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 .
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.
DETAILED ACTION
This action is in response to applicant’s original submittal made on 06/09/2025. Claims 1-19 are pending.
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d).
Specification (Title)
The title of the invention is not descriptive. A new title is required that is clearly indicative of the invention to which the claims are directed to.
Claim Objections
Claims 6 and 15 are objected to because of the following informalities: The variables, “l, t, y” are not defined. Appropriate correction is required.
Claims 7 and 16 are objected to because of the following informalities: line 10 recites “a KD” loss value. The examiner notes that KD in this instance is not defined. Appropriate correction is required.
Claims 9 and 18 are objected to because of the following informalities: variables “CE TDL” are not defined. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1, 10 and 19 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. The examiner contends that the metes and bound of applicant's claim limitation(s) elements of, “fine-tuning”, "to-be-forgotten", “unlearned model”, “applying a knowledge distillation technique”, and “fine-tuned model”, cannot be readily determined. The examiner notes that the claim language of, “fine-tuning”, "to-be-forgotten", “applying a knowledge distillation technique”, “unlearned model” and “fine-tuned model”, creates uncertainty about what is being claimed.
Claims 2 and 11 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. The examiner contends that the metes and bound of applicant's claim limitation(s) elements of, "convolution layer", cannot be readily determined. The examiner notes that the claim language of, "convolution layer", creates uncertainty about what is being claimed.
Claims 4, 13 and 16 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. The examiner contends that the metes and bound of applicant's claim limitation(s) elements of, “fine-tune” and “fine-tuning”, cannot be readily determined. The examiner notes that the claim language of, “fine-tune” and “fine-tuning”, creates uncertainty about what is being claimed.
Claims 7, 8 and 16 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. The examiner contends that the metes and bound of applicant's claim limitation(s) elements of, “smoothed second logit by smoothing the second logit using a softmax function” and “smoothed second logit” cannot be readily determined. The examiner notes that the claim language of, “smoothed second logit by smoothing the second logit using a softmax function” and “smoothed second logit”, creates uncertainty about what is being claimed.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-4, 10-13 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Cai et al. (US Patent Publication No. 2025/0390785 and Cai hereinafter) in view of FANG (CN 116938542) and further in view of LIU et al. (US Patent Publication No. 2021/0142181 and LIU hereinafter).
As to claims 1, 10 and 19, Cai teaches a method for unlearning to be performed by an unlearning apparatus, the method comprising:
receiving to-be-forgotten target data as input (i.e., …teaches in par. 0012 the following: “1: A computer-implemented method for processing unlearning data from a machine learning model having L layers. The method comprises processing a query at a first set of layers of the L layers of the machine learning model to determine interim output. The method further comprises determining whether the interim output is classified as data to unlearn from the machine learning model to data to keep. The method further comprises processing the interim output at a second set of layers of the L layers of the machine learning model to produce a first output in response to determining that the interim output is classified as data to keep for the machine learning model. The method further comprises inputting the interim output to an edited machine learning model formed from the second set of layers to output second output in response to determining that the interim output is classified as the data to unlearn from the machine learning model.” …teaches in par. 0046 the following: “At block 408 the sample i is tokenized (at block 408) to produce a vector of tokens representing the sample. The tokens are inputted (at block 410) to the first set of layers M.sub.1 . . . k 114 to produce activations for the tokens.”.);
extracting a feature vector from the to-be-forgotten target data using a first layer that is a pre-attack layer of a pre-trained model (i.e., …teaches in par. 0045 the following: “For instance, semantically similar may be determined when the selected samples to keep 208 form embedded vectors in an embedded space that are within a predefined distance in the embedded space to embedded vectors of the samples to unlearn 206.” ……teaches in par. 0046 the following: “At block 408 the sample i is tokenized (at block 408) to produce a vector of tokens representing the sample. The tokens are inputted (at block 410) to the first set of layers M.sub.1 . . . k 114 to produce activations for the tokens.);
fine-tuning one of the first model and the second model by applying a knowledge distillation technique to the first model (i.e., …teaches in par. 0047 the following: “the training operations using the samples may be performed by batching the samples into one or more batches, where the feature vectors of the samples in the batch form concurrent vectors or a matrix of sample feature vectors to perform batch gradient descent to train.” …teaches in par. 0049 the following: “illustrates an embodiment of operations performed by the M.sub.edit trainer 212 to train the fine-tuned second set of layers M′.sub.k+1 . . . L 124 and the low-rank adaptation module Δ”. …Teaches in par. 0017 the following: “the edited machine learning model, which receives the interim output classified as data to unlearn, to not output the data to unlearn so that the machine learning model as a whole is trained to not output data that needs to be unlearned.”.).
The system of Cai does not expressly teach:
generating a noisy feature vector by adding noise to the feature vector;
and the second model using the feature vector and the noisy feature vector;
and generating an unlearned model for the to-be-forgotten target data based on the first layer and the fine-tuned model.
In this instance the examiner notes the teachings of prior art reference FANG.
With regards to applicant’s claim limitation element of, “generating a noisy feature vector by adding noise to the feature vector”, FANG teaches: (
PNG
media_image1.png
484
868
media_image1.png
Greyscale
).
With regards to applicant’s claim limitation element of, “and the second model using the feature vector and the noisy feature vector”, FANG teaches as part of his claim 1 the following: (
“step 1, constructing an extended data set: based on the existing clean small sample set of defender for expanding, adding random Gaussian noise to each sample in the sample set, optimizing the clean sample added with noise on the existing target model to obtain the expanded adjacent data set”.
With regards to applicant’s claim limitation element of, “and generating an unlearned model for the to-be-forgotten target data based on the first layer and the fine-tuned model”, FANG teaches in his Abstract the following: “…candidate trigger with outlier L1 norm through outlier detection algorithm, the candidate trigger is considered as the back door trigger of the attacker by the defender, the corresponding label is the target label of the attacker; finally, constructing a forgetting data set, finely adjusting the target model on the forgetting data set, and selecting the proper fine adjustment model as the final target model according to the fine adjustment effect.”.
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention was made to implement the teachings of Cai with the teachings of FANG by having their system comprise an enhanced model maintenance process. One would have been motivated to do so to provide a simple and effective means to maintain model data, wherein the enhanced model maintenance process helps facilitate better data integrity and makes it easier to configure model data.
The system of Cai and FANG does not expressly teach:
setting a second layer, which is a target attack layer of the pre-trained model, as a first model,
and duplicating a second layer to generate a duplicated second layer to be set as a second model.
In this instance the examiner notes the teachings of prior art reference LIU.
With regards to applicant’s claim limitation element of, “setting a second layer, which is a target attack layer of the pre-trained model, as a first model”, teaches in par. 004 the following: “The first layer can be pretrained in a pretraining stage, and the first layer and the second layer can be tuned together using virtual adversarial regularization.”.
With regards to applicant’s claim limitation element of, “and duplicating a second layer to generate a duplicated second layer to be set as a second model”, LIU illustrates in figure 4, figure element 402 a second duplicated layer.”.
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention was made to implement the teachings of Cai and FANG with the teachings of LIU by having their system comprise an enhanced model training process. One would have been motivated to do so to provide a simple and effective means to tune models, wherein the enhanced model training process helps facilitate model regulation and makes it easier to configure the model structure.
As to claims 2 and 11, the system of Cai, FANG and LIU as applied claim 1 above teaches model configuring, specifically neither Cai nor FANG teaches a unlearning method of claim 1, wherein the first layer includes a convolution layer, and the second layer includes a fully-connected layer.
In this instance the examiner notes the teachings of prior art reference LIU.
LIU teaches in par. 0019 the following: “an overall architecture of a layered model, including the number of layers, the connectivity of the layers, and the type of operations performed by individual layers.”. Teaches in par. 0037 the following: “After the embedding layers 104 are pretrained, the embedding layers and the task-specific layer 110 can be tuned together in tuning stage 204.”.
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention was made to implement the teachings of Cai and FANG with the teachings of LIU by having their system comprise an enhanced model training process. One would have been motivated to do so to provide a simple and effective means to tune models, wherein the enhanced model training process helps facilitate model regulation and makes it easier to configure model structure.
As to claims 3 and 12, the system of Cai, FANG and LIU as applied claim 1 above teaches model configuring, specifically neither Cai nor FANG teaches a unlearning method of claim 1, wherein the duplicating the second layer includes: duplicating the second layer for each epoch to set the duplicated second layer as the second model.
In this instance the examiner notes the teachings of prior art reference LIU.
LIU teaches in par. 0038 the following: “One specific tuning approach is to use a stochastic gradient descent technique to train the parameters of the embedding layers 104 and the task-specific layer 110. In each training iteration or epoch, batch selector 214 selects a batch of labeled task-specific data from labeled task-specific examples 216, and outputs these as labeled training samples 218.”.
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention was made to implement the teachings of Cai and FANG with the teachings of LIU by having their system comprise an enhanced model training process. One would have been motivated to do so to provide a simple and effective means to tune models, wherein the enhanced model training process helps facilitate model regulation and makes it easier to configure model structure.
As to claims 4 and 13, the system of Cai, FANG and LIU as applied claim 1 above teaches model configuring, specifically neither Cai nor FANG teaches a unlearning method of claim 1, wherein one of the first model and the second model includes a student model, and the other includes a teacher model, and wherein the fine-tuning includes: fine-tuning the student model.
In this instance the examiner notes the teachings of prior art reference LIU.
With regards to applicant’s claim limitation element of, “wherein one of the first model and the second model includes a student model”, LIU teaches in par. 0020 the following: “there are many machine learning tasks for which there is a relative lack of training data. One broad approach to training a model with limited task-specific training data for a particular task involves “transfer learning.” In transfer learning, a model is first pretrained on another task for which significant training data is available,”.
With regards to applicant’s claim limitation element of, “and the other includes a teacher model, and wherein the fine-tuning includes: fine-tuning the student model”, LIU teaches in par. 0039 the following: “where the model parameters are tuned to obtain an updated model 226 which is then output to the model history 212. The next tuning iteration can proceed by retrieving the previous model 228 from the model history 212 and continuing with tuning iterations until a stopping condition is reached, e.g., the model converges, achieves a threshold accuracy on a test data set, a training budget is exhausted, and or all the labeled task-specific examples 216 have been exhausted.”.
Thus, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention was made to implement the teachings of Cai and FANG with the teachings of LIU by having their system comprise an enhanced model training process. One would have been motivated to do so to provide a simple and effective means to tune models, wherein the enhanced model training process helps facilitate model regulation and makes it easier to configure model structure.
Allowable Subject Matter
Claims 5 and 14 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter: Applicant’s recital of, “a first step of calculating a gradient for a class of the feature vector itself after passing the feature vector through one of the first model and the second model; a second step of generating the noisy feature vector by multiplying a small value epsilon by a sign of the gradient and adding a result of a multiplication to the feature vector; and a third step of generating the noisy feature vector with noise finally added by repeating the first step and the second step.”.
Dependent claims 6 and 15 are allowed by way of their respective dependency on dependent claims 5 and 14.
Claims 7 and 16 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
The following is a statement of reasons for the indication of allowable subject matter: Applicant’s recital of, “generating a first logit by passing the feature vector through the first model; generating a second logit by passing the noisy feature vector through the second model; generating a smoothed second logit by smoothing the second logit using a softmaxfunction; calculating a first loss value from the first logit and the smoothed second logit using a distillation loss function; calculating a second loss value from the first logit and the second logit using a cross- entropy loss function; calculating a KD loss value, which is a final loss value, from the first loss value and the second loss value using a knowledge distillation (KD) loss function; and fine-tuning the second model based on the KD loss value.”.
Dependent claims 8 and 9 are allowed by way of their dependency on dependent claim 7. Dependent claims 17 and 18 are allowed by way of their dependency on dependent claim 16.
Art Made of Record
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure: Mishra et al. (US Patent No. 12,456,020).
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRYAN F WRIGHT whose telephone number is (571)270-3826.
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, Eleni Shiferaw can be reached on (571)272-3867. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BRYAN F WRIGHT/ Examiner, Art Unit 2497
Prosecution Insights