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 .
Status of Claims
Applicant’s submission filed 6/2/23 has been entered. Claims 1-11 are presented for examination.
Claim Objections
Claims 1, 10, 11 objected to because of the following informalities: The claims appear to contain typographical errors. The claims recite: “preforming a first-tier calculation”. Appropriate correction is required.
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 1-11 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
The claims disclose the abstract idea of training a model.
STEP 1
Are the claims directed to a process, machine, manufacture or composition of matter?
The claims are all directed to a statutory category (e.g., a process, machine, manufacture, or composition of matter). The answer is YES.
STEP 2A. Prong 1
Exemplary claim 11 recites the following abstract concepts that are found to include “abstract idea”:
“ A training system for a neural network model, comprising:
---performing a training method so as to train the neural network model, wherein the training method comprises:
-- (a) receiving an image data; (b) preforming a first-tier calculation comprising a first convolution calculation and a first non-linear calculation by using the image data;
(c) performing a combination computation, comprising:
(c1) grouping an output of the first-tier calculation;
--(c2) performing a linear computation and a non-linear computation respectively on different groups of the output of the first-tier calculation; and
--(c3) performing a consolidate computation based on an output of the linear computation and an output of the non-linear computation;
--(d) when a second-tier calculation is determined to be executed, performing the second-tier calculation comprising a second convolution calculation and a second non-linear calculation based on an output of the combination computation; and
--(e) when the second-tier calculation is determined not to be executed, classifying the output of the combination computation;
wherein, the output of the combination computation is generated at least based on the image data after the first-tier calculation and one of a mathematical operator and a non-linear operator.
The remaining limitations are no more than computer elements (i.e., a processor) to be used as a tool to perform this abstract idea.
The recited limitations cover a process that, under its broadest reasonable interpretation, covers subject matter viewed as a certain method of organizing human activity with the additional recitation of generic computer components. For example, but for the “by a processor” language, “performing, receiving, grouping, classifying etc..” in the context of this claim encompasses the user receiving the data performing the calculations, consolidating the results, and classifying the output.
The practice of receiving, grouping, calculating, and outputting data, as well as performing calculation is a commercial or legal interaction long prevalent in our system of commerce. The claims recite the idea of performing various conceptual steps generically resulting in the training/programming a system. As determined earlier, none of these steps recites specific technological implementation details, but instead get to this result by receiving, selecting and determining data. Thus, the claims are directed to a certain method of organizing human activity
STEP 2A, Prong 2
Are there additional elements or a combination of elements in the claim that apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception, such that it is more than a drafting effort designed to monopolize the exception?
The claim recites a processor is used to perform the steps; and performing the linear and non-linear computations.
The processor , the linear and non-linear computations in the steps are recited at a high level of generality, i.e., as a generic processor performing a generic computer function of processing data (receiving, by a processor, image data). This generic processor limitation is no more than mere instructions to apply the exception using a generic computer component.
Accordingly, the additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claim is directed to an abstract idea.
STEP 2B
The next issue is whether the claims provide an inventive concept because the additional elements recited in the claims provide significantly more than the recited judicial exception. Taking the claim elements separately, the function performed by the computer system at each step of the process is purely conventional. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed above with respect to integration of the abstract idea into a practical application, the additional element of using a processor to perform the steps amounts to no more than mere instructions to apply the exception using a generic computer component. Mere instructions to apply an exception using a generic computer component cannot provide an inventive concept. The claim is not patent eligible.
Considered as an ordered combination, the computer components of Applicants' claims add nothing that is not already present when the steps are considered separately. The claimed invention does not focus on an improvement in computers as tools, but rather certain independently abstract ideas that use computers as tools. {Elec. Power, 830 F.3d at 1354). (Step 2B: NO).
There is no indication that indication that the processor is anything other than a generic, off-the-shelf computer component, and the Symantec, TLI, and OIP Techs. Court decisions cited in MPEP 2106.05(d)(II) indicate that mere collection or receipt of data over a network is a well‐understood, routine, conventional function when it is claimed in a merely generic manner (as it is here).
Alternately, this is an example of organizing and manipulating information through mathematical correlations (Digitech).
Independent claims 10, 11 recite similar limitations as claim 1 and are therefore rejected under the same rationale.
The dependent claims when analyzed as a whole are held to be patent ineligible under 35 U.S.C. 101 because the additional recited limitations fail to establish that the claims are not directed to an abstract idea. The claims provide minimal technical structure or components for further consideration either individually or as ordered combinations with the independent claims. As such, additional recited limitations in the dependent claims only refine the identified abstract idea further. Further refinement of an abstract idea does not convert an abstract idea into something concrete.
Accordingly, a conclusion that the collecting step is well-understood, routine, conventional activity is supported under Berkheimer Option 2.
See MPEP 2106.05(d)(II) The courts have recognized the following computer functions as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity.
i. Receiving or transmitting data over a network, e.g., using the Internet to gather data, Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350,1355,112 USPQ2d 1093,1096 (Fed. Cir. 2014) (computer receives and sends information over a network); but see DDR Holdings, LLC v. Hoteis.com, L.P., 773 F.3d 1245, 1258, 113 USPQ2d 1097, 1106 (Fed. Cir. 2014) ("Unlike the claims in Ultramercial, the claims at issue here specify how interactions with the Internet are manipulated to yield a desired result-a result that overrides the routine and conventional sequence of events ordinarily triggered by the click of a hyperlink." (emphasis added));
iv. Storing and retrieving information in memory, Versata Dev. Group, Inc. v. SAP Am., Inc., 793 F.3d 1306,1334,115 USPQ2d 1681,1701 (Fed. Cir. 2015); OIP Techs., 788 F.3d at 1363,115 USPQ2d at 1092-93.
The claims are ineligible.
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 1-11 are rejected under 35 U.S.C. 103 as being unpatentable over BAI et al. US 20180025256 A1, in view of CAO et al. (CN 112766495 A)
Re-claim 1, BAI et al. teach -- A training method for a neural network model, comprising:
(a) receiving an image data;
(see e.g. [0034] During implementation, the server may obtain, in advance, some sample image blocks whose image content includes a character string (for example, an English word) and store them in a training sample library. The sample image blocks whose image content includes a character string may be obtained by combining image content of natural scene image blocks with a preset character string.
(b) preforming a first-tier calculation comprising a first convolution calculation and a first non-linear calculation by using the image data;
(see e.g. [0034] The server may perform processing on image data of a sample image block according to a convolutional neural network algorithm including a to-be-determined parameter, and may obtain an expression of a probability set including a preset quantity. The expression of the probability set includes the plurality of to-be-determined parameters of the convolutional neural network algorithm.
[0036] The convolutional layer includes a to-be-determined parameter of the convolutional neural network model, that is, a convolution kernel and an initial offset parameter. When the convolutional neural network algorithm is used for calculation, a parameter in the convolutional neural network model may be initialized in advance. Initialization of a parameter may be completed randomly, and the initialized parameter may satisfy Gaussian distribution whose average is 0 and whose standard deviation is 0.01. The server may process each pre-stored sample image block based on the convolutional neural network algorithm to obtain an expression, that is, a probability matrix, of a probability set corresponding to each sample image block. That is, the server may perform convolutional processing for a first preset quantity of times, pooling processing for a second preset quantity of times, and linear processing for a third preset quantity of times on each pre-stored sample image block to obtain an expression of a corresponding first feature vector, and then, perform softmax softening processing on the expression of the first feature vector, so as to obtain a probability matrix corresponding to each sample image block.
[0038] . That is, pooling processing (which may be the maximum pooling processing) is separately performed on each convolutional image block and pooling kernels of the first pooling layer, and corresponding first pooling output may be obtained.).
***The Examiner notes pooling is considered non-linear.
BAI et al. do not explicitly teach the following limitations.
However CAO et al. teach (c) performing a combination computation, comprising:
(c1) grouping an output of the first-tier calculation; (c2) performing a linear computation and a non-linear computation respectively on different groups of the output of the first-tier calculation; and
(see e.g. ---obtaining the network parameter of the first neural network layer;
dividing the network parameter into a first parameter set for linear operation, and a second parameter set for nonlinear operation.
--a linear calculation unit, configured to, for the first encryption result, performing homomorphic linear operation, obtaining the first linear calculation result;
In the step, the nonlinear calculation can be performed according to the first linear calculation result and the second parameter set, so as to determine the output result of the first neural network layer.
(c3) performing a consolidate computation based on an output of the linear computation and an output of the non-linear computation;
(see e.g. ---obtaining a first linear calculation result of homomorphic linear operation based on the encryption result from the untrusted execution environment;
sending the first linear calculation result to the TEE, making it according to the first linear calculation result and the second parameter set for nonlinear operation in the first neural network layer, determining the output result of the first neural network layer.
(d) when a second-tier calculation is determined to be executed, performing the second-tier calculation comprising a second convolution calculation and a second non-linear calculation based on an output of the combination computation; and
(see e.g. CAO et al. --Because the first linear calculation result itself is homomorphic encrypted data, so, in one embodiment, can be based on the same-state encryption algorithm corresponding to the key, decrypting the first linear calculation result, obtaining the second linear calculation result; the second linear calculation result is the plaintext linear calculation result obtained according to the input data and linear operation parameter of the neural network layer. then, performing nonlinear calculation according to the second linear calculation result and the second parameter set, determining the output result of the first neural network layer. In one example, the second parameter set comprises a parameter of the activation function of the neuron of the neural network layer, can according to the second linear calculation result and the neuron activation function of the neural network layer, performing nonlinear operation, determining the output result of the first neural network layer.
(e) when the second-tier calculation is determined not to be executed, classifying the output of the combination computation; wherein, the output of the combination computation is generated at least based on the image data after the first-tier calculation and one of a mathematical operator and a non-linear operator.
(see e.g. CAO et al. --Due to the output result of the deep learning model, usually the output result of the last layer in the plurality of neural network layers contained in it. Therefore, according to one embodiment, it also can be the N neural network layer after calculating processing, the output result of the last layer of N neural network layer as the output result of the deep learning mod
In one embodiment, the homomorphic linear operation comprises homomorphic addition operation and or homomorphic multiplication operation.
Therefore, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to modify BAI et al., and include the steps cited above, as taught by CAO et al., in order to protect the privacy of the model operation while improving the efficiency of the model operation (see e.g. CAO et al.).
Re-claims 2, 3 BAI et al. do not teach the limitations as claimed.
However, CAO et al. teach ---The training method according to claim 1, wherein the output of the combination computation is further based on the image data after at least once of the second-tier calculation.
(see e.g. CAO et al. --the second linear calculation result is the plaintext linear calculation result obtained according to the input data and linear operation parameter of the neural network layer).
3. The training method according to claim 1, wherein the second-tier calculation is performed W times, and the training method further comprises:
(see e.g. CAO et al. In another embodiment, it also can be according to a predetermined rule in the TEE, for multiple times loading deep learning model For example, in one example, it can according to the deep learning model order of N neural network layer included, loading N neural network layer multiple times, wherein each loading one or more neural network layer.)
Regarding the following limitation:
(f) determining whether a w+1 time second-tier calculation is performed or not, where w is an integer from 0 to W.
BAI et al. anticipate the limitation by teaching: performing each processing step for a preset quantity of times (see e.g. [0035]). And setting a first preset quantity of times as 5, a second preset quantity of times as 4, a third preset quantity of times as 3. (see e.g. [0037]).
Re-claim 4, BAI et al. teach --The training method according to claim 1, wherein the first convolution calculation is performed N times and an output of the n-1 time first convolution calculation is regarded as the image data for the n time first convolution calculation, where N≥2 and N≥n≥2, wherein the step (c) is performed based on the output of the N time first convolution calculation and an output of the first non-linear calculation.
(see e.g. [0035] performing convolutional processing for a first preset quantity of times,
[0038] after convolutional processing is separately performed on the filled image block and the 96 convolution kernels, output of the first convolutional layer may be obtained, and may be recorded as first convolution output. The first convolution output includes 96 convolutional image blocks, and a size of each convolutional image block is 32*100.)
Re-claim 5, BAI et al. teach --The training method according to claim 1, wherein the first non-linear calculation is performed M times and an output of the m-1 time first non-linear calculation is regarded as the image data for the m time first non-linear calculation, where M≥2 and M≥m≥2, wherein the step (c) is performed based on the output of the M time first non-linear calculation and an output of the first convolution calculation.
(see e.g. [0035] pooling processing for a second preset quantity of times.
[0038] pooling processing (which may be the maximum pooling processing) is separately performed on each convolutional image block and pooling kernels of the first pooling layer, and corresponding first pooling output may be obtained.).
Re-claim 6, BAI et al. teach --The training method according to claim 1, further comprising one or any combination of the following steps: changing a dimension of an output of the first convolution calculation and/or an output of the second convolution calculation; and updating the output of the first convolution calculation and/or the output of the second convolution calculation based on an activation function.
(see e.g. [0042] Subsequently, normalization is performed on the first linear hidden layer output, and a specific process of the normalization is: calculating an average and a variance of the first linear hidden layer output, which may be reducing a numeral of each dimension in the first linear hidden layer output by the average corresponding to the first linear hidden layer output and dividing a result by the variance, that is, normalization may be performed on the first linear hidden layer output.
[0345] FIG. 18 is a flowchart illustrating the method of FIG. 15, with the additional steps of extracting a reduced dimension representation and concatenating the visual data to create a reduced dimension representation of the separated frame;
[0346] FIG. 19 is a flowchart illustrating the method of FIG. 16, with the additional steps of extracting a reduced dimension representation and concatenating the visual data to create a reduced dimension representation of the separated scene).
Re-claim 7, BAI et al. teach --The training method according to claim 1, further comprising one or any combination of the following steps: changing a dimension of an output of the first non-linear calculation and/or an output of the second non-linear calculation; and updating the output of the first non-linear calculation and/or the output of the second non-linear calculation based on an activation function.
(see e.g. [0042] Subsequently, normalization is performed on the first linear hidden layer output, and a specific process of the normalization is: calculating an average and a variance of the first linear hidden layer output, which may be reducing a numeral of each dimension in the first linear hidden layer output by the average corresponding to the first linear hidden layer output and dividing a result by the variance, that is, normalization may be performed on the first linear hidden layer output.)
NOTE: CAO et al. teach --dimension checking unit, configured to: before sending the first linear calculation result to the TEE, the first linear calculation result for dimension checking, determining whether it is a predetermined dimension, determining whether to transmit the first linear calculation result to the TEE according to the checking result.
Re-claim 8, BAI et al. teach --The training method according to claim 1, wherein the mathematical operator comprises one or any combination of the following: addition, subtraction, multiplication, division, and power value.
(see e.g. [0042] A quantity of dimensions of a vector of the first linear hidden layer output is 4096, and processing of the first linear hidden layer is equivalent to multiplying a matrix having a size of 4096*11264 by a feature vector having 11264 dimensions, to obtain the first linear hidden layer output and the vector having a size of 4096.
NOTE: CAO et al. also teach --In one embodiment, the homomorphic linear operation comprises homomorphic addition operation and/or homomorphic multiplication operation.
Re-claim 9, BAI et al. teach --The training method according to claim 1, wherein the non-linear operator comprises one or any combination of the following: acquiring a maximum value, acquiring a minimum value, and acquiring a mean value.
(see e.g. [0042] Subsequently, normalization is performed on the first linear hidden layer output, and a specific process of the normalization is: calculating an average and a variance of the first linear hidden layer output, which may be reducing a numeral of each dimension in the first linear hidden layer output by the average corresponding to the first linear hidden layer output and dividing a result by the variance, that is, normalization may be performed on the first linear hidden layer output.)
Claim 10 recites similar limitations as claim 1 and is therefore rejected under the same arts and rationale.
Claim 11 recites similar limitations as claim 1 and is therefore rejected under the same arts and rationale
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
BORSE et al. (US-20210150347-A1) --Guided training of machine learning models with convolution layer feature data fusion.
WANG (US-20190325621-A1) -Machine Learning for visual processing.
ZHANG (CN-112348161-A) - Training Method Of Neural Network, Training Device Of Neural Network And Electronic Device
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LUNA CHAMPAGNE whose telephone number is (571)272-7177. The examiner can normally be reached M-F 8:00-5:00.
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, Florian Zeender can be reached at 571 272-6790. 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.
/LUNA CHAMPAGNE/Primary Examiner, Art Unit 3627 February 20, 2026