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 .
This action is responsive to the Amendment filed on 4/30/2026.
Claims 1-9, 11-19 are pending in the case. Claim(s) 10, 20 has/have been cancelled.
Response to Arguments
Applicant’s arguments and amendments with regards to the 35 U.S.C. § 112(b) rejection of claim(s) 9, 10, 19 and 20 have been fully considered and are persuasive. Therefore, the 35 U.S.C. § 112(b) rejection of claim(s) 9, 10, 19 and 20 is respectfully withdrawn.
Applicant's arguments and amendments with regards to the 35 U.S.C. § 101 rejection of claim(s) 1-20 have been fully considered and are persuasive. The 35 U.S.C. § 101 rejection of claim(s) 1-20 is respectfully withdrawn.
Applicant's arguments with respect to 35 U.S.C. § 102 and 103 rejection of claim(s) 1-20 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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, 4, 5, 7, 9, 11, 14, 15, 17, 19, are rejected under 35 U.S.C. 103 as being unpatentable over Fan (US 20240232686 A1), in view of Asif (US 20220101184 A1).
Asif was cited in the IDS dated 5/20/2026
Regarding claim 1, Fan teaches a method, comprising (Fan [4-6] method to generate compressed model):
selecting, by one or more processors, layers from a machine learning model (Fan [5] system may use processors to perform method, Fan [75] layers to be compressed are selected based on compression scheme(s), Fan [76] only layers selected based on selected compression scheme(s) are compressed in model);
generating a compressed machine learning model that corresponds to the machine learning model by compressing the selected layers (Fan [75] selected layers are compressed based on compression scheme(s), Fan [76] only layers selected based on selected compression scheme(s) are compressed in model);
...generate first metadata associated with an operation of the machine learning model ...; ... generate second metadata associated with an operation of the compressed machine learning model ...; ... comparing the first metadata with the second metadata (Fan [67, 68, 71, 79, 83] cost function compares performance metrics (metadata) of compressed vs uncompressed models); and
determining whether a behavior of the compressed machine learning model is within a threshold of a behavior of the machine learning model based on the comparison (Fan [71] parameters specify the amount of acceptable change in performance metrics).
Fan does not specifically teach executing the machine learning model to generate first metadata associated with an operation of the machine learning model; executing the compressed machine learning model to generate second metadata associated with an operation of the compressed machine learning model; in response to a change to at least one of a dataset, a codebase, or a model pipeline associated with the machine learning model, comparing the first metadata with the second metadata; determining whether a behavior of the compressed machine learning model is within a threshold of a behavior of the machine learning model based on the comparison; and outputting a result indicating whether the compressed machine learning model maintains the behavior of the machine learning model.
However Asif teaches generating a compressed machine learning model that corresponds to the machine learning model by compressing the selected layers (Asif [37, 43] compressed student model generated from model (teacher) based on layers selected, selected layers may be based on constraints);
executing the machine learning model to generate first metadata associated with an operation of the machine learning model; executing the compressed machine learning model to generate second metadata associated with an operation of the compressed machine learning model (Asif [41] teacher and student models may be executed and performance information for each model determined);
in response to a change to at least one of ... a model pipeline associated with the machine learning model, comparing the first metadata with the second metadata; determining whether a behavior of the compressed machine learning model is within a threshold of a behavior of the machine learning model based on the comparison (Asif [35, 42, 49, 50, 67, 83] based on receiving performance requirements and/or hardware specifications (change in model pipeline and hardware), compressed model(s) may be generated and executed, once executed performance of student is compared to performance of teacher- to determine whether difference between the two is acceptable based on a threshold) ; and
outputting a result indicating whether the compressed machine learning model maintains the behavior of the machine learning mode (Asif [80] if performance is better then student model replaces teacher model).
It would have been obvious to one of an ordinary skill in the art before the effective filing date of the claimed invention, to have incorporated the concept taught by Asif of executing the machine learning model to generate first metadata associated with an operation of the machine learning model; executing the compressed machine learning model to generate second metadata associated with an operation of the compressed machine learning model; in response to a change to at least one of a dataset, a codebase, or a model pipeline associated with the machine learning model, comparing the first metadata with the second metadata; determining whether a behavior of the compressed machine learning model is within a threshold of a behavior of the machine learning model based on the comparison; and outputting a result indicating whether the compressed machine learning model maintains the behavior of the machine learning mode, into the invention suggested by Fan; since both inventions are directed towards determining performance of a model after layer compression, and incorporating the teaching of Asif into the invention suggested by Fan would provide the added advantage of allowing compressed models to be evaluated as a whole based on current specifications and data instead of relying on separate performance information for compressed portions of a model, and the combination would perform with a reasonable expectation of success (Asif [37, 43, 41, 35, 42, 49, 50, 67, 83, 80]).
Regarding claim 4, Fan and Asif teach the invention as claimed in claim 1 above.
Fan further teaches wherein unselected layers in the machine learning model are unchanged in the compressed machine learning model (Fan [76] only layers selected based on selected compression scheme(s) are compressed).
Regarding claim 5, Fan and Asif teach the invention as claimed in claim 4 above.
Fan does not specifically teach determining the change in the codebase of the machine learning model, wherein the change is at least one of a data ETL (Extract-Transform-Load) change, a library update, a library rollback, a codebase change, a hardware change, a pipeline change, a dataset change or combination thereof, wherein the comparing is performed in response to the change.
However Asif teaches determining the change in the codebase of the machine learning model, wherein the change is ... a hardware change, wherein the comparing is performed in response to the change (Asif [3, 34, 46, 47] process may be because of different hardware specification, Asif [35, 42, 49, 50, 67, 83] based on receiving performance requirements and/or hardware specifications (change in model pipeline), compressed model(s) may be generated and executed, once executed performance of student is compared to performance of teacher).
Regarding claim 7, Fan and Asif teach the invention as claimed in claim 1 above.
Fan further teaches selecting layers from the machine learning model based on a pre-defined rule for a class of the machine learning model and automatically generating the compressed machine learning model based on the layers selected by the rule (Fan [75] selected layers are compressed based on compression scheme(s). Fan [65] compression schemes may each be for specific model portions (class)).
Regarding claim 9, Fan and Asif teach the invention as claimed in claim 1 above.
Fan further teaches wherein wherein determining whether a behavior of the compressed machine learning model is within the threshold of behavior of the machine learning model further comprises determining whether the behavior of the compressed machine learning model is... within a threshold of behavior of the compressed machine learning model prior to ...compression...(Fan [68, 71] parameters specify the amount of acceptable change in performance metrics between compressed and uncompressed models, Fan [5, 6, 28, 29, 34, 85] models may be implemented as software components and layers may be model components).
Fan does not specifically teach prior to the change in ... the model pipeline
However Asif teaches wherein determining whether a behavior of the compressed machine learning model is within the threshold of behavior of the machine learning model further comprises determining whether the behavior of the compressed machine learning model is within the threshold of a-behavior of the machine learning model ...prior to the change in the dataset, the codebase, or the model pipeline (Asif [35, 42, 49, 50, 67, 83] based on receiving performance requirements and/or hardware specifications (change in model pipeline and hardware), compressed model(s) may be generated and executed, once executed performance of student is compared to performance of teacher- to determine whether difference between the two is acceptable based on a threshold).
Claim 11 is directed towards a medium storing instructions similar in scope to the instructions performed by the method of claim 1, and is rejected under the same rationale. Fan further teaches a non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations (Fan [4-6]).
Claim(s) 14, 15, 17, 19, is/are dependent on claim 11 above, is/are directed towards a medium storing instructions similar in scope to the instructions performed by the method of claim(s) 4, 5, 7, 9, respectively, and is/are rejected under the same rationale.
Claims 2, 3, 12, 13 are rejected under 35 U.S.C. 103 as being unpatentable over Fan (US 20240232686 A1) in view of Asif (US 20220101184 A1), and further in view of Fraser “EvoSuite: Automatic Test Suite Generation for Object-Oriented Software” dated 2011.
Fraser was cited in the IDS dated 9/8/2025
Regarding claim 2, Fan and Asif teach the invention as claimed in claim 1 above.
Fan further teaches selected layers that have been compressed in the compressed machine learning model.... comparing the metadata from the machine learning model and the metadata from the compressed machine learning model...(Fan [5, 6, 28, 29, 34, 85] models may be implemented as software components (classes) and layers may be model components (classes), Fan [67, 68, 71, 79, 83] cost function compares performance metrics (metadata) of compressed vs uncompressed model, Fan [71] parameters specify the amount of acceptable change in performance metrics).
Fan does not specifically teach automatically performing unit tests to test the selected layers that have been compressed in the compressed model, wherein comparing the metadata from the machine learning model and the metadata from the compressed machine learning model comprises performing the unit tests on the metadata from the machine learning model and the metadata from the compressed machine learning model
However Fraser teaches automatically performing unit tests to test the selected components that have been ...changed in the modified tool..., wherein comparing the metadata from the ... tool .... and the metadata from the ... modified tool... comprises performing the unit tests on the metadata from the ... tool .... and the metadata from the... modified tool... (Fraser Abstract performing unit testing allows adding small and effective sets of assertions that concisely summarize the current behavior, detect deviations from expected behavior, and to capture the current behavior in order to protect against future defects breaking this behavior, Fraser Introduction unit tests may be performed for objects (components) of modified tool using automation, Fraser Section 3, return values for tool objects and modified tool objects are compared to determine failure).
It would have been obvious to one of an ordinary skill in the art before the effective filing date of the claimed invention, to have incorporated the concept taught by Fraser of automatically performing unit tests to test the selected components that have been ...changed in the modified tool..., wherein comparing the metadata from the ... tool .... and the metadata from the ... modified tool... comprises performing the unit tests on the metadata from the ... tool .... and the metadata from the... modified tool..., into the invention suggested by Fan and Asif; since both inventions are directed towards determining the effect of modifications in components on a tool, and incorporating the teaching of Fraser into the invention suggested by Fan and Asif would provide the added advantage of adding small and effective sets of assertions that concisely summarize the current behavior, detect deviations from expected behavior, and to capture the current behavior in order to protect against future defects breaking this behavior, and the combination would perform with a reasonable expectation of success (Fraser Abstract, Introduction, Section 3).
Regarding claim 3, Fan, Asif and Fraser teach the invention as claimed in claim 2 above. Fan does not specifically teach wherein the unit tests include inner model metric unit tests, output metric unit tests, and/or evolution metric unit tests
However Fraser teaches wherein the unit tests include inner model metric unit tests, output metric unit tests, and/or evolution metric unit tests (Fraser Introduction last para and Sections 2, 3 and 4, tests can evaluate classes, objects, changes in return values and mutation testing, whole test suites can target entire coverage criterion).
Claim(s) 12 and 13 is/are dependent on claim 11 above, is/are directed towards a medium storing instructions similar in scope to the instructions performed by the method of claim(s) 2 and 3 respectively, and is/are rejected under the same rationale.
Claims 6 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Fan (US 20240232686 A1) in view of Asif (US 20220101184 A1), and further in view of Dey (US 20220284293 A1).
Regarding claim 6, Fan and Asif teach the invention as claimed in claim 1 above.
Fan further teaches training ... the compressed model (Fan [78-80] compressed model may be trained).
Fan does not specifically teach ... validating the compressed machine learning model
However Dey teaches ... validating the compressed machine learning model (Dey [60] compressed model may be validated with test data, resulting in betterment of accuracy and latency over ... handcrafted model)
It would have been obvious to one of an ordinary skill in the art before the effective filing date of the claimed invention, to have incorporated the concept taught by Dey of ... validating the compressed machine learning model, into the invention suggested by Fan and Asif; since both inventions are directed towards generating compressed model, and incorporating the teaching of Dey into the invention suggested by Fan and Asif would provide the added advantage of betterment of accuracy and latency over ... handcrafted model, and the combination would perform with a reasonable expectation of success (Dey [60]).
Claim(s) 16 is/are dependent on claim 11 above, is/are directed towards a medium storing instructions similar in scope to the instructions performed by the method of claim(s) 6 respectively, and is/are rejected under the same rationale.
Claims 8 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Fan (US 20240232686 A1) in view of Asif (US 20220101184 A1), and further in view of Xie (US 20190370658 A1).
Regarding claim 8, Fan and Asif teach the invention as claimed in claim 1 above.
Fan does not specifically teach wherein subsequent compression operations that select layers that were previously compressed use the previously compressed layers
However Xie teaches wherein subsequent compression operations that select layers that were previously compressed use the previously compressed layers (Xie [19, 24, 26] model may be compressed by incrementally compressing layers and retaining compressed layers from prior increments that satisfy performance criteria).
It would have been obvious to one of an ordinary skill in the art before the effective filing date of the claimed invention, to have incorporated the concept taught by Xie of wherein subsequent compression operations that select layers that were previously compressed use the previously compressed layers, into the invention suggested by Fan and Asif; since both inventions are directed towards compressing a model by compressing layers, and incorporating the teaching of Xie into the invention suggested by Fan and Asif would provide the added advantage of optimizing performance by retaining compression that previously satisfied performance criteria, and the combination would perform with a reasonable expectation of success (Xie [19, 24, 26]).
Claim(s) 18 is/are dependent on claim 11 above, is/are directed towards a medium storing instructions similar in scope to the instructions performed by the method of claim(s) 8 respectively, and is/are rejected under the same rationale.
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 SANCHITA ROY whose telephone number is (571)272-5310. The examiner can normally be reached Monday-Friday 12-8.
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, Usmaan Saeed can be reached at (571) 272-4046. 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.
SANCHITA ROY
Primary Examiner
Art Unit 2146
/SANCHITA ROY/Primary Examiner, Art Unit 2146
Prosecution Insights