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
This action is in reply to the amendments and remarks filed on 06/25/2026.
Claims 1-19 and 21 are pending.
Claims 1-3, 6-11, and 13-19 have been amended.
Claim 20 has been canceled.
Claim 21 has been added.
Response to Arguments
Applicant’s arguments, with respect to the rejection(s) of claim(s) 1-19 and 21 under 35 U.S.C. 101, have been considered but they are not persuasive. The applicant argues that the amended claims “are directed to a particular improvement in the capabilities of a computing device…[and] a practical application of an abstract idea” since restructuring of network modules enable a “single backbone network to be reused across different application scenarios without reconstruction”; and therefore, overcome the 101 rejections. The examiner respectfully disagrees.
The additional elements and use in the claim do not operate to overcome the previous 101 abstract idea rejection since they remain recited at a high level and can be performed in the human mind. The amended operations of training a neural network is deemed as adding the words “apply it” (or an equivalent, i.e., “training”) 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). See 35 U.S.C 101 section for full, updated analysis of claim limitations necessitated by applicant amendments.
Applicant’s arguments, with respect to the rejection(s) of claim(s) 1 and 9 under 35 U.S.C. 103, have been considered but they are not persuasive. Applicant argues that no art of reference teaches the amended claim language of claims 1 and 9 that now states “obtaining a network model based on the to-be-determined network and the information about the at least one sampling structure, wherein the information about the at least one sampling structure determines a structure of the at least one placeholder module by converting the structure of the at least one placeholder module from an initial or empty structure into a structure corresponding to the information about the at least one sampling structure, and wherein the structure corresponding to the information about the at least one sampling structure is defined by one or more structural parameters of the at least one sampling structure in the candidate set”, since “Chen does not discuss a placeholder module or converting an initial or empty structure. Instead, Chen only states that its networks P are initialized randomly and that each network P consists of its architecture P.8 and its fitness P.f”. The examiner respectfully disagrees due to the broadness of the claim language.
Chen, sections 3.2-3.4 and 4 teach performing a block search for the initialized models and changing/replacing aspects of the model, such as kernel size, branches, blocks, etc. (placeholder modules); since “a population of networks P is initialized randomly. Each individual P consists of its architecture P.θ and its fitness P.f. Any architecture against the constraint η would be removed and a substitute would be picked” from the “search space[s]” to be included in the model; then training the model for adjusting parameters of the searched blocks to model settings (converting an initial or empty structure as argued) according to the model layer type needed and performing mutation and crossover of models to obtain “next generation networks”.
Applicant is encouraged to add clarifying amendments to the claim language cannot be read as broadly. See 35 U.S.C 103 section for full mapping of claim limitations necessitated by applicant amendments.
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 1-19 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claims 1, 9, and 17 are respectively drawn to a system, method, and non-transitory computer readable storage medium, hence each falls under one of four categories of statutory subject matter (Step 1). Nonetheless, the claims are directed to a judicially recognized exception of an abstract idea without significantly more.
Claims 1, 9, and 17 recite the following, or analogous, limitations “obtaining an initial backbone network and a candidate set; replacing at least one basic unit in the initial backbone network with at least one placeholder module to obtain a to-be-determined network, wherein the candidate set comprises parameters of a plurality of structures corresponding to the at least one placeholder module; performing sampling based on the candidate set to obtain information about at least one sampling structure; obtaining a network model based on the to-be-determined network and the information about the at least one sampling structure, wherein the information about the at least one sampling structure determines a structure of the at least one placeholder by converting the structure of the at least one placeholder module from an initial or empty structure into a structure corresponding to the information about the at least one sampling structure, and wherein the structure corresponding to the information about the at least one sampling structure is defined by one or more structural parameters of the at least one sampling structure in the candidate set; and applying, when the network model meets a preset condition, the network model as a target…network”. These limitations, as claimed, under its broadest reasonable interpretation, can be evaluated in a human mind, except for the recitation of generic computer components (using artificial intelligence/machine learning, a computer including one or more microprocessors, and a non-transitory computer readable storage medium) (Step 2A). Other than reciting “target neural network”, “memory”, “a processor”, “a non-transitory computer-readable storage medium”, and “training, by one or more processors of a neural network construction device, the target neural network based on a preset data set to obtain a trained target neural network” to perform the exceptions, nothing in the claims preclude the steps from practically being performed in the human mind. For example, a human expert can:
mentally/with the aid of pen and paper obtaining an initial backbone network and a candidate set (e.g. by thinking of/writing out remembering a template calculation and dataset of variables),
mentally/with the aid of pen and paper replacing at least one basic unit in the initial backbone network with at least one placeholder module to obtain a to-be-determined network, wherein the candidate set comprises parameters of a plurality of structures corresponding to the at least one placeholder module (e.g. by thinking of/writing out changing a variable in the calculation to create an updated calculation, and the dataset includes different variable sets corresponding to values),
mentally/with the aid of pen and paper performing sampling based on the candidate set to obtain information about at least one sampling structure (e.g. by thinking of/writing out selecting from the variable dataset to determine possible calculation architectures),
mentally/with the aid of pen and paper obtaining a network model based on the to-be-determined network and the information about the at least one sampling structure, wherein the information about the at least one sampling structure determines a structure of the at least one placeholder by converting the structure of the at least one placeholder module from an initial or empty structure into a structure corresponding to the information about the at least one sampling structure, and wherein the structure corresponding to the information about the at least one sampling structure is defined by one or more structural parameters of the at least one sampling structure in the candidate set (e.g. by thinking of/writing out an updated calculation architecture from the updated calculation and possible calculation architectures, including placement of the chosen variable by adjusting the number of variables in the calculation from a first number according to a number of variables parameter),
mentally/with the aid of pen and paper applying, when the network model meets a preset condition, the network model as a target…network (e.g. by thinking of/writing out the updated calculation architecture being greater than a predetermined criteria value, and deploying the updated calculation architecture as the final model for use).
Thus, the claims recite a mental process (Step 2A, Prong 1).
Claims 1, 9, and 17 include additional elements, “target neural network”, “memory”, “a processor”, “a non-transitory computer-readable storage medium”, “training, by one or more processors of a neural network construction device, the target neural network based on a preset data set to obtain a trained target neural network”, however the recitations of these elements are at a high level of generality, and amount to adding the words “apply it” (or an equivalent) with the judicial exception (i.e., “training…the target neural network based on a preset data set to obtain a trained target neural network”), or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea (i.e., “memory”, “a processor”, “processors”, and “a non-transitory computer-readable storage medium”) (see MPEP 2106.05(f)); and generally linking the user of the judicial exception to a particular technological environment or field of use (i.e., “target neural network”, “neural network construction device”) (see MPEP 2106.05(h)). Hence, each of the additional limitations or in combination do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea (Step 2A, Prong 2; see MPEP 2106.05(f)). The additional elements in the claim do not amount to significantly more than an abstract idea. Furthermore, 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 the integration of the abstract idea into a practical application, the additional elements of using “target neural network”, “memory”, “a processor”, and “a non-transitory computer-readable storage medium” to perform the steps of the independent claims amounts to no more than mere 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; and generally linking the user of the judicial exception to a particular technological environment or field of use; as these cannot provide an inventive concept. (STEP 2B). As such, claims 1, 9, and 17 are not patent eligible.
Dependent claims 2-8, 10-16, and 18-19 are also ineligible for the same reasons given with respect to claims 1, 10, and 17. The dependent claims describe additional mental processes:
mentally/with the aid of pen and paper wherein after obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure, the method further comprises: performing, when the network model does not meet the preset condition, resampling based on the candidate set; updating the information about the at least one sampling structure based on the resampling to obtain updated information; and updating the network model based on the updated information (claims 2, 10, and 18) (e.g. by mentally/writing out the updated calculation architecture being less than a predetermined criteria value, reselecting from the dataset variables, and updating the calculation architecture)
mentally/with the aid of pen and paper wherein before performing sampling based on the candidate set to obtain the information about the at least one sampling structure, the method comprises constructing a parameter space based on the candidate set, wherein the parameter space comprises architecture parameters corresponding to the parameters of the plurality of structures, and wherein performing sampling based on the candidate set to obtain the information about the at least one sampling structure comprises performing sampling on the parameter space to obtain at least one group of sampling parameters corresponding to the at least one sampling structure (claims 3, 11, and 19) (e.g. by mentally/writing out a collection of dataset variables corresponding to values and selecting a subset of relevant architecture variables for updating the calculation architecture)
mentally/with the aid of pen and paper wherein obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure comprises converting the structure of the at least one placeholder module in the to-be-determined network based on the at least one group of sampling parameters in order to obtain the network model (claims 4 and 12) (e.g. by mentally/writing out the updated calculation architecture from the updated calculation and possible calculation architectures, including placement of the chosen variable conformed to the selected variables)
mentally/with the aid of pen and paper wherein before obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure, the method further comprises constructing the plurality of structures based on the candidate set and the to-be-determined network, wherein the plurality of structures forms a structure search space, and wherein obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure comprises searching the network model from the structure search space based on the at least one group of sampling parameters (claims 5 and 13) (e.g. by thinking of/writing out multiple updated architectures for the calculation from the selected variables in the collected dataset)
mentally/with the aid of pen and paper wherein the preset condition comprises a quantity of times of obtaining the network model exceeds a preset quantity of times (claims 6 and 14) (e.g. by thinking of/writing out the predetermined criteria value adjustment is made to create a new calculation compared an expected value)
mentally/with the aid of pen and paper wherein the candidate set comprises a type of an operator (claims 7 and 15) (e.g. by thinking of/writing out the selected variables include types of mathematical functions utilized by the calculation)
mentally/with the aid of pen and paper wherein the target…network is for performing picture recognition (claims 8 and 16) (e.g. by thinking of/writing out the calculation outputs a label regarding an input image)
Again, the dependent claims continued to cover the performance of the limitation in the mind as inherited from the independent claims (Step 2A, Prong 1). The dependent claims 8 and 16 recitation of “target neural network” is again recited at a high level and amounts to generally linking the user of the judicial exception to a particular technological environment or field of use, and do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea (Step 2A, Prong 2). The additional element in the claims do not amount to significantly more than an abstract idea. As discussed above with respect to the integration of the abstract idea into a practical application, the additional elements to perform the steps of in the dependent claims and perform the steps of the claims amount to no more than generally linking the user of the judicial exception to a particular technological environment or field of use, and this cannot provide an inventive concept. (STEP 2B). As such, dependent claims 2-8, 10-16, and 18-20 additional elements or combination of elements do not amount to significantly more than an abstract idea nor provide any inventive concept, nor impose a meaningful limit to integrate the elements into a practical application or significantly more than the judicial exceptions; therefore, the dependent claims are not deemed patent eligible.
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.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-19 and 21 are rejected under 35 U.S.C. 103 as being unpatentable over Chen et al, (“DetNAS: Backbone Search for Object Detection”, 2019), hereinafter Chen, in view of Hua et al, (US Pub 20200257961), hereinafter Hua.
Regarding claims 1, 9, and 17, Chen teaches a method; a neural network construction apparatus, comprising: a memory configured to store program instructions; and a processor coupled to the memory and configured to execute the program instructions to cause the neural network construction apparatus to; and a non-transitory computer-readable storage medium storing a computer program, that when executed by a processor (sections 3.1-3.2 and 4 teach performing the embodiments of the disclosure on one or more “GPU” communicatively coupled to one or more “memory” known to be included in a computer system), cause an apparatus to:
obtaining an initial backbone network and a candidate set (sections 1 and 3.2-3.3 teach a “backbone network, DetNASNet” and performing an object detection in search spaces of blocks);
replacing at least one basic unit in the initial backbone network with at least one placeholder module to obtain a to-be-determined network, wherein the candidate set comprises parameters of a plurality of structures corresponding to the at least one placeholder module (sections 1 and 3.2-3.3 teach block search spaces including corresponding block “numbers”, and for example “[f]or each block to search, there are 4 choices developed from the original ShuffleNetv2 block: changing the kernel size with {3x3, 5x5, 7x7} or replacing the right branch with an Xception block (three repeated separable depthwise 3x3 convolutions)”);
performing sampling based on the candidate set to obtain information about at least one sampling structure (sections 3.2-3.3 teach “we need to recompute batch statistics for each single path (child networks) before each evaluation” performing the operational blocks);
obtaining a network model based on the to-be-determined network and the information about the at least one sampling structure, wherein the information about the at least one sampling structure determines a structure of the at least one placeholder module by converting the structure of the at least one placeholder module from an initial or empty structure into a structure corresponding to the information about the at least one sampling structure, and wherein the structure corresponding to the information about the at least one sampling structure is defined by one or more structural parameters of the at least one sampling structure in the candidate set (sections 3.3-3.4 and 4 teach “a population of networks P is initialized randomly. Each individual P consists of its architecture P.θ and its fitness P.f. Any architecture against the constraint η would be removed and a substitute would be picked” from the “search space[s]” to be included in the model; then training the model for adjusting parameters of the searched blocks to model settings (converting the structure of the at least one placeholder module from an initial or empty structure into a structure corresponding to the information about the at least one sampling structure) according to the model layer type needed and performing mutation and crossover of models to obtain “next generation networks”);
applying, when the network model meets a preset condition, the network model as a target neural network (sections 3.4 and 4 teach “Among these evaluated networks, we select the top |P| as parents to generate child networks. The next generation networks are generated by mutation and crossover half by half under the constraint η. By repeating this process in iterations, we can find a single path θbest with the best validation accuracy or fitness, fbest”); and
training, by one or more processors of a neural network construction device, the target neural network based on a preset data set to obtain a trained target neural network (sections 3.2, 3.4, and 4 teach “To train the one-shot supernet backbone on ImageNet, we use a batch size of 1024 on 8 GPUs for 300k iterations” on a “classification dataset” for tuning the child networks within).
Chen at least implies obtaining a network model based on the to-be-determined network and the information about the at least one sampling structure, wherein the information about the at least one sampling structure determines a structure of the at least one placeholder module by converting the structure of the at least one placeholder module from an initial or empty structure into a structure corresponding to the information about the at least one sampling structure, and wherein the structure corresponding to the information about the at least one sampling structure is defined by one or more structural parameters of the at least one sampling structure in the candidate set; however Hua teaches obtaining a network model based on the to-be-determined network and the information about the at least one sampling structure, wherein the information about the at least one sampling structure determines a structure of the at least one placeholder module by converting the structure of the at least one placeholder module from an initial or empty structure into a structure corresponding to the information about the at least one sampling structure, and wherein the structure corresponding to the information about the at least one sampling structure is defined by one or more structural parameters of the at least one sampling structure in the candidate set (paragraphs 0022, 0029, and 0032-0035 teaches neural network construction using a search space for specifying and implementing candidate cells in a template network in accordance with the set of prediction parameters and training the network for analyzing performance of the cell(s)).
Thus it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to implement Hua’s teachings of candidate cell search and implementation in a template neural network and updating a first and second deep neural network into Chen’s teaching of backbone network editing with search space blocks and model training in order to define “a specific technical implementation which is more computationally efficient…than existing systems that rely on evaluating the performance of a large number of candidate cells by actually training a network having the candidate cell” (Hua, paragraphs 0040).
Regarding claims 2, 10, and 18, The combination of Chen and Hua teach all the claim limitations of claims 1, 9, and 17 above; and further teach wherein after obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure, the method further comprises: performing, when the network model does not meet the preset condition, resampling based on the candidate set; updating the information about the at least one sampling structure based on the resampling to obtain updated information; and updating the network model based on the updated information (Chen, section 3.4, page 12, and Algorithm 2 teach generating models and iteratively calculating the fitness of generated models to meet a determined architecture “constraint”).
Regarding claims 3, 11, and 19, The combination of Chen and Hua teach all the claim limitations of claims 1, 9, and 17 above; and further teach wherein before performing sampling based on the candidate set to obtain the information about the at least one sampling structure, the method comprises constructing a parameter space based on the candidate set, wherein the parameter space comprises architecture parameters corresponding to the parameters of the plurality of structures, and wherein performing sampling based on the candidate set to obtain the information about the at least one sampling structure comprises performing sampling on the parameter space to obtain at least one group of sampling parameters corresponding to the at least one sampling structure (Chen, section 3.4 teaches “At first, a population of networks P is initialized randomly. Each individual P consists of its architecture P.θ and its fitness P.f. Any architecture against the constraint η would be removed and a substitute would be picked”).
Regarding claims 4 and 12, The combination of Chen and Hua teach all the claim limitations of claims 3 and 11 above; and further teach wherein obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure comprises converting the structure of the at least one placeholder module in the to-be-determined network based on the at least one group of sampling parameters in order to obtain the network model (Chen, sections 3.3-3.4 and 5.3 teach conforming searched blocks to model settings and performing mutation and crossover of models to obtain “next generation networks”).
Regarding claims 5 and 13, The combination of Chen and Hua teach all the claim limitations of claims 3 and 11 above; and further teach wherein before obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure, the method further comprises constructing the plurality of structures based on the candidate set and the to-be-determined network, wherein the plurality of structures forms a structure search space, and wherein obtaining the network model based on the to-be-determined network and the information about the at least one sampling structure comprises searching the network model from the structure search space based on the at least one group of sampling parameters (Chen, sections 3.3-3.4 teach “At first, a population of networks P is initialized randomly. Each individual P consists of its architecture P.θ and its fitness P.f. Any architecture against the constraint η would be removed and a substitute would be picked”; and when picking the substitute block structure from the “search space[s]” to be included in the model).
Regarding claims 6 and 14, The combination of Chen and Hua teach all the claim limitations of claims 1 and 9 above; and further teach wherein the preset condition comprises a quantity of times of obtaining the network model exceeds a preset quantity of times (Chen, section 4 teaches “For evolutionary search, the evolution process is repeated for 20 iterations. The population size |P| is 50 and the parents size |P| is 10. Thus, there are 1000 networks evaluated in one search” when the initial models are developed.).
Regarding claims 7 and 15, The combination of Chen and Hua teach all the claim limitations of claims 1 and 9 above; and further teach wherein the candidate set comprises a type of an operator (Chen, section 3.3 teaches block search spaces including blocks that correspond to specific network functions (type of an operator): “For each block to search, there are 4 choices developed from the original ShuffleNetv2 block: changing the kernel size with {3_3, 5_5, 7_7} or replacing the right branch with an Xception block (three repeated separable depthwise 3_3 convolutions) (types)”).
Regarding claims 8 and 16, The combination of Chen and Hua teach all the claim limitations of claims 1 and 9 above; and further teach wherein the target neural network is for performing picture recognition (Hua, paragraphs 0022, 0029, 0032-0035, and claim 3 teaches the trained cell neural network used for tasks including “image or video classification”).
Chen and Hua are combinable for the same rationale as set forth above with respect to claims 1 and 9.
Regarding claim 21, The combination of Chen and Hua teach all the claim limitations of claim 1 above; and further teach wherein the preset condition comprises both an accuracy requirement and at least one hardware constraint of a target hardware device, wherein the hardware constraint comprises at least one of a computational latency threshold, a quantity of floating-point operations per second threshold, or a memory usage limit, and wherein the network model meets the preset condition when the network model satisfies both the accuracy requirement and the hardware constraint (Chen, sections 3.4 and 5.1-5.2 teach determining accuracy/fitness of models and their “FLOPs” achievements for further training and use).
Conclusion
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 CLINT MULLINAX whose telephone number is 571-272-3241. The examiner can normally be reached on Mon - Fri 8:00-4:30 PT.
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, Alexey Shmatov can be reached on 571-270-3428. 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.
/C.M./Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights