DETAILED ACTION
This action is in response to the filing on 07/03/2026. Claims 1, 3-11, and 13-20, 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 § 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, 3-11, and 13-20 are rejected under 35 U.S.C 101 because the claimed invention is directed to an abstract idea without significantly more.
Independent Claims 1 and 11
Step 1:
Claims 1 and 11 recite a method 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:
An optimizing method of semi-supervised learning, applicable to a labeled data set and an unlabeled data set, and the optimizing method comprises — 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.
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set — 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)).
determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set 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.
selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold — 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.
updating the machine learning model according to respective compared results of second confidence scores of the second predicted result of the at least one second sample in the unlabeled data set and the respective pseudo-label threshold corresponding to a category associated with the second confidence scores — 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:
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set — 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)).
determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set 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.
selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold — 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.
updating the machine learning model according to respective compared results of second confidence scores of the second predicted result of the at least one second sample in the unlabeled data set and the respective pseudo-label threshold corresponding to a category associated with the second confidence scores — 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 1 recites the additional elements of:
wherein at least one first sample in the labeled data set is labeled as one of a plurality of categories, and at least one second sample in the unlabeled data set is not labeled as one of the plurality of categories, — This element amounts to no more than insignificant extra-solution activity in the form of insignificant application (see MPEP § 2106.05(g)), and is well-known, understood, routine, conventional activity in view of YI XU et al. ("Dash: Semi-Supervised Learning with Dynamic Thresholding", ICML 2021, September 1, 2021, pp. 1-22. As cited in IDS filed 08/09/2023), Xu discloses self-supervised learning is known in the art for training machine learning models with both labeled and unlabeled data (see Xu, Section 1, para. 2).
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set through a machine learning model — This element amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP § 2106.05(f)).
Claim 11 recites the additional elements of:
A computing apparatus, applicable to a labeled data set and an unlabeled data set, and the computing apparatus comprises — 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.
wherein at least one first sample in the labeled data set is labeled as one of a plurality of categories, and at least one second sample in the unlabeled data set is not labeled as one of the plurality of categories, — This element amounts to no more than insignificant extra-solution activity in the form of insignificant application (see MPEP § 2106.05(g)), and is well-known, understood, routine, conventional activity in view of YI XU et al. ("Dash: Semi-Supervised Learning with Dynamic Thresholding", ICML 2021, September 1, 2021, pp. 1-22. As cited in IDS filed 08/09/2023), Xu discloses self-supervised learning is known in the art for training machine learning models with both labeled and unlabeled data (see Xu, Section 1, para. 2).
a storage device, storing program code; and a processor, coupled to the storage device and loading the program code to execute — 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 generic computer components.
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set through a machine learning model — This element amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP § 2106.05(f)).
Step 2B:
The claims do not contain significantly more than the judicial exception.
Claim 1 recites the additional elements of:
wherein at least one first sample in the labeled data set is labeled as one of a plurality of categories, and at least one second sample in the unlabeled data set is not labeled as one of the plurality of categories, — This element amounts to no more than insignificant extra-solution activity in the form of insignificant application (see MPEP § 2106.05(g)), and is well-known, understood, routine, conventional activity in view of YI XU et al. ("Dash: Semi-Supervised Learning with Dynamic Thresholding", ICML 2021, September 1, 2021, pp. 1-22. As cited in IDS filed 08/09/2023), Xu discloses self-supervised learning is known in the art for training machine learning models with both labeled and unlabeled data (see Xu, Section 1, para. 2).
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set through a machine learning model — This element amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP § 2106.05(f)).
Claim 11 recites the additional elements of:
A computing apparatus, applicable to a labeled data set and an unlabeled data set, and the computing apparatus comprises — 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.
wherein at least one first sample in the labeled data set is labeled as one of a plurality of categories, and at least one second sample in the unlabeled data set is not labeled as one of the plurality of categories, — This element amounts to no more than insignificant extra-solution activity in the form of insignificant application (see MPEP § 2106.05(g)), and is well-known, understood, routine, conventional activity in view of YI XU et al. ("Dash: Semi-Supervised Learning with Dynamic Thresholding", ICML 2021, September 1, 2021, pp. 1-22. As cited in IDS filed 08/09/2023), Xu discloses self-supervised learning is known in the art for training machine learning models with both labeled and unlabeled data (see Xu, Section 1, para. 2).
a storage device, storing program code; and a processor, coupled to the storage device and loading the program code to execute — 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 generic computer components.
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set through a machine learning model — This element amounts to no more than adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP § 2106.05(f)).
As such claims 1 and 11 are not patent eligible.
Dependent Claims 3-10 and 13-20
Step 1:
Claims 3-10 and 13-20 recite a method 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:
Claims 3-10 and 13-20 merely narrow the previously cited abstract idea limitations. For the reasons described above with respect to independent claims 1 and 11 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.
Claims 3 and 13 recite a method and system respectively comprising:
selecting a highest score in the second confidence scores of the second predicted result of the at least one second sample, wherein the highest score corresponds to the first category — 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.
selecting one of the first confidence scores corresponding to the first category from the first predicted result of the at least one first sample labeled as the first category as the pseudo-label threshold for comparing the first category in the at least one second sample — 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.
Claims 4 and 14 recite a method and system respectively comprising:
sorting the first confidence scores of the at least one first sample in the first score table according to the magnitude of the first confidence scores — 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.
selecting a first confidence score from the first score table according to a confidence level index or selecting the highest one of the first confidence scores from the first score table as the pseudo-label threshold — 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.
Claims 5 and 15 recite a method and system respectively comprising:
defining a capacity of the first score table — 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.
in response to an amount of first confidence scores of the at least one first sample added to the first score table being greater than the capacity, deleting some of the first confidence scores of the at least one first sample in the first score table according to a sequence of addition to the first score table — 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.
Claims 6 and 16 recite a method and system respectively comprising:
in response to a highest score in the second confidence scores being less than the pseudo-label threshold, prohibiting the at least one second sample or the corresponding second predicted result from being used to update the machine learning model per 35 U.S.C. 112(b) rejection above) — 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.
In response to the highest score in the second confidence scores not being less than the pseudo-label threshold, allowing the at least one second sample or the corresponding second predicted result to be used for updating the machine learning model per 35 U.S.C. 112(b) rejection above)— 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.
Claims 7 and 17 recite a method and system respectively comprising:
establishing a loss function according to the second confidence scores corresponding to the categories in the second predicted result of the at least one second sample and a label corresponding to the highest score — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept, specifically, mathematical formulas or equations (see MPEP § 2106.04(a)(2)(I)(B)).
updating the machine learning model according to the loss function — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept, specifically, organizing information and manipulating information through mathematical correlations (see MPEP § 2106.04(a)(2)(I)(A)) and mathematical formulas or equations (see MPEP § 2106.04(a)(2)(I)(B)).
Claims 8 and 18 recite a method and system respectively comprising:
establishing another loss function according to the first confidence scores corresponding to the categories in the first predicted result of each of the least one first sample and labeled category — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept, specifically, mathematical formulas or equations (see MPEP § 2106.04(a)(2)(I)(B)).
and updating the machine learning model according to the another loss function — Under its broadest reasonable interpretation, this limitation encompasses the abstract idea of a mathematical concept, specifically, organizing information and manipulating information through mathematical correlations (see MPEP § 2106.04(a)(2)(I)(A)) and mathematical formulas or equations (see MPEP § 2106.04(a)(2)(I)(B)).
Step 2A Prong 2:
This judicial exception is not integrated into a practical application.
Claims 4 and 14 recite the additional element of:
adding the first confidence scores corresponding to the first category from the first predicted result of the at least one first sample labeled as the first category in a first score table — This element amounts to no more than insignificant extra-solution activity in the form of mere data gathering and output (see MPEP § 2106.05(g)), and is well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II), storing and retrieving information in memory).
Claims 9 and 19 recite the additional element of:
removing at least one redundant neuron in an updated machine learning model; and adjusting weight corresponding to neurons other than the at least one redundant neuron in the updated machine learning model. — This element amounts to no more than insignificant extra-solution activity in the form of insignificant application (see MPEP § 2106.05(g)), and is well-understood, routine, conventional activity in view of Hengyuan Hu et al. ("Network Trimming: A Data-Driven Neuron Pruning Approach towards Efficient Deep Architectures", arXiv:1607.03250v1 [cs.NE] 12 Jul 2016), Hu discusses previous work of pruning neural networks (see Hu, p. 2, Section 2) as well as retraining the network after pruning (see Hu, p. 3-4, Section 3.2).
Claims 10 and 20 recite the additional element of:
increasing a parameter capacity of the machine learning model, wherein the parameter capacity is an amount of parameter inputs for the machine learning model — This element amounts to no more than insignificant extra-solution activity in the form of mere data gathering and output (see MPEP § 2106.05(g)), and is well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II), storing and retrieving information in memory).
Step 2B:
The claims do not contain significantly more than the judicial exception.
Claims 4 and 14 recite the additional element of:
adding the first confidence scores corresponding to the first category from the first predicted result of the at least one first sample labeled as the first category in a first score table — This element amounts to no more than insignificant extra-solution activity in the form of mere data gathering and output (see MPEP § 2106.05(g)), and is well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II), storing and retrieving information in memory).
Claims 9 and 19 recite the additional element of:
removing at least one redundant neuron in an updated machine learning model; and adjusting weight corresponding to neurons other than the at least one redundant neuron in the updated machine learning model. — This element amounts to no more than insignificant extra-solution activity in the form of insignificant application (see MPEP § 2106.05(g)), and is well-understood, routine, conventional activity in view of Hengyuan Hu et al. ("Network Trimming: A Data-Driven Neuron Pruning Approach towards Efficient Deep Architectures", arXiv:1607.03250v1 [cs.NE] 12 Jul 2016), Hu discusses previous work of pruning neural networks (see Hu, p. 2, Section 2) as well as retraining the network after pruning (see Hu, p. 3-4, Section 3.2).
Claims 10 and 20 recite the additional element of:
increasing a parameter capacity of the machine learning model, wherein the parameter capacity is an amount of parameter inputs for the machine learning model — This element amounts to no more than insignificant extra-solution activity in the form of mere data gathering and output (see MPEP § 2106.05(g)), and is well-understood, routine, conventional activity (see MPEP § 2106.05(d)(II), storing and retrieving information in memory).
As such claims 3-10 and 13-20 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, 6-8, 11, 13, and 16-18 are rejected under 35 U.S.C. 103 as being unpatentable in view over YI XU et al. ("Dash: Semi-Supervised Learning with Dynamic Thresholding", ICML 2021, September 1, 2021, pp. 1-22. As cited in IDS filed 08/09/2023), hereinafter Xu, in view of SNOW et al. (US 2022/0391765 A1, first cited in IDS filed 12/31/2025), hereinafter Snow, and further in view of Wang, Yidong, et al. ("Freematch: Self-adaptive thresholding for semi-supervised learning." arXiv preprint arXiv:2205.07246), hereinafter Wang.
Regarding claim 1, Xu teaches An optimizing method of semi-supervised learning, applicable to a labeled data set and an unlabeled data set, wherein at least one first sample in the labeled data set is labeled as one of a plurality of categories, and at least one second sample in the unlabeled data set is not labeled as one of the plurality of categories, and the optimizing method comprises (Xu discloses a method of semi-supervised learning a labeled data set Dl and an unlabeled data set Du, where the labeled dataset has labeled samples, and the unlabeled dataset has unlabeled samples [see Xu, p., 4, Section 3.1, para. 2-3]):
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set through a machine learning model (Xu discloses a cross-entropy loss function that determines the cross-entropy between the known label yi and the prediction p(w, xi) for the input pair (x,y) in the labeled data Dl and unlabeled data Du [see Xu, p. 4, Section 3.1, para. 1-2 and Eq. 2]. Thus, predictions for given inputs in the labeled and unlabeled data samples are determined);
determining a pseudo-label threshold according to first confidence scores of the first predicted result of the at least one first sample of the labeled data set (Xu discloses that FixMatch uses unlabeled examples with a fixed confidence prediction using cross entropy loss [see Xu, p. 3, para. 1] and in the proposed Dash method they determine a dynamic pseudo label threshold ρt using the same cross entropy loss confidence as FixMatch on the labeled data [see Xu, p. 3, para. 3; p. 6, para. 2; p. 7, para. 2-p. 8, para. 1]. Thus, the pseudo-label dynamic threshold is determined accorded to confidence scores of the labeled data);
updating the machine learning model according to respective compared results of second confidence scores of the second predicted result of the at least one second sample in the unlabeled data set and the respective pseudo-label threshold corresponding to a category associated with the second confidence scores (Xu discloses that FixMatch uses unlabeled examples with a fixed confidence prediction using cross entropy loss [see Xu, p. 3, para. 1] and in the proposed Dash method they determine a dynamic pseudo label threshold ρt using the same cross entropy loss confidence as FixMatch on the labeled data [see Xu, p. 3, para. 3; p. 6, para. 2; p. 7, para. 2-p. 8, para. 1]. Thus, the pseudo-label dynamic threshold is determined accorded to confidence scores of the labeled data. Further, Xu discloses retaining the unlabeled samples with unsupervised loss smaller than the threshold [see Xu, p. 7, para. 1] in the same manner FixMatch selecting unlabeled example if its confidence prediction is greater than 0.95. [see Xu, p. 2, Figure 1]. Thus, Xu uses the unlabeled samples to update the machine learning model if their cross entropy loss is smaller than the dynamic pseudo-label threshold (i.e. if their confidence is greater than the pseudo-label threshold)).
However, Xu fails to teach determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set; and selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold.
In the same field of endeavor, Snow teaches:
selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold (Snow discloses setting the active learning threshold to the confidence measure of the nth most uncertain prediction [see Snow, para. 69]).
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 selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold as suggested in Snow into Xu because both methods are directed to semi-supervised learning (see Xu, Abstract, see Snow, Abstract). Incorporating the teaching of Snow into Xu would provide advantages in the training and/or performance (e.g. in inference) of a machine learning model where labeled training data is scarce relative to unlabeled data, unlabeled data is abundant, and/or the cost of labeling data (e.g. in terms of time, expertise, resources, etc.) is high (see Snow, para. 66-67).
However, the combination of Xu and Snow fails to teach determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set.
In the same field of endeavor, Wang teaches:
determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set (Wang discloses determining a plurality of local class-specific self-adaptive thresholds based on the model’s learning status, wherein the local thresholds are determined based on the model’s confidence about its predictions [see Wang, p. 2, para. 3-4]).
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set as suggested in Wang into the combination of Xu and Snow because both methods perform semi-supervised learning with pseudo-label threshold (see Xu, Abstract; Wang, p. 2, para. 3-4). Incorporating the teaching of Wang into the combination of Xu and Snow would maximize mutual information between model’s input and output producing confident and diverse predictions on unlabeled data (see Wang, p. 2, para. 4).
Regarding claim 3, the combination of Xu, Snow, and Wang as applied in claim 1 above teaches all the limitations of claim 1 and further teaches:
selecting a highest score in the second confidence scores of the second predicted result of the at least one second sample, wherein the highest score corresponds to the first category (Xu discloses retaining the unlabeled samples with unsupervised loss smaller than the threshold [see Xu, p. 7, para. 1]. Thus, Xu would select at least the highest score in the unlabeled confidence scores because it checks all of the scores);
selecting one of the first confidence scores corresponding to the first category from the first predicted result of the at least one first sample labeled as the first category as the pseudo-label threshold for comparing the first category in the at least one second sample (Snow discloses setting the active learning threshold to the confidence measure of the nth most uncertain prediction [see Snow, para. 69]).
Regarding claim 6, the combination of Xu, Snow, and Wang as applied in claim 1 above teaches all the limitations of claim 1 and further teaches:
prohibiting the at least one second sample with a highest score in the second confidence scores being less than the pseudo-label threshold or the corresponding second predicted result from being used to update the machine learning model; and allowing the at least one second sample with a highest score in the second confidence scores not being less than the pseudo-label threshold or the corresponding second predicted result to be used for updating the machine learning model (Xu discloses that FixMatch uses unlabeled examples with a fixed confidence prediction using cross entropy loss [see Xu, p. 3, para. 1] and in the proposed Dash method they determine a dynamic pseudo label threshold ρt using the same cross entropy loss confidence as FixMatch on the labeled data [see Xu, p. 3, para. 3; p. 6, para. 2; p. 7, para. 2-p. 8, para. 1]. Thus, the pseudo-label dynamic threshold is determined accorded to confidence scores of the labeled data. Further, Xu discloses retaining the unlabeled samples with unsupervised loss smaller than the threshold [see Xu, p. 7, para. 1] in the same manner FixMatch selecting unlabeled example if its confidence prediction is greater than 0.95. [see Xu, p. 2, Figure 1]. Thus, Xu uses the unlabeled samples to update the machine learning model if their cross entropy loss is smaller than the dynamic pseudo-label threshold (i.e. if their confidence is greater than the pseudo-label threshold), and does not the unlabeled samples to update the machine learning model if their cross entropy loss is greater than the dynamic pseudo-label threshold (i.e. if their confidence is lower than the pseudo-label threshold)).
Regarding claim 7, the combination of Xu, Snow, and Wang as applied in claim 6 above teaches all the limitations of claim 6 and further teaches:
establishing a loss function according to the second confidence scores corresponding to the categories in the second predicted result of the at least one second sample and a label corresponding to the highest score; and updating the machine learning model according to the loss function (Xu discloses that FixMatch uses a fixed confidence prediction using cross entropy loss [see Xu, p. 3, para. 1] and in the proposed Dash method they use the same cross entropy loss confidence as FixMatch [see Xu, p 6, para. 2]. Xu further discloses using the unlabeled data's cross entropy loss (i.e. confidence) as part of the loss function used to update the model during stochastic gradient descent [see Xu, p. 5-8, Section 4 and p. 6, Algorithm 1]).
Regarding claim 8, the combination of Xu, Snow, and Wang as applied in claim 7 above teaches all the limitations of claim 7 and further teaches:
establishing another loss function according to the first confidence scores corresponding to the categories in the first predicted result of each of the least one first sample and labeled category; and updating the machine learning model according to the another loss function (Xu discloses that FixMatch uses a fixed confidence prediction using cross entropy loss [see Xu, p. 3, para. 1] and in the proposed Dash method they use the same cross entropy loss confidence as FixMatch [see Xu, p 6, para. 2]. Xu further discloses using the labeled data's cross entropy loss (i.e. confidence) as part of the loss function used to update the model during stochastic gradient descent [see Xu, p. 5-8, Section 4 and p. 6, Algorithm 1]).
Regarding claim 11, Xu teaches an optimizing method of semi-supervised learning, applicable to a labeled data set and an unlabeled data set, wherein at least one first sample in the labeled data set is labeled as one of a plurality of categories, and at least one second sample in the unlabeled data set is not labeled as one of the plurality of categories, and the optimizing method comprises (Xu discloses a method of semi-supervised learning a labeled data set Dl and an unlabeled data set Du, where the labeled dataset has labeled samples, and the unlabeled dataset has unlabeled samples [see Xu, p., 4, Section 3.1, para. 2-3]):
respectively determining a first predicted result of the labeled data set and a second predicted result of the unlabeled data set through a machine learning model (Xu discloses a cross-entropy loss function that determines the cross-entropy between the known label yi and the prediction p(w, xi) for the input pair (x,y) in the labeled data Dl and unlabeled data Du [see Xu, p. 4, Section 3.1, para. 1-2 and Eq. 2]. Thus, predictions for given inputs in the labeled and unlabeled data samples are determined);
determining a pseudo-label threshold according to first confidence scores of the first predicted result of the at least one first sample of the labeled data set (Xu discloses that FixMatch uses unlabeled examples with a fixed confidence prediction using cross entropy loss [see Xu, p. 3, para. 1] and in the proposed Dash method they determine a dynamic pseudo label threshold ρt using the same cross entropy loss confidence as FixMatch on the labeled data [see Xu, p. 3, para. 3; p. 6, para. 2; p. 7, para. 2-p. 8, para. 1]. Thus, the pseudo-label dynamic threshold is determined accorded to confidence scores of the labeled data);
updating the machine learning model according to respective compared results of second confidence scores of the second predicted result of the at least one second sample in the unlabeled data set and the respective pseudo-label threshold corresponding to a category associated with the second confidence scores (Xu discloses that FixMatch uses unlabeled examples with a fixed confidence prediction using cross entropy loss [see Xu, p. 3, para. 1] and in the proposed Dash method they determine a dynamic pseudo label threshold ρt using the same cross entropy loss confidence as FixMatch on the labeled data [see Xu, p. 3, para. 3; p. 6, para. 2; p. 7, para. 2-p. 8, para. 1]. Thus, the pseudo-label dynamic threshold is determined accorded to confidence scores of the labeled data. Further, Xu discloses retaining the unlabeled samples with unsupervised loss smaller than the threshold [see Xu, p. 7, para. 1] in the same manner FixMatch selecting unlabeled example if its confidence prediction is greater than 0.95. [see Xu, p. 2, Figure 1]. Thus, Xu uses the unlabeled samples to update the machine learning model if their cross entropy loss is smaller than the dynamic pseudo-label threshold (i.e. if their confidence is greater than the pseudo-label threshold)).
However, Xu fails to teach A computing apparatus, and the computing apparatus comprises: a storage device, storing program code; and a processor, coupled to the storage device and loading the program code to execute; determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set; and selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold.
It would have been obvious to one of ordinary skill in the art before the effective filing date to use the method of Xu [see Xu, p., 4, Section 3.1, para. 2-3] on a generic computer comprising generic computer components including storage and a processor, the storage comprising program code to cause the processor to execute the method of Xu to teach A computing apparatus, and the computing apparatus comprises: a storage device, storing program code; and a processor, coupled to the storage device and loading the program code to execute.
However, Xu fails to teach determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set; and selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold.
In the same field of endeavor, Snow teaches:
selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold (Snow discloses setting the active learning threshold to the confidence measure of the nth most uncertain prediction [see Snow, para. 69]).
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 selecting one of the first confidence scores of the first predicted result of the at least one first sample as the pseudo-label threshold as suggested in Snow into Xu because both methods are directed to semi-supervised learning (see Xu, Abstract, see Snow, Abstract). Incorporating the teaching of Snow into Xu would provide advantages in the training and/or performance (e.g. in inference) of a machine learning model where labeled training data is scarce relative to unlabeled data, unlabeled data is abundant, and/or the cost of labeling data (e.g. in terms of time, expertise, resources, etc.) is high (see Snow, para. 66-67).
However, the combination of Xu and Snow fails to teach determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set.
In the same field of endeavor, Wang teaches:
determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set (Wang discloses determining a plurality of local class-specific self-adaptive thresholds based on the model’s learning status, wherein the local thresholds are determined based on the model’s confidence about its predictions [see Wang, p. 2, para. 3-4]).
It would have been obvious to one of ordinary skill in the art before the effective filing date to incorporate determining, for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set as suggested in Wang into the combination of Xu and Snow because both methods perform semi-supervised learning with pseudo-label threshold (see Xu, Abstract; Wang, p. 2, para. 3-4). Incorporating the teaching of Wang into the combination of Xu and Snow would maximize mutual information between model’s input and output producing confident and diverse predictions on unlabeled data (see Wang, p. 2, para. 4).
Regarding claim 13, claim 13 contains substantially similar limitations to those found in claim 3 above. Consequently, claim 13 is rejected for the same reasons.
Regarding claim 16, claim 16 contains substantially similar limitations to those found in claim 6 above. Consequently, claim 16 is rejected for the same reasons.
Regarding claim 17, claim 17 contains substantially similar limitations to those found in claim 7 above. Consequently, claim 17 is rejected for the same reasons.
Regarding claim 18, claim 18 contains substantially similar limitations to those found in claim 8 above. Consequently, claim 18 is rejected for the same reasons.
Claims 9 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over YI XU et al. ("Dash: Semi-Supervised Learning with Dynamic Thresholding", ICML 2021, September 1, 2021, pp. 1-22. As cited in IDS filed 08/09/2023), hereinafter Xu, in view of SNOW et al. (US 2022/0391765 A1, first cited in IDS filed 12/31/2025), hereinafter Snow, and further in view of Wang, Yidong, et al. ("Freematch: Self-adaptive thresholding for semi-supervised learning." arXiv preprint arXiv:2205.07246), hereinafter Wang, as applied in claim 1 above, in view of Hengyuan Hu et al. ("Network Trimming: A Data-Driven Neuron Pruning Approach towards Efficient Deep Architectures", arXiv:1607.03250v1 [cs.NE] 12 Jul 2016), hereinafter Hu.
Regarding claim 9, the combination of Xu, Snow, and Wang as applied in claim 1 above teaches all the limitations of claim 1.
However, the combination of Xu, Snow, and Wang fails to teach removing at least one redundant neuron in an updated machine learning model; and adjusting weight corresponding to neurons other than the at least one redundant neuron in the updated machine learning model.
In the same field of endeavor, Hu teaches:
removing at least one redundant neuron in an updated machine learning model (In this paper, we introduce network trimming which iteratively optimizes the network by pruning unimportant neurons based on analysis of their outputs on a large dataset. Our algorithm is inspired by an observation that the outputs of a significant portion of neurons in a large network are mostly zero, regardless of what inputs the network received. These zero activation neurons are redundant, and can be removed without affecting the overall accuracy of the network. [see Hu, Abstract]);
adjusting weight corresponding to neurons other than the at least one redundant neuron in the updated machine learning model (After pruning the zero activation neurons, we retrain the network using the weights before pruning as initialization. We alternate the pruning and retraining to further reduce zero activations in a network. [see Hu, Abstract]).
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 removing at least one redundant neuron in an updated machine learning model; and adjusting weight corresponding to neurons other than the at least one redundant neuron in the updated machine learning model as suggested in Hu into the combination of Xu, Snow, and Wang because both methods are directed to machine learning (see Xu, Abstract; see Hu, Abstract). Incorporating the teaching of Hu into the combination of Xu, Snow, and Wang would achieve high compression ratio of parameters without losing or even achieving higher accuracy than the original network (see Hu, Abstract).
Regarding claim 19, claim 19 contains substantially similar limitations to those found in claim 9 above. Consequently, claim 19 is rejected for the same reasons.
Claims 10 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over YI XU et al. ("Dash: Semi-Supervised Learning with Dynamic Thresholding", ICML 2021, September 1, 2021, pp. 1-22. As cited in IDS filed 08/09/2023), hereinafter Xu, in view of SNOW et al. (US 2022/0391765 A1, first cited in IDS filed 12/31/2025), hereinafter Snow, and further in view of Wang, Yidong, et al. ("Freematch: Self-adaptive thresholding for semi-supervised learning." arXiv preprint arXiv:2205.07246), hereinafter Wang, as applied in claim 1 above, in view of Pascanu et al. (US 2019/0232489 A1), hereinafter Pascanu.
Regarding claim 10, the combination of Xu, Snow, and Wang as applied in claim 1 above teaches all the limitations of claim 1.
However, Xu fails to teach increasing a parameter capacity of the machine learning model, wherein the parameter capacity is an amount of parameter inputs for the machine learning model.
In the same field of endeavor, Pascanu teaches:
increasing a parameter capacity of the machine learning model, wherein the parameter capacity is an amount of parameter inputs for the machine learning model (By taking additional data as input, the second robot-trained DNN allows the neural network system 106 to have a flexibility for adding new capacity, including new input modalities, when transferring to a new task [see Pascanu, para. 52]).
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 increasing a parameter capacity of the machine learning model, wherein the parameter capacity is an amount of parameter inputs for the machine learning model as suggested in Pascanu into the combination of Xu, Snow, and Wang because both methods are directed to machine learning (see Xu, Abstract; see Pascanu, Abstract). Incorporating the teaching of Pascanu into the combination of Xu, Snow, and Wang would be able to accommodate changing network morphology or new input modalities, and advantageous for bridging the reality gap, to accommodate dissimilar inputs between simulation and real sensors (see Pascanu, para. 52).
Regarding claim 20, claim 20 contains substantially similar limitations to those found in claim 10 above. Consequently, claim 20 is rejected for the same reasons.
Allowable Subject Matter
Claims 4-5 and 14-15 objected to as being dependent upon a rejected base claim, but would be allowable if the rejection under 35 U.S.C. 101 is overcome and the claims are rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Response to Amendment
Applicant’s amendment to the claims, filed 07/03/2026, with respect to the objection of claims 1-3 and 11-13 have been fully considered and are accepted, the objections to claims 1-3, and 11-13 are respectfully withdrawn.
Applicant’s amendment to the claims, filed 07/03/2026, with respect to the rejection of claims 6-8 and 16-18 under 35 U.S.C. 112(b) have been fully considered and are accepted, the rejection of claims 6-8, and 16-18 under 35 U.S.C. 112(b) are respectfully withdrawn.
Response to Arguments
Applicant's arguments, filed 07/03/2026, traversing the rejection of claims 1, 3-11, and 13-20 under 35 U.S.C. 101 have been fully considered and are not persuasive. Applicant argues that the claims are not directed to a mental process and are integrated into a practical application, Examiner respectfully disagrees.
With respect to Applicants’ argument that the operations are performed on large amounts of unlabeled data and require repeated generation, storage, comparison, and updating of confidence-score-based thresholds across multiple categories; as Applicant has stated, amended claim 1 requires determining respective pseudo-label thresholds according to confidence scores and updating machine learning models according to comparisons between confidence scores and the thresholds. As identified in the previous office action mailed 05/04/2026 and identified in the 35 U.S.C. 101 section above, the determination of mathematical thresholds 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)). Further, the comparison of category-specific confidence scores and corresponding category-specific thresholds also encompasses a concept that can be performed in the human mind, including observation, evaluation, judgement or opinion. While the aforementioned operations may be performed on large amounts of unlabeled data, the claim language directly recites a mental process. Further, the claim language encompass a mathematical concept, specifically, organizing information and manipulating information through mathematical correlations (see MPEP § 2106.04(a)(2)(I)), as determining mathematical thresholds and comparing mathematical values to those thresholds is inherently manipulating information through mathematical correlations. Thus, even if the claim language were not reasonably performable in the human mind, the claim language would still be directed to a judicial exception as they recite mathematical concepts.
With respect to Applicant’s argument that the claims are integrated into a practical application, Applicant states the purported improvement of avoiding or reducing the data imbalance problem as reflected in para. 28 and 60 of the specification, is founded by determining for each category of the plurality of categories, a respective pseudo-label threshold according to first confidence scores, and selecting one of the first confidence scores as the pseudo-label threshold. However, 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)). Thus, the purported improvement cannot integrate the judicial exception, as the improvement is provided by the judicial exception itself, specifically the steps of determination and comparison as explained previously encompassing a judicial exception in the form of mental processes and mathematical concepts. Applicant further cites to Desjardins and Enfish, LLC v. Microsoft Corp., as precedential decisions where software innovations may provide non-abstract improvements to computer technology and should not be disregarded through overgeneralization. However, as explained, the purported improvement is provided by the judicial exception and thus cannot practically integrate the judicial exception.
Thus, for at least the aforementioned reasons, the rejection of claims 1, 3-11, and 13-20 under 35 U.S.C. 101 is respectfully maintained.
Applicant's arguments, filed 07/03/2026, traversing the rejection of claims 1, 3-11, and 13-20 under 35 U.S.C. 102(a)(1) and 35 U.S.C 103 have been fully considered and are not persuasive.
Applicant’s arguments with respect to the claim language reciting “determining, for each category, of the plurality of categories, a respective pseudo-label threshold according to first confidence scores corresponding to the category in the first predicted result of the at least one first sample of the labeled data set” 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.
Applicant’s further argues that Snow does not directly use a confidence score itself as the threshold and thus fails to teach “selecting one of the first confidence scores of the first predicted results of the at least one first sample as the pseudo-label threshold” citing to para. 50-54 and 69 of Snow, Examiner respectfully disagrees. As cited in para. 69 of Snow, “… to reduce the number of queries to the oracle, for example by setting the active learning threshold t2 (and/or threshold t) to a value corresponding to a confidence no greater than an upper confidence threshold… and setting the active learning threshold t2 (and/or threshold t) to the confidence measure for the nth-most-uncertain prediction (and/or any other suitable value for a confidence measure)” (emphasis added). Thus, Snow directly states setting the threshold t to a value corresponding to the confidence measure of the predictions for the nth-most-uncertain prediction or another suitable value for the confidence measure. While the disclosure of Snow may disclose other embodiments which select, determine, or update the threshold differently, the cited embodiment of Snow in the mapping directly sets the threshold as at least one of the confidence measures of a model’s predictions. Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date to select one of the confidence scores of the predicted results as the pseudo-label threshold.
Thus, for at least the aforementioned reasons, the rejection of claims 1, 3-11, and 13-20 under 35 U.S.C. 103 is respectfully maintained.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Guo, Lan-Zhe, and Yu-Feng Li. ("Class-imbalanced semi-supervised learning with adaptive thresholding." International conference on machine learning. PMLR, 2022.) discloses having adaptive thresholds for class-specific pseudo-labels to effectively train machine learning models
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