or more features from a target data set.
generating one or more masks using the one or more target features – the data scientist generates masks for the target features.
generating a set of task features by applying the one or more masks to the one or more target features – the data scientist applies the mask(s) to the features to generate a set of task features.
and generating a loss using the one or more target features and the set of task features – the loss is a mathematical function (see, e.g., equation (1) of the specification as filed). Alternatively/additionally – the data scientist selects a loss function from the one or more target features and task features.
If a claim, under the broadest reasonable interpretation covers a mathematical relationship between variables or numbers, a numerical formula or equation, or a mathematical calculation, it will be considered as falling within the “mathematical concepts” grouping of abstract ideas. If a claim, under the broadest reasonable interpretation covers concepts that can be performed in the human mind, or by a human using a pen and paper, including observation, evaluation, judgment, or opinion, it will be considered as falling within the “mental processes” grouping of abstract ideas. Additionally, performing mathematical calculations using a formula that could be practically performed in the human mind may be considered to fall within both the mathematical concepts grouping and the mental process grouping. See MPEP § 2106.04(a)(2).
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
Under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
A method of cross-domain adaptive machine learning, comprising – this amounts to generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type. See MPEP § 2106.05(h) and Electric Power, 830 F.3d at 1354, 119 USPQ2d at 1742 (limiting application of abstract idea to power grid data).
training a target domain feature extraction model that comprises a source domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
the source domain feature extraction model having been trained on a source data set – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and the source domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and a mask generation model trained on only the target data set – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and deploying the trained target domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
A method of cross-domain adaptive machine learning, comprising – this amounts to generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type. See MPEP § 2106.05(h) and Electric Power, 830 F.3d at 1354, 119 USPQ2d at 1742 (limiting application of abstract idea to power grid data).
training a target domain feature extraction model that comprises a source domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
the source domain feature extraction model having been trained on a source data set – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and the source domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and a mask generation model trained on only the target data set – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and deploying the trained target domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 2:
The claim recites the following additional mathematical concept elements:
wherein the source domain feature extraction model is trained using a self-supervised loss function – this is describing a mathematical function/formula (the self-supervised loss function is a mathematical formula – see, e.g., Equation (1) in the specification as filed).
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 3:
The claim recites the following additional mathematical concept elements:
wherein the self-supervised loss function comprises a contrastive loss function – this is describing a mathematical function/formula (the contrastive, self-supervised loss function is a mathematical formula – see, e.g., Equation (1) in the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 4:
The claim recites the following additional mathematical concept elements:
augmenting the source data set by performing one or more transformations on one or more samples of the source data set – this is describing a mathematical function/formula (the transformations are a mathematical function – see, e.g., para. [0036] of the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 5:
The claim recites the following additional mental process elements:
generating a set of positive features based on the target data set – the data scientist generates a set of positive features from the target data set.
and generating a set of negative features based on the target data set – the data scientist generates a set of negative features from the target data set.
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
Under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
wherein training the mask generation model comprises – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
wherein training the mask generation model comprises – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
and the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 6:
The claim recites the following additional mental process elements:
generating a set of masks – the data scientist generates a set of masks.
and generating a set of binary masks based on the set of masks – the data scientist generates a set of binary masks from the masks.
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
Under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
using the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
using the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 7:
Under step 2A, prong 1, the claim recites an abstract idea including the following mathematical concept elements:
wherein generating the set of binary masks based on the set of masks comprises: adding logistic noise to the set of masks – this is describing a mathematical function/calculation (producing random noise from a logistic distribution).
applying a nonlinear activation function to the set of masks – this is describing a mathematical function/formula (the nonlinear activation function is a mathematical formula – see, e.g., Equation (2) in the specification as filed).
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 8:
The claim recites the following additional mathematical concept elements:
wherein the nonlinear activation function comprises a sigmoid function – this is describing a mathematical function/formula (see, e.g., Equation (2) in the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 9:
Under step 2A, prong 1, the claim recites an abstract idea including the following mathematical concept elements:
using a loss function comprising a cross-entropy loss component – this is describing a mathematical function/formula (see, e.g., Equation (1) in the specification as filed).
If a claim, under the broadest reasonable interpretation covers a mathematical relationship between variables or numbers, a numerical formula or equation, or a mathematical calculation, it will be considered as falling within the “mathematical concepts” grouping of abstract ideas. Additionally, performing mathematical calculations using a formula that could be practically performed in the human mind may be considered to fall within both the mathematical concepts grouping and the mental process grouping. See MPEP § 2106.04(a)(2).
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
Under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
wherein the mask generation model is trained – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
wherein the mask generation model is trained – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 10:
The claim recites the following additional mathematical concept elements:
wherein the loss function further comprises a maximum entropy loss component based on the set of negative features – this is describing a mathematical function/formula (see, e.g., Equation (4) in the specification as filed).
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 11:
The claim recites the following additional mathematical concept elements:
wherein the loss function further comprises a divergence loss component based on the set of positive features and the set of negative features – this is describing a mathematical function/formula (see, e.g., Equation (5) in the specification as filed).
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 12:
The claim recites the following additional mathematical concept elements:
wherein the loss function further comprises: a first weighting parameter for the cross-entropy loss component; a second weighting parameter for the maximum entropy loss component; and a third weighting parameter for the divergence loss component – this is describing parameters of a mathematical function/formula (see, e.g., Equation (6) in the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 13:
Under step 2A, prong 1, the claim recites an abstract idea including the following mathematical concept elements:
wherein generating the loss comprises using a loss function comprising a regularization loss component – this is describing a mathematical function/formula (see, e.g., Equation (7) in the specification as filed).
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 14:
The claim recites the following additional mathematical concept elements:
wherein the regularization loss component comprises a Euclidean distance function – this is describing a mathematical function/formula (see, e.g., Equation (7) in the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 15:
The claim recites the following additional mathematical concept elements:
wherein the loss function further comprises a cross-entropy loss component – this is describing a mathematical function/formula (see, e.g., Equation (1) in the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 16:
The claim recites the following additional mathematical concept elements:
wherein for a given sample, the cross-entropy loss component is configured to generate a cross-entropy loss value based on a positive feature of the set of task features – this is describing a mathematical function/formula (see, e.g., Equation (1) in the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
Under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
generated by the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
based on the given sample and a classification output generated by a linear classification model based on the given sample – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
generated by the mask generation model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
based on the given sample and a classification output generated by a linear classification model based on the given sample – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 17:
The claim recites the following additional mathematical concept elements:
wherein the loss function further comprises a weighting parameter for the regularization loss component – this is describing parameters of a mathematical function/formula (see, e.g., Equation (6) in the specification as filed)
Accordingly, at step 2A, prong one, the claim is directed to an abstract idea.
The claim does not include any additional elements, under step 2A prong two, or step 2B, except those listed above in prior claim(s). Accordingly, at step 2A, prong two, the claim as a whole does not integrate the judicial exception into a practical application. See MPEP § 2106.04(d). Furthermore, at step 2B, the claim elements both individually and in combination do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 18:
Under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
wherein the target domain feature extraction model comprises a neural network model – this amounts to generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type. See MPEP § 2106.05(h) and Electric Power, 830 F.3d at 1354, 119 USPQ2d at 1742 (limiting application of abstract idea to power grid data).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
wherein the target domain feature extraction model comprises a neural network model – this amounts to generally linking the use of the judicial exception to a particular technological environment or field of use by limiting it to a particular data source or type. See MPEP § 2106.05(h) and Electric Power, 830 F.3d at 1354, 119 USPQ2d at 1742 (limiting application of abstract idea to power grid data).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 19:
Under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
further comprising generating an inference using the trained target domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
further comprising generating an inference using the trained target domain feature extraction model – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 20:
See the rejection of claim 1 above wherein, under step 2A, prong two, the judicial exception is not integrated into a practical application. In particular, the claim recites the additional elements of:
A processing system comprising:
memory comprising computer-executable instructions; and
one or more processors configured to execute the computer-executable instructions and cause the processing system to perform [the method] – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2A, prong two, these additional elements do not integrate the abstract idea into a practical application for the claim as a whole, because it does not impose any meaningful limits on practicing the abstract idea. See MPEP § 2106.04(d).
Under step 2B, the claims do not include additional elements that are sufficient to amount to significantly more that the judicial exception. As discussed above, the claim recites the additional elements of:
A processing system comprising:
memory comprising computer-executable instructions; and
one or more processors configured to execute the computer-executable instructions and cause the processing system to perform [the method] – this amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Accordingly, at step 2B, these additional elements, both individually and in combination, do not amount to significantly more than the judicial exception. See MPEP § 2106.05.
Therefore, the claim is not eligible subject matter under 35 U.S.C. 101.
As per claim 21, see the rejection of claim 2, above.
As per claim 22, see the rejection of claim 3, above.
As per claim 23, see the rejection of claim 4, above.
As per claim 25, see the rejection of claim 7, above.
As per claim 26, see the rejections of claims 9-11, above.
As per claim 27, see the rejection of claim 12, above.
As per claim 28, see the rejections of claims 13-14, above.
As per claim 29, see the rejection of claim 19, above.
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.
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.
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.
Claim(s) 1, 5, 9-12, 18-20, 26, 27, and 29 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jin et al. (Feature Alignment and Restoration for Domain Generalization and Adaptation, June 2020, pgs. 1-15 – cited in an IDS) in view of Harikumar (US 2022/0156992).
As per claim 1, Jin teaches a method of cross-domain adaptive machine learning, comprising: training a target domain feature extraction model that comprises a source domain feature extraction model, the source domain feature extraction model having been trained on a source data set [a Feature Alignment and Restoration (FAR) framework is used to tune an unsupervised domain adaptation (UDA) model (pg. 1, abstract; etc.) including a feature extraction model for a target domain(s) (tuning a target domain feature extraction model) trained from a backbone network including feature extraction (pg. 3, fig. 2; etc.); where the backbone network is the source domain feature extraction model trained on the source data set (see, e.g., pgs. 3-4, sections 3-3.1; where Fs1 is a feature map extracted from data of a first source domain; etc.)], wherein the training comprises:
generating one or more target features using a target data set and the source domain feature extraction model [a Feature Alignment and Restoration (FAR) framework is used to tune an unsupervised domain adaptation (UDA) model (pg. 1, abstract; etc.) including a feature extraction model for a target domain(s) (tuning a target domain feature extraction model) trained from a backbone network including feature extraction (pg. 3, fig. 2; etc.) which extracts features from the target data (pg. 3, section 3; etc.)];
generating one or more masks using the one or more target features and a mask generation model [In the FA phase, we first leverage spatial and channel attention to adaptively select features from the extracted feature map F of the backbone network and enforce across domain alignment constraints on them. Then, the FR step distills task-relevant (discriminative) features from the residual to compensate for the aligned features. Moreover, in the FR phase, we design a dual ranking entropy loss constraint to promote the distillation of the task-relevant features from the residual, trained on unlabeled images from a target domain (pgs. 3-4, sections 3-3.1; pg. 5, section 3.2; etc.); where the spatial/channel attention selecting features is the mask generation model];
generating a set of task features by applying the one or more masks to the one or more target features [In the FA phase, we first leverage spatial and channel attention to adaptively select features from the extracted feature map F of the backbone network (pgs. 3-4, sections 3-3.1; pg. 5, section 3.2; etc.); where the selected features are the set of task features generated by applying the spatial/channel attention (masks)]; and
generating a loss using the one or more target features and the set of task features [we design a dual ranking entropy loss constraint to promote the distillation of the task-relevant features from the residual, trained on unlabeled images from a target domain (pgs. 3-4, sections 3-3.1; pg. 5, section 3.2; etc.)]; and
deploying the trained target domain feature extraction model [the model is deployed and tested on a target domain (pg. 3, section 3; pg. 6, table 1; etc.)].
While Jin teaches mask generation (see above) and the target models are trained using a target domain set of unlabeled images (target data set) (pg. 3, section 3; pg. 5, section 3.2; etc.), it has not been relied upon for teaching a mask generation model trained on only the target data set.
Harikumar teaches a mask generation model trained on only the target data set [a mask refinement network is trained to predict fine-attention masks for specific classes present in target images of a training dataset or other target dataset(s) (para. 0025, 0030, etc.); for the target domain of Jin (see also below)].
Jin and Harikumar are analogous art, as they are within the same field of endeavor, namely optimizing features for classification of images.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to train a mask refinement network on a target image dataset for generating fine-attention masks, as taught Harikumar, for generating attention/masks for feature selection in the system/method taught by Jin.
Harikumar provides motivation as [training the model to produce fine-attention masks for specific images produces higher resolution masks and finer segmentation details on the target images (para. 0025, etc.)]. Also, it has been held that the omission of an element and its function (in this case, training on only the target data set, instead of source and target domains) in a combination where the remaining elements perform the same function as before involves only routine skill in the art. In re Karlson, 136 USPQ 184.
As per claim 5, Jin/Harikumar teaches wherein training the mask generation model comprises: generating a set of positive features based on the target data set and the mask generation model; and generating a set of negative features based on the target data set and the mask generation model [the FA and FR phases are used to disentangle the residual, R, into task-relevant (positive) features R+ and task-irrelevant (negative) features R- while training on the target domain image data set (Jin: pg. 4, section 3.1; see “Feature Alignment (FA) Phase” and “Feature Restoration (FR) Phase”, etc.)].
As per claim 9, Jin/Harikumar teaches wherein the mask generation model is trained using a loss function comprising a cross-entropy loss component based on the set of positive features [the FA/FR model loss function includes a weighted summation of alignment loss Lalign, dual ranking entropy loss LDRE, and basic classification loss (i.e., cross-entropy loss) Lcls (Jin: pg. 5, section 3.2; pg. 11, supplementary 2; etc.)].
As per claim 10, Jin/Harikumar teaches wherein the loss function further comprises a maximum entropy loss component based on the set of negative features [dual entropy loss functions are calculated from the enhanced feature vector f+, the contaminated feature vector f-, and the reference feature vector f; where the loss for the contaminated feature vector is L-DRE (maximum entropy loss component) (Jin: pg. 4, section 3.1 and equation (2), etc.)].
As per claim 11, Jin/Harikumar teaches wherein the loss function further comprises a divergence loss component based on the set of positive features and the set of negative features [dual entropy loss functions are calculated from the enhanced feature vector f+, the contaminated feature vector f-, and the reference feature vector f; which includes a divergence between the feature vectors for enhanced and contaminated features (Jin: pg. 4, section 3.1 and equation (2); pg. 7, section 4.4; etc.)].
As per claim 12, Jin/Harikumar teaches wherein the loss function further comprises: a first weighting parameter for the cross-entropy loss component; a second weighting parameter for the maximum entropy loss component; and a third weighting parameter for the divergence loss component [the FA/FR model loss function includes a weighted summation of alignment loss Lalign, dual ranking entropy loss LDRE, and basic classification loss (i.e., cross-entropy loss) Lcls (Jin: pg. 5, section 3.2; pg. 11, supplementary 2; etc.)].
As per claim 18, Jin/Harikumar teaches wherein the target domain feature extraction model comprises a neural network model [feature extraction can utilize a DNN (Jin: pg. 3, fig. 2; etc.)].
As per claim 19, Jin/Harikumar teaches generating an inference using the trained target domain feature extraction model [the trained model produces inferences and is compared to multiple baseline models (Jin: pgs. 5-6, sections 4-4.2 and Table 1; pg. 8, Tables 4-5; etc.)].
As per claim 20, see the rejection of claim 1, above, wherein Jin/Harikumar also teaches a processing system, comprising: a memory comprising computer-executable instructions; and one or more processors configured to execute the computer-executable instructions and cause the processing system to perform an operation comprising [the method] [the trained model produces inferences and is compared to multiple baseline models (Jin: pgs. 5-6, sections 4-4.2 and Table 1; pg. 8, Tables 4-5; etc.); which requires implementing the model using instructions stored in a memory and executed by a processor].
As per claim 26, see the rejections of claims 9-11, above.
As per claim 27, see the rejection of claim 12, above.
As per claim 29, see the rejection of claim 19, above.
Claim(s) 2-4 and 21-23 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jin et al. (Feature Alignment and Restoration for Domain Generalization and Adaptation, June 2020, pgs. 1-15 – cited in an IDS) in view of Chen et al. (A Simple Framework for Contrastive Learning of Visual Representations, July 2020, pgs. 1-20 – cited in an IDS).
As per claim 2, Jin/Harikumar teaches the method of claim 1, as described above.
While Jin/Harikumar also teaches wherein the source domain feature extraction model is trained with a loss function (see above), it has not been relied upon for teaching wherein the source domain feature extraction model is trained using a self-supervised loss function.
Chen wherein the source domain feature extraction model is trained using a self-supervised loss function [a linear classifier is trained for image classification on self-supervised representations using a contrastive loss function (pg. 1, abstract; pg. 2, section 2.1; etc.), which includes the model learning generalizable features (i.e., feature extraction) (pg. 5, section 3.1; etc.); where the loss function is thus a self-supervised loss function; for the source domain feature extraction model of Jin, above].
Jin/Harikumar and Chen are analogous art, as they are within the same field of endeavor, namely building/training image classifier models.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to utilize the self-supervised contrastive loss function for training the feature extraction and classifier model(s), as taught by Chen, for the loss function used to train the source domain feature extraction model in the system taught by Jin/Harikumar.
Chen provides motivation as [the classifier trained using self-supervised contrastive loss outperforms prior models, including supervised models, without requiring labeled data (pg. 1, abstract and section 1; etc.)].
As per claim 3, Jin/Harikumar/Chen teaches wherein the self-supervised loss function comprises a contrastive loss function [a backbone network, including a source domain feature extraction model, is trained on the source data set using a multi-part loss function (see, e.g., pgs. 3-4, sections 3-3.1; where Fs1 is a feature map extracted from data of a first source domain; pg. 5, section 3.2; etc.); where the classifier is trained for image classification on self-supervised representations using a contrastive loss function (Chen: pg. 1, abstract; pg. 2, section 2.1; etc.)].
As per claim 4, Jin/Harikumar/Chen teaches augmenting the source data set by performing one or more transformations on one or more samples of the source data set [the training data set is augmented using various transformations, such as random cropping, resizing, random color distortion, random Gaussian blur, etc. (Chen: pg. 2, section 2.1; etc.)].
As per claim 21, see the rejection of claim 2, above.
As per claim 22, see the rejection of claim 3, above.
As per claim 23, see the rejection of claim 4, above.
Claim(s) 6 and 24 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jin et al. (Feature Alignment and Restoration for Domain Generalization and Adaptation, June 2020, pgs. 1-15 – cited in an IDS), in view of Harikumar (US 2022/0156992), and further in view of Li et al. (US 2021/0046861).
As per claim 6, Jin/Harikumar teaches generating a set of masks using the mask generation model [To enable the content-adaptive disentanglement, we use a spatial and channel attention module (similar to the attention in the FA, see Supplementary) as the gate (generated masks) to obtain task-relevant feature R+ and the remaining task-irrelevant feature R- (Jin: pg. 4, section 3.1; see “Feature Alignment (FA) Phase” and “Feature Restoration (FR) Phase”, etc.)];
While Jin/Harikumar teaches a mask generation model generating a set of fine-attention masks (see above), it has not been relied upon for teaching generating a set of binary masks based on the set of masks.
Li teaches generating a set of binary masks based on the set of masks [the DNN may be trained to output a binary confidence mask indicating confidence that a pixel belongs to a class, using the activation function output, which may include ReLU or sigmoid functions (paras. 0025-26); based on the activation function output from the attention gates (masks) of Jin, above].
Jin/Harikumar and Li are analogous art, as they are within the same field of endeavor, namely building/training image classifiers.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to generate a set of binary masks, for the classifier, from the activation function outputs in the trained DNN, as taught by Li, from the activation function outputs generated from the fine-attention mask generation in the DNN classifier taught by Jin/Harikumar.
Li provides motivation as [the binary masks produced by the DNN provide more accurate identification and localization for items/segments in the image(s) (para. 0005, etc.)].
As per claim 24, see the rejections of claims 5-6, above.
Claim(s) 7, 8, and 25 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jin, Harikumar, and Li as applied to claims 6 and 24, above, and further in view of Van den Oord et al. (US 2018/0365554).
As per claim 7, Jin/Harikumar/Li teaches wherein generating the set of binary masks based on the set of masks comprises: applying a nonlinear activation function to the set of masks [the channel attention gates are followed by activation function layers, including a ReLU or sigmoid (nonlinear) activation functions (Jin: pg. 10, supplementary 1; etc.); where the DNN may be trained to output a binary confidence mask indicating confidence that a pixel belongs to a class, using the activation function output, which may include ReLU or sigmoid functions (Li: paras. 0025-26)].
While Jin/Harikumar/Li teaches that the generated masks may include noise (see, e.g., Li: para. 0026), it has not been relied upon for teaching adding logistic noise to the set of masks.
Van den Oord teaches adding logistic noise to the set of masks [the feedforward network can add noise to generated outputs using a random noise vector that includes a noise value for each sample in the output, where the noise values in the noise vector come from a logistic distribution (i.e., logistic noise) (paras. 0065-66, etc.); for the set of masks of Jin/Li, above].
Jin/Harikumar/Li and Van den Oord are analogous art, as they are within the same field of endeavor, namely training convolutional neural networks.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to include added logistic noise to the output vector generation of the DNN, as taught by Van den Oord, to generating the set of masks in the DNN taught by Jin/Li.
Van den Oord provides motivation as [using added (logistic) noise facilitates parallel processing and faster output generation (para. 0006, etc.) and allows the neural network to be effectively conditioned on the context input to generate a high-quality output example while having fewer parameters and being less computationally complex (para. 0024, etc.)].
As per claim 8, Jin/Harikumar/Li/Van den Oord teaches wherein the nonlinear activation function comprises a sigmoid function [the channel attention gates are followed by activation function layers, including a ReLU or sigmoid (nonlinear) activation functions (Jin: pg. 10, supplementary 1; etc.); where the DNN may be trained to output a binary confidence mask indicating confidence that a pixel belongs to a class, using the activation function output, which may include ReLU or sigmoid functions (Li: paras. 0025-26)].
As per claim 25, see the rejection of claim 7, above.
Claim(s) 13-17 and 28 is/are rejected under 35 U.S.C. 103 as being unpatentable over Jin et al. (Feature Alignment and Restoration for Domain Generalization and Adaptation, June 2020, pgs. 1-15 – cited in an IDS), in view of Harikumar (US 2022/0156992), and further in view of Li et al. (US 2019/0258925 – hereinafter Shen to distinguish it from the reference cited above).
As per claim 13, Jin/Harikumar teaches generating the loss [we design a dual ranking entropy loss constraint to promote the distillation of the task-relevant features from the residual, trained on unlabeled images from a target domain (pgs. 3-4, sections 3-3.1; pg. 5, section 3.2; etc.)].
While Jin/Harikumar teaches training the target domain feature extraction model using a loss function (see above), it has not been relied upon for teaching using a loss function comprising a regularization loss component.
Shen teaches using a loss function comprising a regularization loss component [the loss function includes a number of weighted components, including a regularization loss function component (paras. 0103-104; equation (9); etc.)].
Jin/Harikumar and Shen are analogous art, as they are within the same field of endeavor, namely building/training an attention controlled neural network for image classification.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to include the regularization loss function component in the loss function, as taught by Shen, in the loss function used for training the target domain feature extraction model in the system taught by Jin/Harikumar.
Shen provides motivation as [adding the regularization loss improves the accuracy of the loss function in training the neural network (para. 0103, etc.)].
As per claim 14, Jin/Harikumar/Shen teaches wherein the regularization loss component comprises a Euclidean distance function [the regularization loss term includes a distance function (Shen: paras. 0103-104, etc.) which can be based on Euclidean distance (Shen: para. 0061, etc.)].
As per claim 15, Jin/Harikumar/Shen teaches wherein the loss function further comprises a cross-entropy loss component [the FA/FR model loss function includes a weighted summation of alignment loss Lalign, dual ranking entropy loss LDRE, and basic classification loss (i.e., cross-entropy loss) Lcls (Jin: pg. 5, section 3.2; pg. 11, supplementary 2; etc.); where the overall loss function can include a cross-entropy loss function (Shen: para. 0068, etc.)].
As per claim 16, Jin/Harikumar/Shen teaches wherein for a given sample, the cross-entropy loss component is configured to generate a cross-entropy loss value based on a positive feature of the set of task features generated by the mask generation model based on the given sample and a classification output generated by a linear classification model based on the given sample [the FA/FR model loss function includes a weighted summation of alignment loss Lalign, dual ranking entropy loss LDRE, and basic classification loss (i.e., cross-entropy loss) Lcls (Jin: pg. 5, section 3.2; pg. 11, supplementary 2; etc.); where the classification loss is determined from outputs of the FA/FR (mask generation) and classification models (Jin: pg. 3, fig. 2; etc.); where the overall loss function can include a cross-entropy loss function (Shen: para. 0068, etc.)].
As per claim 17, Jin/Harikumar/Shen teaches wherein the loss function further comprises a weighting parameter for the regularization loss component [the loss function includes a number of components weighted with a hyper-parameter, including a regularization loss function component (Shen: paras. 0103-104; equation (9); etc.)].
As per claim 28, see the rejections of claims 13-14, above.
Response to Arguments
Applicant's arguments, regarding the rejections under 35 U.S.C. 101, filed 9 April 2026 have been fully considered but they are not persuasive.
Applicant argues that the claimed invention provide a solution to a problem with “conventional machine learning” by training a “feature extraction model” and “training a mask generator (e.g., a mask generation model) to tune the feature extraction model.” Applicant further argues that this facilitates training of a feature extraction model on the source domain and training a mask generator on the target domain to tune output of the feature extraction model (e.g., select salient features from the output of the feature extractor), and allows for machine learning models trained with “relatively few training samples for the target domain to…achieve a high task accuracy.”
However, applicant has described an improvement to selecting features. As described above, this is a mental process/mathematical concepts. Therefore, (assuming that the invention provides these advantages) this amounts to an improvement to an abstract idea rather than to a computer or technology. See MPEP 2106.05(a). It appears that any benefits to the computer itself are based solely on the use of an improvement to the abstract idea(s), using generic computer components to apply the abstract idea(s). Additionally, to find a valid improvement to a computer or technology the specification must disclose the improvement and the claim must include the necessary components to realize the improvement. MPEP 2106.05(d)(1).
Applicant also argues that the claimed invention integrates the abstract idea into a practical application—namely, training and deploying machine learning models, which involves adapting the machine learning models to different domains using few training samples.
However, simply training and deploying machine learning models amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Applicant also argues that the practical application improves the technical field of machine learning models that include a feature extraction model that may be trained with mask generators. Applicant argues that “the mask generators may select salient features from output of the source domain feature extractor based on a target domain. The generated mask(s) can improve training of the feature extractor by forcing it to focus on selected features.”
However, applicant has described an improvement to selecting features. As described above, this is a mental process/mathematical calculations. Therefore, (assuming that the invention provides these advantages) this amounts to an improvement to an abstract idea rather than to a computer or technology. See MPEP 2106.05(a). It appears that any benefits to the computer itself are based solely on the use of an improvement to the abstract idea(s), using generic computer components to apply the abstract idea(s). Additionally, to find a valid improvement to a computer or technology the specification must disclose the improvement and the claim must include the necessary components to realize the improvement. MPEP 2106.05(d)(1).
Applicant also argues that the present claims are not merely instructions to apply an exception using a generic computer component, but provide a specific improvement in machine learning technology that integrates the abstract idea into a practical application—that is the ability to train machine learning models with increased accuracy and decreased reliance on large data sets. Applicant further compares the claimed invention to the decision in Ex parte Desjardins, and argues that the claimed invention can be “applied to a wide variety of machine learning tasks, and can generally improve the accuracy of models in any number of task domains.”
While the examiner has identified (above) what constitutes the abstract idea and what is drawn to conventional components, the Federal Circuit has also indicated that mere automation of manual processes or increasing the speed of a process, where these purported improvements come solely from the capabilities of a general-purpose computer are not sufficient to show an improvement in computer functionality. FairWarning IP, LLC v. latric Sys., 839 F.3d 1089, 1095, 120 USPQ2d 1293, 1296 (Fed. Cir. 2016); Credit Acceptance Corp. v. Westlake Services, 859 F.3d 1044, 1055, 123 USPQ2d 1100, 1108-09 (Fed. Cir. 2017). The Federal Circuit has also indicated that a claim must include more than conventional implementation on generic components or machinery to qualify as an improvement to an existing technology. Affinity Labs of Tex. v. DirecTV, LLC, 838 F.3d 1253, 1264-65, 120 USPQ2d 1201, 1208-09 (Fed. Cir. 2016); TLI Communications LLC v. AVAuto, LLC, 823 F.3d 607, 612-613, 118 USPQ2d 1744, 1747-48 (Fed. Cir. 2016). Claims must also include more than just instructions to perform the method on a generic component or machinery to qualify as an improvement to an existing technology (MPEP § 2106.05(a)).
Applicant’s arguments directed to the amendments made to the independent claims are addressed by the updated rejections, above. Examiner notes that applying generic machine learning models amounts to no more than a recitation of the words "apply it" (or an equivalent) including mere instructions to implement an abstract idea or other exception on a computer, and/or at most generally linking the use of the judicial exception to a particular technological environment or field of use. See MPEP § 2106.05(f) and (h).
Applicant also argues that “conventional solutions require a significant amount of data for training and may be limited to the specific environments in which these models are trained,” “tend to over-fit to the particular samples used and will perform poorly…for other new data samples” and that the claimed invention provides an improvement that addresses these issues.
However, applicant has described an improvement to selecting features and applying selected loss function(s). As described above, this is a mental process/mathematical concepts. Therefore, (assuming that the invention provides these advantages) this amounts to an improvement to an abstract idea rather than to a computer or technology. See MPEP 2106.05(a). It appears that any benefits to the computer itself are based solely on the use of an improvement to the abstract idea(s), using generic computer components to apply the abstract idea(s). Additionally, to find a valid improvement to a computer or technology the specification must disclose the improvement and the claim must include the necessary components to realize the improvement. MPEP 2106.05(d)(1).
Applicant’s arguments, see the remarks, filed 9 April 2026, with respect to the rejection(s) of claim(s) 1, 5, 9-12, 18-20, 26, 27, and 29 under 35 U.S.C. 102 have been fully considered and are persuasive in view of the amendments made to the independent claims. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Harikumar, which has been relied upon for teaching a mask generation model trained on only the target data set (see above).
Conclusion
The following is a summary of the treatment and status of all claims in the application as recommended by M.P.E.P. 707.07(i): claims 1-29 are rejected.
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Yosinski et al. (How transferable are features in deep neural networks, Nov 2014, pgs. 1-5 – cited in an IDS) – discloses determining the transferability of features and using transferred features to initialize a network.
Jin et al. (Style Normalization and Restitution for Domain Generalization and Adaptation, 3 Jan 2021, pgs. 1-16 – cited in an IDS) – discloses DG/UDA by adding SNR blocks/layers to a network, including attention on positive/negative features and dual causality loss function(s).
Bousmalis et al. (Unsupervised Pixel-Level Domain Adaptation with Generative Adversarial Networks, Aug 2017, pgs. 1-15) – discloses a system/method training models for pixel-level UDA, including generating and applying binary masks.
Saito et al. (Maximum Classifier Discrepancy for Unsupervised Domain Adaptation, April 2018, pgs. 1-12) – discloses UDA utilizing discrepancy loss, including maximizing and minimizing discrepancy between classifiers.
Reisser (US 2019/0354865) – discloses quantizing weights of a neural network using a distribution modeled using logistic noise.
Puscas (US 11,544,532) – discloses training a neural network including a sigmoid activation function used to generate a mask that produces a binary mask.
Abrol (US 2021/0312674) – discloses training an SAE post processing model for source to target domain mask generation.
Lin (US 11,449,079) – discloses control techniques including attention mask generation models.
The examiner requests, in response to this Office action, that support be shown for language added to any original claims on amendment and any new claims. That is, indicate support for newly added claim language by specifically pointing to page(s) and line number(s) in the specification and/or drawing figure(s). This will assist the examiner in prosecuting the application.
When responding to this office action, Applicant is advised to clearly point out the patentable novelty which he or she thinks the claims present, in view of the state of the art disclosed by the references cited or the objections made. He or she must also show how the amendments avoid such references or objections. See 37 CFR 1.111(c).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GEORGE GIROUX whose telephone number is (571)272-9769. The examiner can normally be reached M-F 10am-6pm.
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, Omar Fernandez Rivas can be reached at 571-272-2589. 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.
/GEORGE GIROUX/Primary Examiner, Art Unit 2128
