FINAL REJECTION, SECOND DETAILED ACTION
Status of Prosecution
The present application, 18/332,589 filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
The application was filed in the Office on June 9, 2023 and claims priority to Indian application IN202311031350, filed on May 2, 2023.
The Office mailed a first detailed action, non-final rejection on April 15, 2026.
Applicant initiated an interview on May 28, 2026. Subsequent amendments and remarks and arguments were filed on June 5, 2026, the subject of the instant action.
Claims 1-20 are pending. Claims 1, 17 and 20 are independent.
Status of Claims
Claims 1-3, 11-18 and 20 are rejected under 35 U.S.C. § 103 as being unpatentable over Givental et al., (“Givental”) United States Patent Application Publication 2021/0264025A1, published on Aug. 26, 2021 in view of Shafran et al. (“Shafran”), United States Patent Application Publication 2024/0338611, published on Oct. 10, 2024 in further view of Woo et al.(“Woo”), United States Patent Application Publication 2025/0139408A1, published on May 1, 2025.
Claims 4-10 and 19 are rejected under 35 U.S.C. § 103 as being unpatentable over Givental in view of Shafran in view of Woo in further view of Abdhi Taghi Abad et al.(“ Abdhi Taghi Abad”), United States Patent Application Publication 2020/0110969, published on April 9, 2020.
Response to Remarks and Arguments
Examiner thanks Applicant’s representatives for the courtesies extended during the May 28, 2026 interview. Examiner also thanks Applicant for the remarks and arguments submitted.
Examiner has considered the amendments and arguments made and has newly rejected the claims as noted below, as necessitated by Applicant’s amendment. Specifically, the application of Shafran et al. (“Shafran”), United States Patent Application Publication 2024/0338611, published on Oct. 10, 2024 teaches the amended claim element of a plurality of models.
The claims stand rejected.
Claims 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.
A.
Claims 1-3, 11-18 and 20 are rejected under 35 U.S.C. § 103 as being unpatentable over Givental et al., (“Givental”) United States Patent Application Publication 2021/0264025A1, published on Aug. 26, 2021 in view of Shafran et al. (“Shafran”), United States Patent Application Publication 2024/0338611, published on Oct. 10, 2024 in further view of Woo et al.(“Woo”), United States Patent Application Publication 2025/0139408A1, published on May 1, 2025.
As to Claim 1, Givental teaches: A computing system comprising:
one or more processors (Givental: par. 0053, processor); and
one or more computer-readable media having thereon computer-executable instructions that are structured such, when executed by the one or more processors (Givental: par. 0053, processor with computer readable storage medium and instructions stored thereon), the computing system would be configured to performa method for configuring compute resources to perform tasks of a given inference task type, the method comprising:
for a given inference task type, for each of a plurality of machine-learning model combinations, (Givental: par. 0046, the models may perform inference tasks; par. 0059-61, a plurality of machine learning models [130] may be used and selected from), estimating 2) an accuracy of the machine-learning model combination in performing tasks of the given inference task type, the accuracy obtained by comparing against a ground truth or a result of a randomly seeded split; (Givental: pars. 0060-61, the accuracy of the classification performance of the models may be measured and considered in selection and generation of the ensemble [190], comparing against ground truth data);
selecting a machine-learning model combination for the given inference task type according to at least one of 1) the estimated compute level of the model combination and 2) the estimated accuracy of the model combination (Givental: par. 0067, a top ML model is selected for inclusion in the ensemble, based on factors including and related to scoring; par. 0028, the model prediction ranking is based on performance factors of the corresponding ML models); and
in response to the selection, configuring an inference component to respond to a task request of the given inference task type by using the selected model combination such that the inference components respond to task requests of the given inference type using the selected model combination (Givental: par. 0083, the selected ML models processes log data to generate a classification output (i.e. task request)).
PNG
media_image1.png
809
757
media_image1.png
Greyscale
Givental may not explicitly teach: each of the plurality of machine-learning models comprising a plurality of machine-learning models.
Shafran teaches in general concepts related to detecting errors by training a plurality of models and then selecting from subsets of models a final ensemble for use (Shafran: Abstract). Specifically, Shafran teaches that model groups are composed of a plurality of trained models (Shafran: Fig. 3, pars. 0076-77, the trained models [310(X)] are used to create model groups [312(Y)]). A best ensemble is selected from these different groups of models based on an evaluation metric (Shafran: par. 0081).
PNG
media_image2.png
676
504
media_image2.png
Greyscale
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Givental disclosures and teachings by allowing for the groups of models instead of individual models for the selection of the best ensemble as taught and suggested by Shafran. Such a person would have been motivated to do so with a reasonable expectation of success to allow for the extra nuance of plurality of models to be used in the selection of best ensembles.
Givental and Shafran may not explicitly teach: for a given inference task type, for each of a plurality of machine-learning model combinations, estimating 1) a compute level for performing tasks of the given inference task type using the machine-learning model combination.
Woo teaches in general concepts related to performing inference operations of machine learning models (Woo: Abstract). Specifically, Woo teaches that for determining which machine learning models to include in combination, the latency and computation resources needed for performing inference operations (i.e. compute level) is considered (Woo: par. 0055).
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Givental-Shafran disclosures and teachings by also considering the compute levels for a given inference task as taught and suggested by Woo. Such a person would have been motivated to do so with a reasonable expectation of success to optimize the performance of inference operations in a prioritized fashion (Woo: par. 0013).
As to Claim 2, Givental, Shafran and Woo teach the elements of claim 1.
Woo further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to configure the inference component with compute resources to perform the compute level within a given latency for a given rate of task requests of the given inference task type (Woo: par. 0065, the sub models used can be scheduled with latency considerations, allowing the task to be performed within the restrictions).
As to Claim 3, Givental, Shafran and Woo teach the elements of claim 2.
Woo further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to select the compute resources that would perform the compute level within the given latency for the given rate of task requests of the given inference task type (Woo: par. 0065, Examiner asserts the scheduling of the computer resources is a selection of the resources as claimed).
As to Claim 11, Givental, Shafran and Woo teach the elements of claim 1.
Woo further teaches: the compute level being expressed as a time that is sufficient for a given compute power to perform tasks of the given inference type (Woo: par. 0006, a “priority time period” for performing the priority inference operations).
As to Claim 12, Givental, Shafran and Woo teach the elements of claim 1.
Woo further teaches: the compute level being expressed as a compute power that is sufficient to perform tasks of the given inference type within a given time (Woo: par. 0035, the amount of time to perform the inference operations),.
As to Claim 13, Givental, Shafran and Woo teach the elements of claim 1.
Givental further teaches: the plurality of combinations being restricted to combinations having no more than a maximum number of machine-learning models (Givental: par. 0031, a number of machine learning models, which inherently has a maximum).
As to Claim 14, Givental, Shafran and Woo teach the elements of claim 13.
Givental further teaches: the maximum number of machine-learning models being four (Examiner asserts that this is merely design choice to select four).
As to Claim 15, Givental, Shafran and Woo teach the elements of claim 13.
Givental further teaches: the maximum number of machine-learning models being three (Examiner asserts that this is merely design choice to select three).
As to Claim 16, Givental, Shafran and Woo teach the elements of claim 1.
Givental further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to do the following for each of at least some of the plurality of machine-learning model combinations:
for each of a plurality of aggregation methods for the machine-learning model combination, estimate an accuracy of the machine-learning model combination in performing tasks of the given inference task type with the aggregation method (Givental: par. 0066, the combination of the models’ results, the classifications may be different and are selected based on a suitable function to generate an overall confident result),
the selection of the machine-learning model combination for the given inference task type also selecting the aggregation method for the machine-learning model combination (Givental: par. 0066, similarly, this would be applicable).
As to Claim 17, Givental, Shafran and Woo teach the elements of claim 1.
Givental further teaches: the accuracy of the machine-learning model combination in performing tasks of the given inference task type comprising splitting input data into a first subset of input data and a second subset of input data (Givental: par. 0027, the training dataset may be split into two subsets), the accuracy of the model combination being measured by a conformity between the output of the model combination when provided with the first subset of the input data and the output of the machine-learning model combination when provided with the second subset of the input data (Givental: pars. 0028-29, loss functions are calculated for each subset to determine the conformity and thus the accuracy).
As to Claim 18, it is rejected for similar reasons as claim 1.
As to Claim 20, it is rejected for similar reasons as claim 1.
B.
Claims 4-10 and 19 are rejected under 35 U.S.C. § 103 as being unpatentable over Givental et al., (“Givental”) United States Patent Application Publication 2021/0264025A1, published on Aug. 26, 2021 in view of Woo et al.(“Woo”), United States Patent Application Publication 2025/0139408A1, published on May 1, 2025 in view of Shafran et al. (“Shafran”), United States Patent Application Publication 2024/0338611, published on Oct. 10, 2024 in further view of Abdhi Taghi Abad et al.(“ Abdhi Taghi Abad”), United States Patent Application Publication 2020/0110969, published on April 9, 2020.
As to Claim 4, Givental, Shafran and Woo teach the elements of claim 2.
Givental, Shafran and Woo may not explicitly teach: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to:
determine a batch size of the task requests based on the compute resources.
Abdhi Taghi Abad teaches in general concepts related to adjusting the combination othe samples in a training batch or training set (Abdhi Taghi Abad: Abstract). Specifically Abdhi Taghi Abad teaches that training batch sizes are determined based on the demands on the computing resources available (Abdhi Taghi Abad: par. 0044).
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Givental-Woo disclosures and teachings by considering the batch size of the task requests similarly to the training sets of Abdhi Taghi Abad. Such a person would have been motivated to do so with a reasonable expectation of success to capture the essence of making dynamic adjustments to allow for smooth operations.
As to Claim 5, Givental, Shafran, Woo and Abdhi Taghi Abad teach the elements of claim 4.
Woo and Abdhi Taghi Abad as combined further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to select the compute resources that would perform the compute level within the given latency for the given rate of task requests of the given inference task type and with a given ability of the compute resources to handle batching (Woo: par. 0065, the sub models used can be scheduled with latency considerations, allowing the task to be performed within the restrictions; Abdhi Taghi Abad: par. 0044, the batching is also considered for the resources).
As to Claim 6, Givental, Shafran, Woo and Abdhi Taghi Abad teach the elements of claim 4.
Woo as combined further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to:
respond to receiving a plurality of task requests to perform tasks of the given inference type by scheduling the plurality of task requests with each machine-learning model of the selected machine-learning model combination (Woo: par. 0065, the sub models used can be scheduled with latency considerations, allowing the task to be performed within the restrictions).
As to Claim 7, Givental, Woo, Shafran and Abdhi Taghi Abad teach the elements of claim 6.
Abdhi Taghi Abad further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to perform scheduling by:
accumulating task requests, and while accumulating requests determining whether the accumulated task requests have reached the determined batch size, and determining whether further accumulation of task requests would increase latency of any task request at risk of exceeding the given latency (Abdhi Taghi Abad: par. 0044, the batching is also considered for the resources. Examiner asserts that the computing resources that would be considered would also consider the increase of latency).
As to Claim 8, Givental, Woo, Shafran and Abdhi Taghi Abad teach the elements of claim 7.
Abdhi Taghi Abad further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to perform scheduling by:
if determining either that the accumulated task requests have reached the determined batch size, or that further accumulation of task requests would increase latency of any task request at risk of exceeding the given latency, submitting the accumulated task requests to the model combination (Abdhi Taghi Abad: par. 0044).
As to Claim 9, Givental, Woo, Shafran and Abdhi Taghi Abad teach the elements of claim 8.
Abdhi Taghi Abad further teaches: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to perform scheduling by:
reordering the task requests to group according to groups based on input size (Abdhi Taghi Abad: par. 0014, the combination of the samples in the training batches may be adjusted accordingly based on the need to improve the effectiveness and accuracy of the model, which would include the input size).
As to Claim 10, Givental, Shafran and Woo teach the elements of claim 1.
Givental and Woo may not explicitly teach: the computer-executable instructions being structured such that, when executed by the one or more processors, the computing system would be adapted to:
reorder the task requests to group according to groups based on input size.
Abdhi Taghi Abad teaches in general concepts related to adjusting the combination othe samples in a training batch or training set (Abdhi Taghi Abad: Abstract). Specifically Abdhi Taghi Abad teaches that training batch sizes are determined based on the demands on the computing resources available (Abdhi Taghi Abad: par. 0044). The groups of training batches may be combined in different ways (i.e. grouped) (Abdhi Taghi Abad: par. 0014).
It would have been obvious to a person having ordinary skill in the art at a time before the effective filing date of the application to have modified the Givental-Woo disclosures and teachings by considering the input size of the task requests in grouped fashion similarly to the training sets of Abdhi Taghi Abad. Such a person would have been motivated to do so with a reasonable expectation of success to capture the essence of making dynamic adjustments to allow for smooth operations.
As to Claim 19, it is rejected for similar reasons as claim 7.
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.
Additional relevant prior art made of the record:
Zhan et al., US PG Publication 2020/0090073 (Mar. 19, 2020) (describing generating a machine learning model with model parameter combinations);
Lee et al., US PG Publication 2024/0143976 (May 2, 2024) (describing accuracy and scoring of multiple models).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JAMES T TSAI whose telephone number is (571)270-3916. The examiner can normally be reached M-F 8-5 Eastern.
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, Viker Lamardo can be reached at 571-270-5871. 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.
/JAMES T TSAI/ Primary Examiner, Art Unit 2147
Prosecution Insights