DETAILED ACTION
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 .
Response to Amendment
This office action is responsive to amendment filed on 03/23/2026. Claims 1, 3-5, 7-10, and 13 are pending. The amendment have overcome the claim objection and 112(b) as set forth in the previous office action.
Response to Arguments
In response to applicant’s argument regarding rejection under 35 U.S.C 101 on page 9-10 that amended independent claim 1 does not recite any mathematical equation, formula or abstract mathematical relationship, rather claim 1 recites a method performed by a processor for calibrating quantization in a neural network deployed in a device configured to perform image processing and also asserted on page 11 that Examiner’s reliance on equations disclosed in the Specification to characterize the claim as reciting mathematical calculation is inconsistence with MPEP 2106.04(a)(2)(I) which requires that the mathematical concepts itself be recited in the claim.
Examiner respectfully disagrees because MPEP 2106.04(a)(2)(C) recites “A claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping …That is, a claim does not have to recite the word "calculating" in order to be considered a mathematical calculation. For example, a step of "determining" a variable or number using mathematical methods or "performing" a mathematical operation may also be considered mathematical calculations when the broadest reasonable interpretation of the claim in light of the specification encompasses a mathematical calculation”. Thus, the claimed invention recites a step of quantizing of the calibration data set comprises dividing data, and determining a total quantization difference metric for the quantization and determining an optimized truncated threshold cover mathematical calculations, relationship, and/or formula, when broadest reasonable interpretation in light of the specification (see at least figure 6 step 630, [0086-0087] describe the step 630 is performed by using mathematical equations 3-9, and 11-12. [0086-0087] describes total quantization difference metric may be calculated with reference to equation 5, wherein equation 6-7 are used to determine quantization difference metric DistQ and DistC of the quantized data part DQ and truncated data part DC. [0075] also describes equations 6-7 calculate the quantization difference metric DistQ and quantization difference metric DistC based on noise amplitude AQ and correlation coefficient EQ). Furthermore, merely reciting a method performed by a processor for calibrating quantization in a neural network, wherein the neural network is deployed in a device configured to perform image processing at a high level of generality, which amounts to no more than mere instructions to apply the judicial exception (e.g., quantization data set) using computer component as MPEP 2106.05(f) recites “simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not integrate a judicial exception into a practical application or provide significantly more.” and at most considered as mere generally linking the use of the judicial exception (e.g., quantization data set) into a particular technological environment or field of use, such as neural network for image processing (see MPEP 2106.05(h)).
Applicant further asserted on page 10-11 a sequence of implementation level processing operations that define a structured calibration workflow that recite calibrating quantization in neural network deployed in an image processing device and this ordered sequence results in selection of a hardware-usable quantization parameter that is subsequently used by the processor when executing the neural network in a device configured for image processing, thereby governing practical deployment of the neural network.
Examiner respectfully disagrees because beside the step of receiving data, which is considered to be insignificant extra solution activity under step 2A prong two and determined to be well-understood, routine, and conventional under step 2B, the rest of the sequence of operations are characterized as the abstract idea under step 2A prong one. Thus, such sequence of mathematical operations are still mathematical concepts being implemented on a computer component for imaging processing, and as explained above, such additional element of image processing device in neural network does not integrate the judicial exception into a practical application as it amounts to no more than mere instructions to apply the judicial exception using computer component and at most considered as mere generally linking the use of the judicial exception (e.g., quantization data set) into a particular technological environment or field of use, such as neural network for image processing as explained above.
Applicant further asserted on page 12 “Amended independent claim 1, as a whole, imposes meaningful limits on any such mathematical concept by embedding it within a specific quantization calibration framework in a neural network deployed in a device configured for image processing.”
Examiner respectfully disagrees because merely reciting quantization calibration framework in a neural network deployed in a device configured for image processing does not integrate the judicial exception into a practical application since such limitations amount to no more than mere instructions to apply the judicial exception using computer component and at most considered as mere generally linking the use of the judicial exception (e.g., quantization data set) into a particular technological environment or field of use, such as neural network for image processing as explained above (see MPEP 2106.05(f) and 2106.05(h)).
Applicant further asserted on page 13, “MPEP § 2106.04(d)(II) indicates that "[e]xaminers evaluate integration into a practical application by: (1) identifying whether there are any additional elements recited in the claim beyond the judicial exception(s); and (2) evaluating those additional elements individually and in combination to determine whether they integrate the exception into a practical application " It should be apparent that the phrase "in combination" refers to a combination with all the other elements of the claim and not merely in combination with the other "additional elements." Otherwise, an analysis that looks only at the "additional limitations" by themselves and that does not consider the interplay between those limitations and the limitations that encompass the alleged abstract idea identified at Prong One, directly violates the Supreme Court's mandate to evaluate the claims "as a whole." Alice Corp., 573 U.S. at 218 n.3; see also Diamond v. Diehr, 450 U.S. 175, 188 (1981) and MPEP § 2106.04(d)(III).”
Examiner has analyzed the claimed as a whole under step 2A prong two and step 2B and determined that the additional elements fail to integrate the judicial exception into a practical application or provide significantly more. Furthermore, the instant claimed invention is not similar to Diamond v Diehr as the claim in Diehr recites additional elements such as the steps of installing rubber in a press, closing the mold, constantly measuring the temperature in the mold, and automatically opening the press at the proper time, which the Courts found them to be meaningful because they sufficiently limited the use of the mathematical equation to the practical application of molding rubber products. In contrast, the instant claim merely recites a method for calibrating quantization in a neural network deployed in a device configured to perform image processing, and such additional elements are recited at a high level of generality, amount to no more than mere instructions to apply the judicial exception using computer component and at most considered as mere generally linking the use of the judicial exception (e.g., quantization data set) into a particular technological environment or field of use, such as neural network for image processing as explained above.
Applicant further asserted on page 14-16 that the claimed invention provides an improvement in the functioning of a computer, where the quantization in neural network is required to reduce computation volume and save computation and storage resources and also compared to Desjardins case.
Examiner respectfully disagrees because any arguably improvements, such as optimizing quantization parameters, reducing computation volume, saving computation resources, saving storage resources, shortening the processing cycle, are a direct consequence of performing the mathematical algorithm as recited in the claims or recited in [0005], wherein a solution of optimizing quantization parameters using a new quantization difference metric, and [0078] describes the steps of dividing input data into quantized data and truncated data and determining the total quantization difference metric used, which facilitate optimization of the quantization parameters and provide higher quantization precision. MPEP 2106.05(a) recites “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements”. Thus, steps of quantizing the calibration dataset including dividing, determining a total quantization difference metric and determining an optimized truncated threshold are characterized as judicial exception (e.g., mathematical concept), and such limitations alone cannot provide the improvement. Furthermore, the claimed invention is different from the Desjardins case as the claim of Desjardins is directed to using machine learning and perform training in neural network model, but the instant claimed invention is directed to a quantization method for a neural network, without reciting any use of machine learning or training neural network.
Applicant further asserted on page 17, “Similar to the Desjardins case, the improvement here lies in the internal functioning of the computer system itself, namely, the manner in which neural network parameters are calibrated and deployed in devices configured to perform image-processing with high inference precision and computational efficiency. The claimed invention therefore improves processor operation, memory utilization, and inference execution efficiency in practical image processing systems. Accordingly, the claims are directed to a technological improvement in computer functionality, not to an abstract mathematical concept.”
Examiner respectfully disagrees because as explained above, any arguably improvement is a direct consequence of performing the sequence of mathematical operations as recited in the claims and the MPEP 2106.05(a) state that judicial exception alone cannot provide the improvement and the limitation of neural network being deployed on a device configured to perform image process is recited at a high level of generality that do not integrate the judicial exception into a practical application as explained above. Furthermore, the instant claim is not similar to the Desjardins case because the claim of Desjardins relates to training a machine learning models, where parameters are adjusted based on training machine learning models. However, the instant claim does not recite any limitation related to training machine learning model or neural network. At most, the instant claim recites the use of quantization operation of calibrating data set in a neural network, which is mere generally linking the use of judicial exception into a field of use or a particular technological environment of neural network.
Applicant further asserted on page 18-20 for step 2B analysis and that claim 1 recites a sequence of operations that is non-generic, non-conventional quantization calibration architecture, such as the claim requires dividing data, separately determining a quantization difference, determining each quantization difference metric and also asserted that these limitations are not well-understood, routine, or conventional in the field, but instead recite a quantization calibration arrangement that confines the claimed invention to a specific, hardware-adapted implementation in device configured to perform image processing consistent with the inventive concept framework articulated in BASCOM and MPEP 2106.05.
Examiner respectfully disagrees because the search for an inventive concept under step 2B is to evaluate whether the additional elements amount to an inventive concept. However, beside the step of receiving data, the sequence of operations as identified in page 18-19 are characterized as the abstract idea under step 2A prong one, rather than additional elements under step 2A prong two. Thus, even though such sequence of operations are non-generic and non-conventional, they are still abstract idea. MPEP 2106.04(I) “The Supreme Court’s decisions make it clear that judicial exceptions need not be old or long-prevalent, and that even newly discovered or novel judicial exceptions are still exceptions” and MPEP 2106.05(I) “a claim for a new abstract idea is still an abstract idea”. In other words, the arrangement of the sequence of operations is the arrangement of mathematical operations, not additional elements as articulated in BASCOM.
Applicant further asserted on page 21, “When considered as a whole, independent claim 1 recites a concrete technological implementation that improves how neural networks are calibrated. By determining an optimized truncated threshold that minimizes a total quantization difference metric derived from both noise amplitude and correlation characteristics, the claimed invention improves quantization parameter selection so as to maintain desired inference precision for image processing while reducing computation volume and storage requirements associated with high-precision floating-point representations. These improvements directly impact processor operation and resource utilization in devices executing neural networks and therefore constitute concrete technological benefits tied to processor functionality, rather than abstract mathematical results.”
Examiner respectfully disagrees because when considered as a whole, any arguably improvement is a direct consequence of performing the sequence of mathematical operations as recited in the claims, as explained above.
Applicant further asserted on page 21, “Moreover, the Office Action does not identify any prior art, evidence, or citation establishing that the specific quantization calibration framework recited in amended independent claim 1, is well-understood, routine, or conventional in the field. Notably, no rejection under 35 U.S.C. §§ 102 or 103 has been asserted against the previously presented claims. While patent eligibility is a distinct inquiry, the absence of any cited prior art or factual support demonstrating the claimed invention reinforces that there is no evidentiary basis for concluding that the recited limitations, alone or in ordered combination, are well-understood, routine, and conventional in the art.”
Examiner respectfully disagrees because MPEP 2106.05(I) explicitly recites “The search for a § 101 inventive concept is thus distinct from demonstrating § 102 novelty."). In addition, the search for an inventive concept is different from an obviousness analysis under 35 U.S.C. 103.” Furthermore, as explained above, what is not well-understood, routine, or conventional, such as the sequence of operations identified in Remarks page 18-19, is characterized as the abstract idea under step 2A prong one. Thus, under step 2B, Examiner is not required to show evidence for the abstract idea to be well-understood, routine, or conventional because the search for the inventive concept under step 2B is to evaluate the additional elements. MPEP 2106.04(I) “The Supreme Court’s decisions make it clear that judicial exceptions need not be old or long-prevalent, and that even newly discovered or novel judicial exceptions are still exceptions” and MPEP 2106.05(I) “a claim for a new abstract idea is still an abstract idea”.
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, 3-5,7-10, and 13 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Claim 1 recites a method for calibrating quantization in a neural network
Under Prong One of Step 2A of the USPTO current eligibility guidance (MPEP 2106), the claim recites limitations that cover mathematical calculations, relationship, and/or formula, such as a method for calibrating quantization in a neural network comprises quantizing the calibration data set by using a truncated threshold, wherein the calibration data set includes a plurality of batches of data, the calibration data set is to calibrate quantization noise, wherein the quantizing of the calibration data set comprises dividing the calibration data set into quantized part data and truncated part data based on the truncated threshold (see at least figure 6 step 620 [0083] describes the quantization may be performed using equation 10, see [0076] for equation 10, [0086] describes input data is divided into quantized part and truncated part with reference to the equations 3-4, and the limitation of the calibration data set is to calibrate quantization noise is merely recited as a result of performing the abstract idea of calibrating quantization); determining a total quantization difference metric for the quantization of the calibration data set based on a quantization difference metric of the quantized part data and a quantization difference metric of the truncated part data, wherein the determining of the total quantization difference metric comprises: determining the quantization difference metric for the quantized part data, determining the quantization difference metric for the truncated part data, the determination of the quantization difference metric for the quantized part data and the determination of the quantization difference metric for the truncated part data are based on at least two of a quantization noise amplitude, and a correlation coefficient of the quantization noise with the calibration data set, the quantization noise amplitude represents a difference in absolute numerical values of quantization errors, the correlation coefficient of the quantization noise with the calibration data set represents a relationship between different representations of the quantized part data and the truncated part data in terms of the quantization errors and distribution of the calibration data set with respect to the truncated threshold; and (see figure 6 step 630, [0086-0087] describe the step 630 is performed by using mathematical equations 3-9, and 11-12. [0086-0087] describes total quantization difference metric may be calculated with reference to equation 5, wherein equation 6-7 are used to determine quantization difference metric DistQ and DistC of the quantized data part DQ and truncated data part DC. [0075] also describes equations 6-7 calculate the quantization difference metric DistQ and quantization difference metric DistC based on noise amplitude AQ and correlation coefficient EQ); and determining an optimized truncated threshold based on the total quantization difference metric, wherein the optimized truncated threshold is used for quantizing the calibration data set, and the optimized truncated threshold minimizes the total quantization difference metric (see figure 6 step 640 [0088-0089] describe step 640 is performed by selecting the truncated threshold that minimizes the total quantization difference metrics. [0033] further describes neural network is a mathematical model that mimics structures and functions of a biological neural network. The limitation of the optimized truncated threshold minimizes the total quantization difference metric is merely recited as a result of performing the abstract idea of determining the optimized truncated threshold). Therefore, the claim includes limitations that fall within the “Mathematical Concepts” grouping of abstract ideas. Accordingly, the claim recites an abstract idea.
Under Prong Two of Step 2A, this judicial exception is not integrated into a practical application. The claim additionally recites a processor for calibrating quantization in a neural network, wherein the neural network is deployed in a device configured to perform image processing, execution of a neural network operation on the device configured to perform the image processing, and controlling an input/output unit to output the optimized truncated threshold. However, the additional elements are recited at a high level of generality, i.e., as computer component performing computer function of receiving, processing, and inputting/outputting data. Furthermore, the steps of receiving a calibration data set from a neural network and outputting the optimized result are considered as insignificant extra and post solution activities (e.g., mere data gathering). Moreover, the limitation of neural network employed in a device and execution of a neural network operation on the device configured to perform image processing is at most considered as mere generally linking the use the judicial exception into a particular technological environment or field of use, such as neural network for image processing. Such additional elements fail to provide a meaningful limitation on the judicial exception, and amount to no more than mere instructions to apply the exception using generic computer element. Thus, the claim is directed to an abstract idea.
Under Step 2B, as discussed with respect to Prong Two of Step 2A, the additional elements in the claim amount no more than mere instructions to apply the exception using a generic component. The same conclusion is reached in step 2B, i.e., mere instruction to apply an exception on a generic element cannot integrate a judicial exception into a practical application at step 2A or provide an inventive concept that is furnished by an element or combination of elements that is recited in the claim in addition to (beyond) the judicial exception. The steps of receiving the calibration data set and outputting the optimized truncated threshold are insignificant extra/post solution activities in step 2A, and are determined to be well-understood, routine, conventional activity in the field. Court decisions cited in MPEP 2106.05(d)(II) section (i), indicate that mere receiving or transmitting data over a network, is well-understood, routing, conventional function when it is claimed in a merely generic manner. Thus, the additional element fails to ensure the claim as a whole amount to significantly more than the judicial exception itself. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101.
Claim 3 further recites wherein the quantizing of the calibration data set by using the truncated threshold includes: quantizing the calibration data set by using a plurality of candidate truncated thresholds in a search space of the truncated threshold, respectively. Such limitations cover mathematical calculations, relationship, and/or formula under step 2A prong one ([0083] describes step S620 performs quantization on the input data D using the truncated threshold. The quantization of the input data may be performed in various manners. For example, the quantization may be performed using equation (10) as described above). Thus, the claim does not recite additional element that would integrate the judicial exception into a practical application under step 2A prong two or ensure the claim as a whole amount to significantly more than the judicial exception itself under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101.
Claim 4 further recites wherein the determining of the total quantization difference metric for the quantization of the calibration data set further includes: dividing, for each candidate truncated threshold in the plurality of candidate truncated thresholds, the calibration data set into the quantized part data and the truncated part data based on whether absolute values of data in the calibration data set are less than the candidate truncated threshold or greater than or equal to the candidate truncated threshold; determining the quantization difference metric of the quantized part data and the quantization difference metric of the truncated part data, respectively; and determining the corresponding total quantization difference metric based on the quantization difference metric and the quantization difference metric. Such limitations cover mathematical calculations, relationship, and/or formula under step 2A prong one (see figure 6 step 630 [0086-0087]). Thus, the claim does not recite additional element that would integrate the judicial exception into a practical application under step 2A prong two or ensure the claim as a whole amount to significantly more than the judicial exception itself under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101.
Claim 5 further recites wherein the quantization difference metric of the quantized part data and the quantization difference metric of the truncated part data is determined based on a quantization noise amplitude of the quantized part data and a correlation coefficient of a quantization noise of the quantized part data with the quantized part data, and the quantization difference metric of the truncated part data is determined based on a quantization noise amplitude of the truncated part data, and EC a correlation coefficient of a quantization noise of the truncated part data with the truncated part data. Such limitations cover mathematical calculations, relationship, and/or formula under step 2A prong one (perform mathematical operations to calculate DistQ and DistC, see at least [0075] equations 6 and 7). Thus, the claim does not recite additional element that would integrate the judicial exception into a practical application under step 2A prong two or ensure the claim as a whole amount to significantly more than the judicial exception itself under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101.
Claim 7 further recites wherein the total quantization difference metric is determined by: combining the quantization difference metric of the quantized part data and the quantization difference metric of the truncated part data. Such limitations cover mathematical calculations, relationship, and/or formula under step 2A prong one (performing combining operation or addition operation of DistQ and DISTc). Thus, the claim does not recite additional element that would integrate the judicial exception into a practical application under step 2A prong two or ensure the claim as a whole amount to significantly more than the judicial exception itself under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101.
Claim 8 further recites wherein the determining of the optimized truncated threshold based on the total quantization difference metric includes: selecting, from the plurality of candidate truncated thresholds, a candidate truncated threshold that minimizes the total quantization difference metric as the optimized truncated threshold. Such limitations cover mathematical calculations, relationship, and/or formula under step 2A prong one (see at least figure 6 step 640 [0088] describes determining the optimized truncated threshold). Thus, the claim does not recite additional element that would integrate the judicial exception into a practical application under step 2A prong two or ensure the claim as a whole amount to significantly more than the judicial exception itself under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101.
Claim 9 further recites wherein the search space of the truncated threshold is determined based on at least a maximum value of the calibration data set, and the candidate truncated threshold is determined at least partially based on preset search precision. Such limitations cover mathematical calculations, relationship, and/or formula under step 2A prong one (determining the search space of the truncated threshold and the candidate truncated threshold, see at least [0093-0095]). Thus, the claim does not recite additional element that would integrate the judicial exception into a practical application under step 2A prong two or ensure the claim as a whole amount to significantly more than the judicial exception itself under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101
Claim 10 further recites wherein the total quantization difference metric is based on a total quantization difference metric of all batches of data. Such limitations cover mathematical calculations, relationship, and/or formula under step 2A prong one (such as further describing the calibration data set and the calculation of the total quantization difference metric). Thus, the claim does not recite additional element that would integrate the judicial exception into a practical application under step 2A prong two or ensure the claim as a whole amount to significantly more than the judicial exception itself under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C. 101.
Claim 13 recites a method claim having similar limitation as the method clam 1. Thus, it is rejected for the same reasons.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HUY DUONG whose telephone number is (571)272-2764. The examiner can normally be reached Mon-Friday 7:30-5:30.
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, Andrew Caldwell can be reached at (571) 272-3702. 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.
/HUY DUONG/Examiner, Art Unit 2182 (571)272-2764
/ANDREW CALDWELL/Supervisory Patent Examiner, Art Unit 2182
Prosecution Insights