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 .
Allowable Subject Matter
Claims 3 and 6-9 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.
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 11-19 are rejected under 35 U.S.C. 101 because the claims encompass non-statutory subject matter.
The claims recite a “computer readable medium,” which is construed to cover both transitory and non-transitory media under the broadest reasonable interpretation consistent with the specification.1 Transitory, propagating signals per se constitute non-statutory subject matter.2 Because the full scope of the claims encompasses non-statutory subject matter, the claims as a whole are non-statutory.
The claims also recite a “computer readable medium comprising computer executable instructions for reducing machine learning models for target hardware, the instructions for…,” which is a computer program. Computer programs per se are considered to be non-statutory subject matter. See MPEP 2106.01.
It is suggested that claim 11 be amended to recite a “non-transitory computer readable medium comprising computer executable instructions for reducing machine learning models for target hardware that, when executed by one or more processors, cause the one or more processors to perform
Appropriate correction is required.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1, 5, 11, 14, and 20 are rejected under 35 U.S.C. 102(a)(1) and (a)(2) as being anticipated by Jiang et al. (US Pub. 20210397963).
Referring to claim 1, Jiang discloses A computer-implemented method for reducing machine learning models for target hardware, the method comprising:
providing a model, a set of training data, and a training threshold [pars. 59-61 and 77; a neural network model, training dataset, and a gradient threshold (i.e., when a target loss converges) are provided to a training apparatus];
determining a search space for reducing the model with a pruning function and a pruning factor, wherein the pruning function increases compression along a depth of the model, and the compression increases are based on the pruning factor [pars. 68-74; layers (i.e., depth) of the neural network model are compressed individually using a binary mask M (i.e., pruning function) and a pruning ratio p (i.e., pruning factor); for each layer, a matrix (i.e., tensor) representing input/output channels ci with height, weight, and depth axes k1, k2, and k3 is reshaped (e.g., a 5D weight tensor is reshaped into a 2D matrix tensor); within the reshaped tensor (i.e., search space), a set of pruning micro-structures bs is selected for pruning according to the pruning ratio (i.e., the top p% of the blocks in the layer)], by:
bounding the pruning function with two or more constraints [pars. 22 and 68-74; the reshaped tensor is shaped based on desired dimensions (i.e., constraints) within which to select the set of pruning micro-structures that align with underlying hardware design];
determining, based on the two or more constraints, boundaries for the pruning factor, the determined boundaries defining at least in part the search space [pars. 68-74; note the reduced dimensions (i.e., boundaries) of the reshaped tensor based on the desired dimensions, where the set of pruning micro-structures is selected from within the reshaped tensor];
training the model to learn a reduced model by iteratively [pars. 74-77; the training dataset is passed as input to the (pre-trained) neural network model, and the neural network model is re-trained in an iterative process]:
updating model parameters based on the pruning function and the pruning factor and within the search space [pars. 74-77; the set of pruning micro-structures is selected from within the reshaped tensor, and corresponding weights W* and (binary) mask M*, are updated by iteratively minimizing the target loss];
evaluating the updated model based on the set of training data and the training threshold [pars. 74-77; a gradient of the target loss is iteratively calculated based on each input x and estimated output y until the gradient threshold (i.e., when the target loss converges)]; and
providing the reduced model to a target hardware [pars. 21 and 22; the (re-trained) neural network model is deployed on a hardware device (having the underlying hardware design)].
Referring to claim 5, Jiang discloses The method of claim 1, wherein the pruning function is a linear or exponential function [pars. 74 and 75; the binary mask is a linear function (because applying the mask involves multiplying each weight coefficient by 0 or 1, which means that the output is directly proportional to the input)].
Referring to claim 11, see at least the rejection for claim 1. Jiang further discloses A computer readable medium comprising computer executable instructions for reducing machine learning models for target hardware, the instructions for performing the claimed steps [fig. 2, processor 220 and storage component 240 / memory 230].
Referring to claim 14, see the rejection for claim 5.
Referring to claim 20, see at least the rejection for claim 1. Jiang further discloses A device comprising a processor and memory, the memory comprising computer executable instructions for reducing machine learning models for target hardware, the instructions causing the processor to perform the claimed steps [fig. 2, processor 220 and storage component 240 / memory 230].
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.
Claims 2 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Bai et al. (US Pub. 20200210113).
Referring to claim 2, Jiang does not appear to explicitly disclose The method of claim 1, the method comprising determining a granularity of the search space based on a number of searching steps.
However, Bai discloses The method of claim 1, the method comprising determining a granularity of the search space based on a number of searching steps [pars. 44 and 51; in an iterative process, steps within a search space are increased to more extensively explore the search space in a finer granularity].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the neural network model taught by Jiang so that the number of searching steps in the re-training can be increased as taught by Bai, with a reasonable expectation of success. The motivation for doing so would have been to optimize the search (e.g., explore more if the cost/ target loss is still acceptable) [Bai, par. 44].
Referring to claim 12, see the rejection for claim 2.
Claims 4, 10, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Jiang in view of Rao et al. (US Pub. 20230091667).
Referring to claim 4, Jiang does not appear to explicitly disclose The method of claim 1, wherein the training threshold is a target accuracy.
However, Rao discloses The method of claim 1, wherein the training threshold is a target accuracy [pars. 82-84; one or more performance indicators are used to evaluate training of a candidate neural network; examples of the performance indicators include model inference accuracy].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the neural network model taught by Jiang so that the re-training is evaluated based on one or more performance indicators such as model inference accuracy as taught by Rao, with a reasonable expectation of success. The motivation for doing so would have been to ensure that the neural network model satisfies performance constraints [Rao, par. 85].
Referring to claim 10, Jiang does not appear to explicitly disclose The method of claim 1, the method comprising: employing knowledge distillation to train the model for the target hardware for classification tasks.
However, Rao discloses The method of claim 1, the method comprising: employing knowledge distillation to train the model for the target hardware for classification tasks [pars. 12, 40, and 87; a knowledge distillation operation is executed on a second (i.e., smaller) neural network for a specific computer vision task to be performed on an electronic device].
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the neural network model taught by Jiang so that a knowledge distillation operation is executed on the neural network model as taught by Rao, with a reasonable expectation of success. The motivation for doing so would have been to improve the accuracy of the neural network model in performing a specific computer vision task [Rao, par. 87].
Referring to claim 19, see the rejection for claim 10.
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GRACE PARK whose telephone number is (571)270-7727. The examiner can normally be reached M-F 8AM-5PM.
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, TAMARA KYLE can be reached at (571)272-4241. 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.
/Grace Park/Primary Examiner, Art Unit 2144
1 Official Gazette Notice 1351 OG 212, dated February 23, 2010, states “the broadest reasonable interpretation of a claim drawn to a computer readable medium…typically covers forms of non-transitory tangible media and transitory propagating signals per se in view of the ordinary and customary meaning of computer readable media, particularly when the specification is silent.”
2 See In re Nuijten, 500 F.3d 1346, 1356-57 (Fed. Cir. 2007) (transitory embodiments are not directed to statutory subject matter) and Interim Examination Instructions for Evaluating Subject Matter Eligibility Under 35 U.S.C. § 101, Aug. 24, 2009; p. 2.
Prosecution Insights