DETAILED ACTION
This action is in response to the amendment filed 04/16/2026. Claims 1, 3-7, 9-13, 15-19, 21-25, 27-30 are pending and have been examined.
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 .
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1, 7, 13, 19, and 25 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Regarding claim 1, claim 1 recites “the one or more invalid data values are pre-designated.” The specification does not support this. The paragraph that provides the most support is paragraph 0046, which states “this can be accomplished, as least in part, by setting values of remaining pixels when written to a scratchpad, or scratch space, to have special pixel values, or invalid pixel values. In at least one embodiment, these can be special or invalid data values for non-pixel values, such as those that would never by generated by relevant hardware during relevant computations or operations. In at least one embodiment, a special or invalid value in this context refers to a value that would not be produced by an operation to be performed, or determined to be a valid input value for an operation to be performed. In at least one embodiment, this might include a value such as "-0" (or another value that may be a "not a number" (NaN) value and that would not be produced by a typical operation, such as a value considered to be a numeric data type that can be interpreted as a value that is undefined or unrepresentable. In at least one embodiment, this might include a value with a letter for an operation that involves only integers or floating point numerical values, among other such options.” This paragraph supports invalid data values, but does not support that the invalid data values are pre-designated.
Regarding claim 7, claim 7 recites “the one or more invalid data values are pre-designated.” The specification does not support this. The paragraph that provides the most support is paragraph 0046, which states “this can be accomplished, as least in part, by setting values of remaining pixels when written to a scratchpad, or scratch space, to have special pixel values, or invalid pixel values. In at least one embodiment, these can be special or invalid data values for non-pixel values, such as those that would never by generated by relevant hardware during relevant computations or operations. In at least one embodiment, a special or invalid value in this context refers to a value that would not be produced by an operation to be performed, or determined to be a valid input value for an operation to be performed. In at least one embodiment, this might include a value such as "-0" (or another value that may be a "not a number" (NaN) value and that would not be produced by a typical operation, such as a value considered to be a numeric data type that can be interpreted as a value that is undefined or unrepresentable. In at least one embodiment, this might include a value with a letter for an operation that involves only integers or floating point numerical values, among other such options.” This paragraph supports invalid data values, but does not support that the invalid data values are pre-designated.
Regarding claim 13, claim 13 recites “the one or more invalid data values are pre-designated.” The specification does not support this. The paragraph that provides the most support is paragraph 0046, which states “this can be accomplished, as least in part, by setting values of remaining pixels when written to a scratchpad, or scratch space, to have special pixel values, or invalid pixel values. In at least one embodiment, these can be special or invalid data values for non-pixel values, such as those that would never by generated by relevant hardware during relevant computations or operations. In at least one embodiment, a special or invalid value in this context refers to a value that would not be produced by an operation to be performed, or determined to be a valid input value for an operation to be performed. In at least one embodiment, this might include a value such as "-0" (or another value that may be a "not a number" (NaN) value and that would not be produced by a typical operation, such as a value considered to be a numeric data type that can be interpreted as a value that is undefined or unrepresentable. In at least one embodiment, this might include a value with a letter for an operation that involves only integers or floating point numerical values, among other such options.” This paragraph supports invalid data values, but does not support that the invalid data values are pre-designated.
Regarding claim 19, claim 19 recites “the one or more invalid data values are pre-designated.” The specification does not support this. The paragraph that provides the most support is paragraph 0046, which states “this can be accomplished, as least in part, by setting values of remaining pixels when written to a scratchpad, or scratch space, to have special pixel values, or invalid pixel values. In at least one embodiment, these can be special or invalid data values for non-pixel values, such as those that would never by generated by relevant hardware during relevant computations or operations. In at least one embodiment, a special or invalid value in this context refers to a value that would not be produced by an operation to be performed, or determined to be a valid input value for an operation to be performed. In at least one embodiment, this might include a value such as "-0" (or another value that may be a "not a number" (NaN) value and that would not be produced by a typical operation, such as a value considered to be a numeric data type that can be interpreted as a value that is undefined or unrepresentable. In at least one embodiment, this might include a value with a letter for an operation that involves only integers or floating point numerical values, among other such options.” This paragraph supports invalid data values, but does not support that the invalid data values are pre-designated.
Regarding claim 25, claim 25 recites “the one or more invalid data values are pre-designated.” The specification does not support this. The paragraph that provides the most support is paragraph 0046, which states “this can be accomplished, as least in part, by setting values of remaining pixels when written to a scratchpad, or scratch space, to have special pixel values, or invalid pixel values. In at least one embodiment, these can be special or invalid data values for non-pixel values, such as those that would never by generated by relevant hardware during relevant computations or operations. In at least one embodiment, a special or invalid value in this context refers to a value that would not be produced by an operation to be performed, or determined to be a valid input value for an operation to be performed. In at least one embodiment, this might include a value such as "-0" (or another value that may be a "not a number" (NaN) value and that would not be produced by a typical operation, such as a value considered to be a numeric data type that can be interpreted as a value that is undefined or unrepresentable. In at least one embodiment, this might include a value with a letter for an operation that involves only integers or floating point numerical values, among other such options.” This paragraph supports invalid data values, but does not support that the invalid data values are pre-designated.
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, 3-7, 9-13, 15-19, 21-25, 27-30 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1:
Subject Matter Eligibility Analysis Step 1:
Claim 1 recites a processor and is thus an article of manufacture, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 1 recites
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (This limitation could encompass a human mentally replacing a value in their head with an invalid value and is thus an evaluation.)
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated (This limitation could encompass a human mentally replacing a data value in their head with a data value that’s unable to be generated and is thus an evaluation.)
detect the one or more invalid data values retained in the input data provided to the one or more operations (This limitation could encompass a human mentally detecting invalid data values and is thus an evaluation.)
cause the one or more operations… to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (This limitation could encompass a human mentally causing operations to be performed without considering the invalid data values and is thus an evaluation.)
Claim 1 therefore recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 1 further recites additional elements of
A processor, comprising: one or more circuits (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Generated by the one or more circuits during one or more operations of one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
of the one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, Claim 1 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional element of Claim 1 does not provide significantly more than the abstract idea itself, taken alone and in combination, because
A processor, comprising: one or more circuits uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Generated by the one or more circuits during one or more operations of one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
of the one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, Claim 1 is subject-matter ineligible.
Regarding Claim 3:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 3 recites the same mental processes as in claim 1 and, therefore, recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 3 recites additional elements of
wherein the one or more circuits are further to cause a plurality of data values to be loaded into one or more registers, wherein the plurality of data values include one or more valid data values and the one or more invalid data values (This additional element does not integrate the abstract idea into a practical application because it recites insignificant extra-solution activity of data gathering(see MPEP 2106.05(g)).)
Therefore, Claim 3 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
These additional elements of Claim 3 do not provide significantly more than the abstract idea itself, taken alone and in combination, because
wherein the one or more circuits are further to cause a plurality of data values to be loaded into one or more registers, wherein the plurality of data values include one or more valid data values and the one or more invalid data values is the well understood, routine, and conventional activity of “transmitting or receiving data over a network” (see MPEP 2106.05(d)(II); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network)).
Therefore, Claim 3 is subject-matter ineligible.
Regarding Claim 4:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 4 recites
utilize the plurality of data values from one or more registers to perform one or more operations in a sequence (This limitation could encompass a human mentally performing operations with the data values and is thus an evaluation.)
wherein individual operations of the sequence are able to identify the invalid data values (This limitation could encompass a human mentally identifying the invalid data values via the operations and is thus an observation.)
Claim 4 therefore recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 4 recites additional elements of
wherein the one or more circuits are further to utilize data values (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, Claim 4 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional element of Claim 4 does not provide significantly more than the abstract idea itself, taken alone and in combination, because
wherein the one or more circuits are further to utilize data values uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, Claim 4 is subject-matter ineligible.
Regarding Claim 5:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 5 recites
wherein at least one individual operation of the one or more operations is enabled to propagate the invalid data values. (This limitation could encompass a human mentally propagating or replacing the data values and is thus an evaluation.)
Claim 5 therefore recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 5 has no additional elements that would integrate the abstract idea into a practical application, and therefore, claim 4 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
Since there are no additional elements to provide significantly more than the abstract idea itself, taken alone and in combination, claim 5 is subject-matter ineligible.
Regarding Claim 6:
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 6 recites the same mental processes as in claim 1 and, therefore, recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 6 recites additional elements of
wherein the one or more neural networks are to perform one or more convolutions using a data set, including the invalid data values, using only valid data values. (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, Claim 6 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional element of Claim 6 does not provide significantly more than the abstract idea itself, taken alone and in combination, because
wherein the one or more neural networks are to perform one or more convolutions using a data set, including the invalid data values, using only valid data values uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, Claim 6 is subject-matter ineligible.
Regarding Claim 7:
Subject Matter Eligibility Analysis Step 1:
Claim 7 recites a system, comprising one or more processors and is thus an apparatus, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 7 recites
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (This limitation could encompass a human mentally replacing a value in their head with an invalid value and is thus an evaluation.)
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated (This limitation could encompass a human mentally replacing a data value in their head with a data value that’s unable to be generated and is thus an evaluation.)
detect the one or more invalid data values retained in the input data provided to the one or more operations (This limitation could encompass a human mentally detecting invalid data values and is thus an evaluation.)
cause the one or more operations… to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (This limitation could encompass a human mentally causing operations to be performed without considering the invalid data values and is thus an evaluation.)
Claim 7 therefore recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 7 further recites additional elements of
A system, comprising: one or more processors (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Generated by the one or more circuits during one or more operations of one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
of the one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, Claim 7 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional element of Claim 7 does not provide significantly more than the abstract idea itself, taken alone and in combination, because
A system, comprising: one or more processors uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Generated by the one or more circuits during one or more operations of one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
of the one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, Claim 7 is subject-matter ineligible.
Regarding claim 9, claim 9 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 10, claim 10 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 11, claim 11 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 12, claim 12 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Regarding Claim 13:
Subject Matter Eligibility Analysis Step 1:
Claim 13 recites a method and is thus a process, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 13 recites
causing one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (This limitation could encompass a human mentally replacing a value in their head with an invalid value and is thus an evaluation.)
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated (This limitation could encompass a human mentally replacing a data value in their head with a data value that’s unable to be generated and is thus an evaluation.)
detecting the one or more invalid data values retained in the input data provided to the one or more operations (This limitation could encompass a human mentally detecting invalid data values and is thus an evaluation.)
causing the one or more operations… to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (This limitation could encompass a human mentally causing operations to be performed without considering the invalid data values and is thus an evaluation.)
Claim 13 therefore recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 13 further recites additional elements of
Generated by the one or more circuits during one or more operations of one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
of the one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, Claim 13 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional element of Claim 13 does not provide significantly more than the abstract idea itself, taken alone and in combination, because
Generated by the one or more circuits during one or more operations of one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
of the one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, Claim 13 is subject-matter ineligible.
Regarding claim 15, claim 15 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 16, claim 16 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 17, claim 17 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 18, claim 18 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Regarding Claim 19:
Subject Matter Eligibility Analysis Step 1:
Claim 19 recites a non-transitory machine-readable medium and is thus an article of manufacture, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 19 recites
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (This limitation could encompass a human mentally replacing a value in their head with an invalid value and is thus an evaluation.)
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated (This limitation could encompass a human mentally replacing a data value in their head with a data value that’s unable to be generated and is thus an evaluation.)
detect the one or more invalid data values retained in the input data provided to the one or more operations (This limitation could encompass a human mentally detecting invalid data values and is thus an evaluation.)
cause the one or more operations… to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (This limitation could encompass a human mentally causing operations to be performed without considering the invalid data values and is thus an evaluation.)
Claim 19 therefore recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 19 further recites additional elements of
a non-transitory machine-readable medium having stored thereon a set of instructions, which if performed by one or more processors, cause the one or more processors to at least: cause one or more data values to be replaced (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Generated by the one or more circuits during one or more operations of one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
of the one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, Claim 19 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional element of Claim 19 does not provide significantly more than the abstract idea itself, taken alone and in combination, because
a non-transitory machine-readable medium having stored thereon a set of instructions, which if performed by one or more processors, cause the one or more processors to at least: cause one or more data values to be replaced uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Generated by the one or more circuits during one or more operations of one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
of the one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, Claim 19 is subject-matter ineligible.
Regarding claim 21, claim 21 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 22, claim 22 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 23, claim 23 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 24, claim 24 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Regarding Claim 25:
Subject Matter Eligibility Analysis Step 1:
Claim 25 recites a data processing system, comprising: one or more processors and is thus an apparatus, one of the four statutory categories of patentable subject matter.
Subject Matter Eligibility Analysis Step 2A Prong 1:
Claim 25 recites
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (This limitation could encompass a human mentally replacing a value in their head with an invalid value and is thus an evaluation.)
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated (This limitation could encompass a human mentally replacing a data value in their head with a data value that’s unable to be generated and is thus an evaluation.)
detect the one or more invalid data values retained in the input data provided to the one or more operations (This limitation could encompass a human mentally detecting invalid data values and is thus an evaluation.)
cause the one or more operations… to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (This limitation could encompass a human mentally causing operations to be performed without considering the invalid data values and is thus an evaluation.)
Claim 25 therefore recites an abstract idea.
Subject Matter Eligibility Analysis Step 2A Prong 2:
Claim 25 further recites additional elements of
An data processing system, comprising: one or more processors to: cause one or more data values to be replaced (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Generated by the one or more circuits during one or more operations of one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
of the one or more neural networks (This additional element does not integrate the abstract idea into a practical application because it recites a generic computer on which to perform the abstract idea, e.g., “apply it on a computer” (see MPEP 2106.05(f)).)
Therefore, Claim 25 is not integrated into a practical application.
Subject Matter Eligibility Analysis Step 2B:
The additional element of Claim 25 does not provide significantly more than the abstract idea itself, taken alone and in combination, because
An data processing system, comprising: one or more processors to: cause one or more data values to be replaced uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Generated by the one or more circuits during one or more operations of one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
of the one or more neural networks uses a computer as a tool to perform the abstract idea cannot provide significantly more (see MPEP 2106.05(f)).
Therefore, Claim 25 is subject-matter ineligible.
Regarding claim 27, claim 27 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 28, claim 28 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 29, claim 29 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 30, claim 30 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
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.
The factual inquiries 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, 3-7, 9-13, 15-19, 21-25, 27-30 are rejected under 35 U.S.C. 103 as being unpatentable over Meyer (US 12039330 B1) (“Meyer”), in view of Shivarathri (US 12153596 B1) (“Shivar”).
Regarding claim 1, Meyer teaches
a processor, comprising one or more circuits (Meyer, page 27, column 12, lines 63-67, “Execution engine 700 can be part of a data processor (e.g. a data processing integrated circuit device such as a processor, a graphics processor, a digital signal processor, a tensor processor, a neural network accelerator, or other types of application specific integrated circuits)”),
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (Meyer, page 27, column 12, line 4-5 “the first top N values [data values] have been replaced by the minimum value [invalid data value]” where “the minimum value [invalid data value] can be a value representing negative infinity (e.g., when operating on signed floating-point numbers), zero (e.g., when operating on positive numbers), or a value representing the largest magnitude negative integer (e.g., when operating on signed integers), etc.” (page 29, col 15, line 33- 38) and “At block 1010, a Match-Replace-N machine instruction is executed to replace the N largest values in the input tensor with a minimum value to generate the next input tensor for the next iteration” (Meyer, page 30, column, 18, lines 14-17).).
Meyer does not teach, but Shivar does teach
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of one or more neural networks (Shivar, page 12, column 11, lines 39-43, “The anomalous data [invalid data values] in the input data set may include incorrect data types of data entries, empty data fields of data entries, extra data fields in the data set, missing data fields in the data set [values that are unable to be generated], or the like” wherein “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that since the invalid data values are in the input data, they are pre-designated.);
detect the one or more invalid data values retained in the input data provided to the one or more operations of the one or more neural networks (Shivar, page 9, column 5, lines 44-51, “Identification of the anomalous data [invalid data] 114 may be performed on the input data set 106 based on the context metadata 108 and/or the associated context profile 110 of the input data set 106. For instance, if a data value in the data set 106 is a string type and the context metadata 108 and/or context profile 110 indicates that the data value is expected to be an integer type, the data value may be determined to be anomalous data 114” where “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that detecting the one or more invalid data is determining whether the data value is anomalous. Examiner further notes that the input data in which the invalid data value is detected is provided to the operation of extraction. Examiner notes that the operation of extraction uses the one or more neural network within the process.).;
cause the one or more operations of the one or more neural network to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (Shivar, page 4, Figure 3,
PNG
media_image1.png
1016
652
media_image1.png
Greyscale
and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41). Examiner notes that the anomalous data or invalid data is corrected before the extraction operation at 320. Thus, the extraction operation is performed without processing the invalid data. Examiner further notes that the valid data is the data that is not anomalous.).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to have the invalid data correspond to values that are unable to be generated. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 3, Meyer in view of Shivar teach the processor of claim 1. Meyer further teaches
the one or more circuits are further to cause a plurality of data values to be loaded into one or more registers (Meyer, page 23, column 8, line 11-14, “the set of largest values are first loaded into the feedback registers of the pipeline”).
wherein the plurality of data values include one or more valid data values and the one or more invalid data values (Meyer, page 24, column 6, line 4-7, “Fig 3b at cycle 3, ALU stage-3… stores -[infinity] [invalid data value] in the feedback register. ALU stage-2... stores 21.1 [valid data value] in the feedback register”).
Regarding claim 4, Meyer in view of Shivar teaches the processor of claim 3 (and thus the rejection of claim 3 is incorporated). Meyer also teaches
wherein the one or more circuits are further to utilize the plurality of data values from the one or more registers to perform one or more operations in a sequence (Meyer, page 22, column 2, line 16, “Each ALU stage [one or more circuits] can perform various arithmetic operations on a set of inputs [plurality of data values] and provides the result to the next ALU stage” where “[30.0, 21.2, 33.1, -2.5, 31.0, 40.2, 17.3, 40.2] represent[] an input tensor [plurality of data values]” (Meyer, page 24, column 5, lines 21-22) and “each ALU [circuit] is configured to select the feedback value [data value] from the feedback register” (Meyer, page 24, column 5, lines 52-53).
Meyer does not teach, but Shivar does teach
wherein individual operations of the sequence are able to identify the invalid data values (Shivar, page 9, column 5, lines 44-51, “Identification of the anomalous data [invalid data] 114 may be performed on the input data set 106 based on the context metadata 108 and/or the associated context profile 110 of the input data set 106. For instance, if a data value in the data set 106 is a string type and the context metadata 108 and/or context profile 110 indicates that the data value is expected to be an integer type, the data value may be determined to be anomalous data 114” [process of determining that the data value is of the wrong type = sequence of operations]).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to be able to identify invalid data values. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 5, Meyer in view of Shivar teaches the processor of claim 4 (and thus the rejection of claim 4 is incorporated). Meyer further teaches
wherein at least one individual operation of the one or more operations is enabled to propagate the invalid data values (Meyer, page 24, column 5, lines 35-37, “the feedback registers of all ALU stages are initialized [one or more operations] to a minimum value [invalid data values] prior to streaming [propagating] the input values into the pipeline”).
Meyer does not teach, but Shivar does teach
wherein at least one individual operation of the one or more operations is enabled… replace the invalid data values (Shivar, abstract, “a correction model is applied to the anomalous data [invalid data] to generate corrected data [replace invalid data values] based on the identified context”).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to be able to replace invalid data values. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 6, Meyer in view of Shivar teaches the processor of claim 1 (and thus the rejection of claim 1 is incorporated). Meyer further teaches
wherein the one or more neural networks are to perform one or more convolutions (Meyer, page 34, col 25, line 22-24 “For example, for a convolutional neural network, convolutions from multiple channels can be summed to produce an output activation for a single channel.”)
Meyer does not teach, but Shivar does teach
using a data set, including the invalid data values, using only valid data values (Shivar, page 12, col 11, lines 36 - 43 “the identified context is used by the system to determine data in the input data set that is expected [valid data values] and data that is not expected, or anomalous [invalid data values], as described herein. The anomalous data in the input data set may include incorrect data types of data entries, empty data fields of data entries, extra data fields in the data set, missing data fields in the data set, or the like” wherein “a correction model is applied to the anomalous data to generate corrected data [using only valid data values]”(Shivar, 12, column 11, lines 48-49)).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to be able to replace invalid data values. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 7, Meyer teaches
a system, comprising: one or more processor (Meyer, page 27, column 12, lines 63-67, “Execution engine 700 can be part of a data processor (e.g. a data processing integrated circuit device such as a processor, a graphics processor, a digital signal processor, a tensor processor, a neural network accelerator, or other types of application specific integrated circuits)”),
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (Meyer, page 27, column 12, line 4-5 “the first top N values [data values] have been replaced by the minimum value [invalid data value]” where “the minimum value [invalid data value] can be a value representing negative infinity (e.g., when operating on signed floating-point numbers), zero (e.g., when operating on positive numbers), or a value representing the largest magnitude negative integer (e.g., when operating on signed integers), etc.” (page 29, col 15, line 33- 38) and “At block 1010, a Match-Replace-N machine instruction is executed to replace the N largest values in the input tensor with a minimum value to generate the next input tensor for the next iteration” (Meyer, page 30, column, 18, lines 14-17).).
Meyer does not teach, but Shivar does teach
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of one or more neural networks (Shivar, page 12, column 11, lines 39-43, “The anomalous data [invalid data values] in the input data set may include incorrect data types of data entries, empty data fields of data entries, extra data fields in the data set, missing data fields in the data set [values that are unable to be generated], or the like” wherein “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that since the invalid data values are in the input data, they are pre-designated.);
detect the one or more invalid data values retained in the input data provided to the one or more operations of the one or more neural networks (Shivar, page 9, column 5, lines 44-51, “Identification of the anomalous data [invalid data] 114 may be performed on the input data set 106 based on the context metadata 108 and/or the associated context profile 110 of the input data set 106. For instance, if a data value in the data set 106 is a string type and the context metadata 108 and/or context profile 110 indicates that the data value is expected to be an integer type, the data value may be determined to be anomalous data 114” where “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that detecting the one or more invalid data is determining whether the data value is anomalous. Examiner further notes that the input data in which the invalid data value is detected is provided to the operation of extraction. Examiner notes that the operation of extraction uses the one or more neural network within the process.).;
cause the one or more operations of the one or more neural network to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (Shivar, page 4, Figure 3,
PNG
media_image1.png
1016
652
media_image1.png
Greyscale
and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41). Examiner notes that the anomalous data or invalid data is corrected before the extraction operation at 320. Thus, the extraction operation is performed without processing the invalid data. Examiner further notes that the valid data is the data that is not anomalous.).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to have the invalid data correspond to values that are unable to be generated. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 9, claim 9 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 10, claim 10 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 11, claim 11 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 12, claim 12 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Regarding claim 13, Meyer teaches
a method, comprising: causing one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (Meyer, page 27, column 12, line 4-5 “the first top N values [data values] have been replaced by the minimum value [invalid data value]” where “the minimum value [invalid data value] can be a value representing negative infinity (e.g., when operating on signed floating-point numbers), zero (e.g., when operating on positive numbers), or a value representing the largest magnitude negative integer (e.g., when operating on signed integers), etc.” (page 29, col 15, line 33- 38) and “At block 1010, a Match-Replace-N machine instruction is executed to replace the N largest values in the input tensor with a minimum value to generate the next input tensor for the next iteration” (Meyer, page 30, column, 18, lines 14-17).).
Meyer does not teach, but Shivar does teach
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of one or more neural networks (Shivar, page 12, column 11, lines 39-43, “The anomalous data [invalid data values] in the input data set may include incorrect data types of data entries, empty data fields of data entries, extra data fields in the data set, missing data fields in the data set [values that are unable to be generated], or the like” wherein “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that since the invalid data values are in the input data, they are pre-designated.);
detecting the one or more invalid data values retained in the input data provided to the one or more operations of the one or more neural networks (Shivar, page 9, column 5, lines 44-51, “Identification of the anomalous data [invalid data] 114 may be performed on the input data set 106 based on the context metadata 108 and/or the associated context profile 110 of the input data set 106. For instance, if a data value in the data set 106 is a string type and the context metadata 108 and/or context profile 110 indicates that the data value is expected to be an integer type, the data value may be determined to be anomalous data 114” where “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that detecting the one or more invalid data is determining whether the data value is anomalous. Examiner further notes that the input data in which the invalid data value is detected is provided to the operation of extraction. Examiner notes that the operation of extraction uses the one or more neural network within the process.).;
causing the one or more operations of the one or more neural network to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (Shivar, page 4, Figure 3,
PNG
media_image1.png
1016
652
media_image1.png
Greyscale
and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41). Examiner notes that the anomalous data or invalid data is corrected before the extraction operation at 320. Thus, the extraction operation is performed without processing the invalid data. Examiner further notes that the valid data is the data that is not anomalous.).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to have the invalid data correspond to values that are unable to be generated. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 15, claim 15 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 16, claim 16 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 17, claim 17 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 18, claim 18 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Regarding claim 19, Meyer teaches
A non-transitory machine-readable medium having stored thereon a set of instructions, which if performed by one or more processors cause the one or more processors to at least: (Meyer, page 37, column 31, lines 1-13, “Computer-readable storage media are physical mediums that are capable of storing data in a format that can be read by a device such as the host processor 1572. Computer-readable storage media can be non-transitory. Non-transitory computer-readable media can retain the data stored thereon when no power is applied to the media….In various examples the data stored on computer-readable storage media can include program instructions”),
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (Meyer, page 27, column 12, line 4-5 “the first top N values [data values] have been replaced by the minimum value [invalid data value]” where “the minimum value [invalid data value] can be a value representing negative infinity (e.g., when operating on signed floating-point numbers), zero (e.g., when operating on positive numbers), or a value representing the largest magnitude negative integer (e.g., when operating on signed integers), etc.” (page 29, col 15, line 33- 38) and “At block 1010, a Match-Replace-N machine instruction is executed to replace the N largest values in the input tensor with a minimum value to generate the next input tensor for the next iteration” (Meyer, page 30, column, 18, lines 14-17).).
Meyer does not teach, but Shivar does teach
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of one or more neural networks (Shivar, page 12, column 11, lines 39-43, “The anomalous data [invalid data values] in the input data set may include incorrect data types of data entries, empty data fields of data entries, extra data fields in the data set, missing data fields in the data set [values that are unable to be generated], or the like” wherein “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that since the invalid data values are in the input data, they are pre-designated.);
detect the one or more invalid data values retained in the input data provided to the one or more operations of the one or more neural networks (Shivar, page 9, column 5, lines 44-51, “Identification of the anomalous data [invalid data] 114 may be performed on the input data set 106 based on the context metadata 108 and/or the associated context profile 110 of the input data set 106. For instance, if a data value in the data set 106 is a string type and the context metadata 108 and/or context profile 110 indicates that the data value is expected to be an integer type, the data value may be determined to be anomalous data 114” where “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that detecting the one or more invalid data is determining whether the data value is anomalous. Examiner further notes that the input data in which the invalid data value is detected is provided to the operation of extraction. Examiner notes that the operation of extraction uses the one or more neural network within the process.).;
cause the one or more operations of the one or more neural network to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (Shivar, page 4, Figure 3,
PNG
media_image1.png
1016
652
media_image1.png
Greyscale
and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41). Examiner notes that the anomalous data or invalid data is corrected before the extraction operation at 320. Thus, the extraction operation is performed without processing the invalid data. Examiner further notes that the valid data is the data that is not anomalous.).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to have the invalid data correspond to values that are unable to be generated. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 21, claim 21 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 22, claim 22 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 23, claim 23 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 24, claim 24 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Regarding claim 25, Meyer teaches
An data processing system, comprising one or more processors (Meyer, page 27, column 12, lines 63-67, “Execution engine 700 can be part of a data processor (e.g. a data processing integrated circuit device such as a processor, a graphics processor, a digital signal processor, a tensor processor, a neural network accelerator, or other types of application specific integrated circuits)”),
cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks (Meyer, page 27, column 12, line 4-5 “the first top N values [data values] have been replaced by the minimum value [invalid data value]” where “the minimum value [invalid data value] can be a value representing negative infinity (e.g., when operating on signed floating-point numbers), zero (e.g., when operating on positive numbers), or a value representing the largest magnitude negative integer (e.g., when operating on signed integers), etc.” (page 29, col 15, line 33- 38) and “At block 1010, a Match-Replace-N machine instruction is executed to replace the N largest values in the input tensor with a minimum value to generate the next input tensor for the next iteration” (Meyer, page 30, column, 18, lines 14-17).).
Meyer does not teach, but Shivar does teach
wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of one or more neural networks (Shivar, page 12, column 11, lines 39-43, “The anomalous data [invalid data values] in the input data set may include incorrect data types of data entries, empty data fields of data entries, extra data fields in the data set, missing data fields in the data set [values that are unable to be generated], or the like” wherein “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that since the invalid data values are in the input data, they are pre-designated.);
detect the one or more invalid data values retained in the input data provided to the one or more operations of the one or more neural networks (Shivar, page 9, column 5, lines 44-51, “Identification of the anomalous data [invalid data] 114 may be performed on the input data set 106 based on the context metadata 108 and/or the associated context profile 110 of the input data set 106. For instance, if a data value in the data set 106 is a string type and the context metadata 108 and/or context profile 110 indicates that the data value is expected to be an integer type, the data value may be determined to be anomalous data 114” where “input data sets 106 may be obtained from various data sources, such as customers or other entities that have been provided access to the system 100. Alternatively, or additionally, the system [one or more circuits] 100 may include a 15 buffer data store in which input data sets 106 are stored prior to processing and from which the extraction layer 102 obtains the input data sets 106 based on the configuration of the extraction layer 102 [one or more operations]” (Shivar, page 8, column 4, lines 11 - 19) and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41) and where “the system 100 may be configured to enable the auditing or reviewing of the performance of the correction model 116 as it operates to improve its performance” (Shivar, page 9, column 6, lines 62-65). Examiner notes that detecting the one or more invalid data is determining whether the data value is anomalous. Examiner further notes that the input data in which the invalid data value is detected is provided to the operation of extraction. Examiner notes that the operation of extraction uses the one or more neural network within the process.).;
cause the one or more operations of the one or more neural network to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data (Shivar, page 4, Figure 3,
PNG
media_image1.png
1016
652
media_image1.png
Greyscale
and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41). Examiner notes that the anomalous data or invalid data is corrected before the extraction operation at 320. Thus, the extraction operation is performed without processing the invalid data. Examiner further notes that the valid data is the data that is not anomalous.).
Meyer and Shivar are considered analogous because they are in the same field of machine learning, and they are both data processing systems. It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Meyer to have the invalid data correspond to values that are unable to be generated. Doing so would “enable[] the system to be flexible in what types and/or formats of data it is configured to receive and to automatically respond to detected issues with the data, avoiding manual review which may impede the data processing” (Shivar, page 8, column 3, line 4-7).
Regarding claim 27, claim 27 recites substantially similar limitations to claim 3, and is therefore rejected under the same analysis.
Regarding claim 28, claim 28 recites substantially similar limitations to claim 4, and is therefore rejected under the same analysis.
Regarding claim 29, claim 29 recites substantially similar limitations to claim 5, and is therefore rejected under the same analysis.
Regarding claim 30, claim 30 recites substantially similar limitations to claim 6, and is therefore rejected under the same analysis.
Response to Arguments
On page 9, Applicant argues:
Meyer in view of Shivarathri fails to teach to cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks. The Office Action relies on Meyer to teach causing one or more data values to be replaced by one or more invalid data values. Meyer teaches replacing the N largest values in a tensor with a minimum value (e.g., negative infinity, zero, max. integer value) so that the next M largest values in the tensor can be determined without recollecting the initial N values. However, the values used in Meyer to replace the N largest values in the tensor are not invalid data types, contrary to the Office Action's interpretation. The data values taught by Meyer are valid data types used specifically because they represent lowest possible values in the tensor so that they will not be evaluated as one of the M next largest values. Additionally, Shivarathri, even in combination with Meyer, also fails to teach causing one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks. Shivarathri teaches a system that includes a correction model used to correct "anomalous data" detecting in an input data set by replacing the anomalous data with corrected data. As with Meyer's data, the replacement data in Shivarathri are not invalid data values. Instead, Shivarathri specifically corrects the detected anomalous data with, by definition, valid data. Thus, Shivarathri teaches replacing data values with valid (e.g., correct) data values.
Regarding the Applicant’s argument that Meyer in view of Shivarathri fails to teach “cause one or more data values to be replaced by one or more invalid data values in input data for one or more neural networks,” Examiner respectfully disagrees. Specifically, Examiner respectfully notes that under broadest reasonable interpretation, negative infinity is an invalid data value that is used to replace data values in input data. Paragraph 0046 of the instant specification supports negative infinity as invalid data with “a special or invalid value in this context refers to a value that would not be produced by an operation to be performed, or determined to be a valid input value for an operation to be performed. In at least one embodiment, this might include a value such as "-0" (or another value that may be a "not a number" (NaN) value and that would not be produced by a typical operation, such as a value considered to be a numeric data type that can be interpreted as a value that is undefined or unrepresentable” (paragraph 0046).
On pages 9-10, Applicant argues:
Meyer in view of Shivarathri also fails to teach wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of the one or more neural networks. The Office Action acknowledges that Meyer does not teach this feature, referring to Shivarathri's discussion of "anomalous data in the input data set may include incorrect data types of data entries, empty data fields of data entries, extra data fields in the data set, missing data fields in the data set, or the like" citing column 11, lines 39-43 of Shivarathri. At the cited passage, Shivarathri describes anomalous data in an input data set, where the anomalous data can include incorrect data types, empty data fields, extra data fields and missing data fields. Firstly, Shivarathri's anomalous data values are not described as, and are not, data values replacing other data values. Applicant's claim requires causing data values to be replaced by pre-designated invalid data values. Shivarathri's anomalous data values are not used to replace other data values, as required by Applicant's claim.
Additionally, the anomalous data values in Shivarathri are not invalid data values that are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of the one or more neural networks. Shivarathri describes the anomalous data as incorrect data (e.g., incorrect data types, empty fields, missing fields, extra fields, etc.) and teaches correcting the anomalous data. Incorrect data types, empty fields, missing fields and extra fields are all examples of valid data values, even if incorrect in a specific context. In response to Applicant's previous remarks, the Office Action states, "the anomalous data, such as the empty data values in Shivarathri map to values that are unable to be generated by the one or more circuits during one or more operations of the one or more neural network" and further states, "[s]ince the data is anomalous, the data is not correct and therefore invalid," citing column 5, lines 56- 61 of Shivarathri (Office Action, p. 47). However, Shivarathri does not describe the anomalous data values as pre-designated or as values corresponding to values that are unable to be generated by the one or more circuits during one or more operations of one or more neural networks. The Office Action is merely speculating about whether or not Shivarathri's anomalous data values are able to be generated by operations of a neural network resulting from the combination of Meyer and Shivarathri. Even under the Office Action's interpretation (i.e., in which Shivarathri's anomalous data is equivalent to Meyer's replacement minimum values), Shivarathri fails to teach wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of the one or more neural networks.
Regarding Applicant’s argument that the prior art of record does not teach “wherein the one or more invalid data values are pre-designated and correspond to values that are unable to be generated by the one or more circuits during one or more operations of the one or more neural networks,” Examiner respectfully disagrees. Specifically Examiner respectfully notes that since the invalid data values, or anomalous data, in Shivarathri is in the input data, the anomalous data is pre-designated. Examiner further notes that as described in the above 103 rejection, Meyer discloses replacing data with invalid data and Shivarathri discloses that the invalid data values are pre-designated. Additionally, Examiner notes that pre-designated invalid data values are not supported by the specification as noted in the 112(a) rejection above.
Examiner further respectfully notes that the anomalous data, such as the empty data values in Shivarathri map to values that are unable to be generated by the one or more circuits during one or more operations of the one or more neural networks. Since the data is anomalous, the data is not correct and therefore invalid (Shivar, page 9, column 5, lines 56-61, “Further, anomalous data 114 ma include anomalies such as metadata values that are unexpected (e.g., typos in metadata labels or field names of the data set) or otherwise found to be incorrect.”). Examiner further respectfully notes that Shivarathri discloses operations of storing and extracting data via a system or circuit (Shivarathri, page 8, column 4, lines 11-19). In both operations, data is not generated since the operation of storing and extracting data is storing and reading data. Examiner further notes specifically, that the “correction model 116 may be configured to receive the input data set 106, identify the anomalous data 114 as described above, and output corrected data 118 to the input data set 106 as output” (Shivarathri, page 9, column 6, lines 1-4). Additionally, in all of these operations of the correction model, the data is not changed nor is data generated. Examiner further notes that since the extraction operation occurs on the data that the correction model identifies, under broadest reasonable interpretation, the extraction operation is of the neural network.
On pages 11-12, Applicant argues:
In further regard to claim 1, Meyer in view of Shivarathri also fails to teach to cause the one or more operations of the one or more neural networks to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data, as in Applicant's claim. Meyer's system clearly processes the replacement minimum values, which the Office Action equates to the invalid data values of claim 1, when identifying the next M largest values in the tensor ( otherwise Meyer's process would not function properly). The Office Action also acknowledges that Meyer does not teach this feature and refers to operation 320 in FIG. 3 of Shivarathri. However, Shivarathri's correction model is the only neural network described in Shivarathri and the correction model does not perform its operations using valid data values without processing the invalid data retained in the input data. Under the Office Action's interpretation (i.e., in which Shivarathri's anomalous data is equivalent to Meyer's replacement minimum values, which Applicant maintains is an incorrect interpretation of the cited references), the correction model clearly processes the anomalous data since it detects it and corrects it. Note, the Office Action acknowledges that Shivarathri's correction model processes the anomalous data by detecting and correcting it (see, e.g., Office Action, pp. 21-23). Under the Office Action's alternative interpretation that Shivarathri's extraction process corresponds to the one or more operations of Applicant's claim, Shivarathri also fails to teach this feature. Shivarathri' s extraction process receives as input the corrected input data without the anomalous data and thus cannot be considered to perform operations using valid data values of the input data without processing the one or more invalid data values retained in the input data.
Additionally, operation 320 is not an operation of Shivarathri' s neural network. As described in Shivarathri regarding FIG. 3, operation 320 is not an operation of Shivarathri's correction model neural network. Nor is the data extracted at 320 "input data" to an operation of the neural network where the input data includes both valid data values and the one or more invalid data values. In fact, the Office Action explicitly relies on the anomalous data in Shivarathri having already been corrected before operation 320. In response to Applicant's previous remarks, the Office Action notes, "the anomalous data or invalid data is corrected before the extraction operation at 320" and concludes, "[t]hus, the extraction operation is performed without considering the invalid data" (Office Action, p. 50). The Office Action further states, "under broadest reasonable interpretation 'of the input data' requires the data values to be derived from the input data" and that "both the valid and invalid data values are derived from the input data" (Office Action, p. 50). However, Applicant's amended claim recites, in part, "cause the one or more operations of the one or more neural networks to be performed using valid data values of the input data without processing the one or more invalid data values retained in the input data" (italics added). Thus, even under the Office Action's interpretation, Shivarathri's extraction operation fails to teach this feature of claim 1.
Moreover, the combination of Meyer and Shivarathri fails to teach this feature of claim 1. A system based on the combination of Meyer and Shivarathri would clearly process Meyer's replacement minimum values and/or Shivarathri's anomalous data since a) the replacement values in Meyer must be processed (as minimum values) for Meyer's system to function properly and b) the entire point of Shivarathri's system is to detect and correct anomalous data in the input data.
Regarding the Applicant’s argument that the prior art of record does not teach “to cause the one or more operations of the one or more neural networks to be performed using valid data values of the input data without processing the one or more invalid data values”, Examiner respectfully disagrees. Specifically, Shivarathri teaches this limitation (Shivar, page 4, Figure 3, see below
PNG
media_image1.png
1016
652
media_image1.png
Greyscale
and “the training of the correction model 116 includes machine learning techniques that use, for instance, a trained regressor such as a random decision forest, a directed acyclic graph, a support vector machine, a convolutional neural network or other neural network” (Shivar, page 9, column 6, lines 37-41). ) Examiner notes that the anomalous data or invalid data is corrected before the extraction operation at 320. Thus, the extraction operation, which is one or more operations of the neural network, is performed without processing the invalid data. Examiner further notes that the valid data is the data that is not anomalous. Examiner additionally notes that since the extraction operation occurs on the data that the correction model identifies, under broadest reasonable interpretation, the extraction operation is of the neural network.
Examiner further notes that as described in the above 103 rejection, Meyer discloses replacing data with invalid data and Shivarathri discloses that the invalid data values retained in the input data. Under broadest reasonable interpretation, the operations of the one or more neural networks are performed using valid data values of the input data without processing invalid data values that were retained in the input data. In other words, the invalid data values, under BRI, do not need to be in the input data when the processing is occurring. Instead, the invalid data values must have been in the input data at some point prior to processing.
On page 13, Applicant argues:
While Applicant disagrees with this rejection, Applicant has made certain clarifying amendments. Applicant submits that the claims provide an improvement in computer technology. For example, the specific features recited in Applicant's claims provide an improvement to computer technology by potentially minimizing storage requirements, increasing computation speed, and/or saving power, as noted at least at paras. [0002], [0048] and [0049] of Applicant's specification. Thus, the recent precedential PT AB Decision Ex Parte Desjardins et al., Appeal 2024-000567 (ARP Sept. 26, 2025) (precedential) is exactly on point. Moreover, claims presenting an improvement to computer functionality have self-evident eligibility under the streamlined analysis shown as Pathway A of MPEP 2106(111) such that additional stages of an eligibility analysis are not required. Even if additional analysis is undertaken, however, the claims are not directed to a judicial exception under Pathway B of MPEP 2106(111), at least because the claimed subject matter integrates any abstract idea into a practical application, which would again make it unnecessary to proceed to additional steps of an eligibility analysis. Even if the claims could be considered directed to an abstract idea, which Applicant does not concede, the claims would be patent-eligible based on Pathway C of MPEP 2106(111) because additional elements recited in the claim amount to substantially more than any such abstract idea. Accordingly, Applicant's amended claims are patent eligible at least on this basis as an improvement in computer technology.
In the response to arguments, the Office Action argues that the "clarifying details in paragraphs 0002, 0048, and 0049 of the instant specification are not explicitly claimed, and therefore, cannot provide an improvement of minimizing storage requirements, increasing speed of computation, and save power" (Office Action, p. 51). However, the passages from the specification were cited to explain benefits and improvement resulting from the claims, not as features of the claims as written.
Moreover, the claims recite a specific "ordered combination" of features to replace data values to be used by one or more neural networks by invalid data values that are unable to be generated by the one or more circuits during one or more operations of the one or more neural networks, detect the one or more invalid data values in input data provided to the one or more operations of the neural network, and cause the one or more operations of the one or more neural networks to be performed using valid data values of the input data without considering the one or more invalid data values of the input data. Such an "ordered combination" provides "significantly more" than a mere abstract idea.
Regarding the Applicant’s argument that claims 1, 3-7, 9-13, 15-19, 21-25, 27-30 are patent eligible, Examiner respectfully disagrees. Specifically, Examiner respectfully notes that the clarifying details in paragraphs 0002, 0048, and 0049 of the instant specification are not explicitly claimed, and therefore, cannot provide an improvement of minimizing storage requirements, increasing speed of computation, and save power. Additionally, Examiner respectfully notes that this “ordered combination” is not recited in the claims and thus cannot provide an improvement or significantly more.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Jung et al. (US 2022/0284263 A1) discloses detecting invalid data in order to zero-pad the image.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KAITLYN R LAU whose telephone number is (571)272-1429. The examiner can normally be reached Monday - Thursday: 8:00 am - 6:00 pm EST.
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, Michelle Bechtold can be reached on (571) 431-0762. 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.
/K.R.L./Examiner, Art Unit 2148 /MICHELLE T BECHTOLD/Supervisory Patent Examiner, Art Unit 2148
