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 Applicant’s amendment filed on 03/09/2026. Claims 1, 3-4, and 6-7 are pending.
Response to Arguments
In response to Applicant’s argument regarding rejection 35 U.S.C. 101 on page 6, “In the present case, amended independent claim 1 does not recite any mathematical formulas, equations, or abstract mathematical relationships. Instead, the claim defines a computer- implemented data normalization process performed during forward propagation of a deep learning neural network model deployed on an artificial intelligence processor chip.”
Examiner respectfully disagrees because the claim recites limitations that cover the mathematical formulas, equations, or abstract mathematical relationships under step 2A prong one, such as normalization operation includes, determining a scaling factor, determining a first product of the scaling factor and the input data, and executing a method of operator splice (see at least specification page 6 describes mathematical equation for determining scaling factor
β
, and see page 7 describes mathematical equation for calculating normalized result y using method of operator splicing as illustrated in figure 3). The additional elements recited in the claim, such as a computer-implemented method for data normalization in a deep learning neural network model deployed on an artificial intelligence processor chip is merely recited at a high level of generality, e.g., computer component performing computer function that is amount to no more than mere instructions to apply the judicial exception (e.g., normalization operation in a deep learning neural network model) using computer component (e.g., AI processor chip) and further considered as mere generally linking the use of the judicial exception (e.g., normalization operation) into a particular technological environment or field of use, such as deep learning neural network for AI chip. Thus, such additional elements fail to integrate the judicial exception into a practical application under step 2A prong two or provide an inventive concept under step 2B.
Applicant further asserted on page 6-7, “More specifically, claim 1 recites a sequence of implementation-level operations including: receiving a digital image in a specific tensor format (Number Channel Height Width, NCHW) as input data; determining a scaling factor of the input data based on a maximum value of quantized data type of the input data and a maximum value of the input data; determining a first product of the scaling factor and the input data; executing a method of operator splicing that processes the first product through a sequence of tensor operations to determine a second product; and determining a normalization result corresponding to an L2 normalization operator, wherein the normalization result is the second product produced by the operator splicing process. These limitations define how normalization is concretely carried out within a neural network execution pipeline, including the use of a particular data layout, quantized data constraints, tensor operations, and an operator splicing mechanism that replaces a conventional normalization operator. The claim is thus directed to a specific computational workflow implemented within a machine-learning execution environment, rather than to a mathematical concept performed in the abstract.
Examiner respectfully disagrees because a sequence of implementation level operations, such as determining scaling factor, determining, a first product, executing a method of operator splicing to determine a second product, and determining a normalization result is characterized as mathematical concept under step 2A prong one as such limitations are merely reciting a sequence of mathematical operations as described on page 6-7 and figure 3 of the specification. Even though such sequence of operations for the normalization operation is novel and replaces the conventional normalization operator, such sequence of operations is still abstract idea (see 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, new math is still math. Furthermore, the claim recites the step of receiving digital image, but such step of receiving data is recited at a high level of generality and at most considered as insignificant extra solution activity under step 2A prong two as mere data gathering and determined to be well-understood, routine, and conventional activity under step 2B (see MPEP 2106.05(d)(II)(i) receiving or transmitting data over a network is conventional activity as recognized by the Courts), and the data format and digital image recited in the claim merely describe the format and data type of the input data being performed in the mathematical operations.
Applicant further asserted on page 7, “While the claimed method may employ arithmetic operations as part of its execution, such operations are embedded within and subordinate to a larger technological process for executing neural network normalization. The claim does not seek to monopolize any mathematical principle, nor does it describe a calculation segregated from a technological context. Instead, it recites concrete processing steps tied to data normalization in a deep learning neural network model deployed on an artificial intelligence processor chip.”
Examiner respectfully disagrees because the claim recites a forward propagation process of a deep learning neural network model, which is a mathematical algorithm/model under step 2A prong one, and wherein the model deployed on an artificial intelligence (AI) processor chip, which is recited at a high level of generality, e.g., computer component performing computer function. Alternatively, when a forward propagation process of a deep learning neural network model employed on an AI processor chip is characterized as additional element, such recitation is at most considered as generally linking the use of the judicial exception to a particular technological environment or field of use, such as the field of neural network or AI (see MPEP 2106.05(h)).
Applicant further asserted on page 7, “the Examiner, on pages 4 and 5, relies heavily on equations and mathematical operations disclosed in the Specification to conclude that the claim covers "mathematical calculations." Applicant respectfully submits that this analysis is inconsistent with MPEP guidance. Under MPEP § 2106.04(a), the determination of whether a claim recites a mathematical concept must be based on what is actually claimed, not on unclaimed embodiments, figures, or equations described in the Specification … necessarily involves arithmetic at some level.”
Examiner respectfully disagrees because MPEP 2106.04(a)(2)(C) states “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.” Thus, the step of determining a scaling factor of the input data when interpreted in light of the specification cover mathematical calculation, and the step of determining a product of the scaling factor and the input data is also considered as mathematical calculation because in mathematics, a product of A and B means a multiplication of A and B. Furthermore, the steps of determining a scaling factor, a first product, and a second product using a method of operator slicing are operations within a normalization operation, which is also a mathematical operation to normalize data. Accordingly, under BRI in light of the specification, such limitations cover mathematical concept grouping of the abstract idea, and the claim recites mathematical operations, not just merely involve mathematical operation.
Applicant further asserted on page 8-9, “The claim, as a whole, imposes meaningful limits on any such mathematical concept by embedding it within a specific data normalization processing pipeline executed in a normalization lay of a deep learning neural network model”.
Examiner respectfully disagrees because merely recite performing a method of normalization in a normalization layer of a neural network model does not provide a meaningful limits on the mathematical concept, but it mere generally linking the use of the judicial exception to a particular technological environment or field of use, such as the field of neural network or AI (see MPEP 2106.05(h)).
Applicant further asserted on page 10-12 that amended claim provide a technological improvement that integrate the judicial exception into a practical application.
Examiner respectfully disagrees because any arguably improvements, such as computation accuracy and data overflow prevention effect, are a direct consequence of performing the mathematical concept as recited in claim 1 and as recognized by the applicant in page 11-12, see page 6 line 19-21 describes that “By setting the scaling factor and introducing the maximum value of the quantized data type, the computation accuracy of scaling factor can be improved, the computation amount of scaling factor can be reduced, and the data overflow prevention effect can be improved. MPEP 2106.05(a) “It is important to note, the judicial exception alone cannot provide the improvement. The improvement can be provided by one or more additional elements. See the discussion of Diamond v. Diehr, 450 U.S. 175, 187 and 191-92, 209 USPQ 1, 10 (1981)) in subsection II, below. In addition, the improvement can be provided by the additional element(s) in combination with the recited judicial exception.” Accordingly, the setting scaling factor and introducing the maximum value of the quantized data type (e.g., judicial exception) alone cannot provide the improvement (e.g., improve computation accuracy and data overflow prevention effect). Therefore, the claimed solution is rooted in the mathematical algorithm of normalizing operation to minimize overflow of the normalized input data rather than rooted in computer technology. Furthermore, the executing normalization using a method of operator splicing comprises processing the first product through a sequence of tensor operations is also considered as mathematical concept, thus the arguably improvements of reduces the amount of computation, complexity of normalization are also a direct consequence of performing such method of operator splicing.
Applicant further asserted on page 13, “Amended claim 1 does not merely recite "setting a scaling factor". Instead, amended independent claim 1 recites a forward-propagation execution sequence in a deep learning neural network, deployed in an AI processor chip, including operator splicing that processes first product through a sequence of tensor operations to produce the normalization result. The Examiner's analysis improperly treats the scaling factor determination as the claimed invention itself, while disregarding the remaining limitations that define how normalization is actually carried out within a machine-learning execution pipeline. When viewed as an ordered combination, the claim is directed to a specific implementation of deep neural network in AI processor chips, not to a mathematical concept performed in the abstract. Accordingly, the Examiner's conclusion that the claim is "rooted in the mathematical algorithm" is based on an incomplete and legally improper characterization of the claim.”
Examiner respectfully disagrees because the limitation of a forward propagation process of the deep learning neural network model employed by AI processor chip is at most considered as mere generally linking the use of the judicial exception (e.g., normalization operation) into a technological environment or field of use, which is neural network model for AI processor chip. Furthermore, the limitation of operator splicing that processes first product through a sequence of tensor operations to produce the normalization result is characterized as the abstract idea under step 2A prong one. Thus, when analyze the claim as a whole under step 2A prong two and step 2B, the claims do not integrate the judicial exception into a practical application or provide significantly more. Moreover, Examiner did not treat the scaling factor determination as the claimed invention itself, but Examiner analyzed the claimed as a whole and determined that the arguably improvements comes from the mathematical concept of performing normalization operation (e.g., judicial exception) as explained above, and MPEP 2106.06(a) states that judicial exception alone cannot provide the improvement.
Applicant further asserted on page 14, “According to the MPEP, when evaluating additional elements to determine whether they amount to an inventive concept … Even assuming that certain individual operations recited in claim 1, such as scaling, multiplication, or normalization, may have been known in isolation, amended independent claim 1 does not claim those operations in isolation. Rather, claim 1 recites a non-generic and non- conventional execution arrangement for performing normalization within a deep learning neural network model deployed on an AI processor chip”
Examiner respectfully disagrees because as applicant recited above and MPEP 2106.05 requires the search for the inventive concept is to evaluate additional elements of the claim. However, the operations, such as scaling, multiplication, or normalization recited in the claim, are characterized as the abstract idea under step 2A prong one, rather than the additional elements under step 2A prong two. Accordingly, even though the claim recites those operations for perform normalization in a novel way or arrangement, they are still abstract idea (see 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”).
Applicant further asserted on page 15, “This ordered combination of limitations defines a specific hardware-aware execution strategy for normalization within a deep learning neural network model deployed on an Al processor chip. The claim therefore recites how normalization is carried out on an AI processor chip during model execution, not merely the mathematical concept of normalization itself. These limitations are not well-understood, routine, or conventional activities in the field, but instead recite an unconventional execution arrangement that confines the claimed normalization to a particular, hardware-adapted implementation on an AI processor chip, consistent with the inventive-concept framework articulated in BASCOM and MPEP § 2106.05. Consistent with BASCOM, the inventive concept here lies in the non-conventional and non-generic ordered combination of scaling, operator splicing, and tensor operations that together implement normalization in a manner adapted to Al processor chip constraints. When considered as a whole, independent claim 1 recites a technological solution that improves the operation and execution efficiency of the AI processor chip by defining a specific normalization execution architecture, rather than claiming an abstract mathematical idea in isolation.”
Examiner respectfully disagrees because the AI processor chip is recited at a high level of generality and amount to no more than instructions to apply the judicial exception using computer component, rather than specific or particular hardware implementation. Furthermore, as explained above, the search for the inventive concept under step 2B is to evaluate additional elements, not abstract idea, thus, the non-conventional and non-generic ordered combination of scaling, operator splicing, and tensor operations together implement normalization is still abstract idea under step 2A prong one, and cannot be used in the search for an inventive concept under step 2B. Accordingly, the claimed invention is not consistent with BASCOM because in BASCOM, the inventive concept lies in the non-conventional and non-generic of the additional element, whereas the instant claim is referring to the combination of abstract idea.
Applicant further asserted on page 15-16, “Moreover, the Office Action does not identify any prior art, evidence, or citation establishing that the claimed execution arrangement, particularly the hardware-aware scaling, operator splicing. and tensor operation sequence, 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 explicitly states “The search for a 101 inventive concept is thus distinct from demonstrating 102 novelty”. Furthermore, even though the combination of operations, such as scaling, operator splicing and tensor operations sequence are not well-understood, routine, and conventional, they are still abstract idea under step 2A prong one. 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 order words, even though the combination of operations are novel, but they are still judicial exception, and under step 2B, the search for inventive concept is to evaluate whether the additional elements is well-understood, routine, and conventional, not the judicial exception.
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, 3-4, and 6-7 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract without significantly more.
Claim 1 recites a method for performing data normalization processing.
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 data normalization comprising in a forward propagation process of the deep learning neural network model, executing a normalizing operation on the input data, when a format of the digital image is in a Number Channel Height Width (NCHW) format, wherein the NCHW format of the digital image is associated with a batch of the digital image, a channel of the digital image, a height of the digital image, and a width of the digital image, wherein the normalizing operation includes: determining a scaling factor of the input data according to a maximum value of quantized data type of the input data and a maximum value of the input data, wherein the maximum value of the quantized data type refers to a maximum value within a value range that is represented by a data type of the input data; determining a first product of the scaling factor and the input data; and executing a method of operator splicing to determine a second product based on the first product; wherein he method of operator splicing comprises processing the first product through a sequence of tensor operations and determining a normalization result corresponding to an L2normalization operator in a normalization layer of the deep learning neural network model, wherein the normalization result is the second product. Such limitations cover mathematical calculations, relationship, and/or formula (see at least figure 2, specification page 6 line 10-15 discloses formula for computing scaling factor, also see page 5 -13-15 describes the maximum value of the quantized data, page 7 line 3-5 discloses formula for calculating the L2 normalization operator in a normalization layer of a deep learning neural network, wherein a forward propagation process of a deep learning neural network is a mathematical algorithm/model that processes data, and the data operated on are data having 4 dimensional data format. Also see figure 3 page 8-9 describing the operation of a method of operator splicing. The digital image is merely recited as a data type to be processed using the mathematical operations). 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 computer-implemented method deployed on an artificial intelligence (AI) processor chip comprising receiving a digital image as input data. However, the additional element is recited at a high level of generality, i.e., as a computer implemented method that receiving data, wherein the step of receiving is considered as insignificant extra solution activity because it amounts to mere data gathering. Alternatively, the claim recites a deep learning neural network model deployed on an AI processor chip, a forward propagation process of the deep learning neural network model and the digital image as input data, such limitations are recited at a high level of generality and at most are considered as mere generally linking the use of the judicial exception (e.g., normalization operation) into a particular technological environment or field of use, such as image processing for deep neural network and AI. Such element fails to provide a meaningful limitation on the judicial exception, and amount to no more than mere instructions to apply the exception using generic computer. Thus, the claim is directed to an abstract idea.
Under Step 2B, the claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. As discussed previously with respect to the step 2A prong two, the additional elements in the claim amount to no more mere instructions to apply the exception and the step of receiving data is determined to be well-understood, routine and conventional (see MPEP 2106.05(d)(II)(i) Receiving or transmitting data over a network). Thus, the claim does not 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, and 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 the method of the operator splicing that includes performing a squaring operation, computing a first square, using an addition operation, computing a reciprocal, using a broadcast multiplication and taking the second product as the normalization result. Such limitation covers the mathematical calculations, relationship, and/or formula of performing squaring, addition, and broadcast multiplication operations under step 2A prong one (see at least figure 3) and does not provide any additional elements that would integrate the judicial exception into a practical application under step 2A prong two 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 under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C 101.
Claim 4 further recites the step of performing broadcast multiplication. Such limitation covers the mathematical calculations, relationship, and/or formula of performing summing, multiplying, dividing operations (see at least page 9 line 15-25 describes the specific method of using the broadcast multiplication to compute the second product) under step 2A prong one and does not provide any additional elements that would integrate the judicial exception into a practical application under step 2A prong two 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 under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C 101.
Claim 6 further recites operation modes of the L2normalization operator include an instance mode and a channel mode. Such limitation covers the mathematical calculations, relationship, and/or formula of performing different modes of the L2Normalization operator (see at least page 8 line 1-6 describes when input data is in NCHW format, that refers to instance mode and page 10 line 3-9 describes input in NHWC format, which refers to a channel mode) under step 2A prong one and does not provide any additional elements that would integrate the judicial exception into a practical application under step 2A prong two 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 under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C 101.
Claim 7 further recites the computer-implemented method for quantizing the input data. Such limitation covers the mathematical calculations, relationship, and/or formula of performing quantizing the input data (see at least page 5 line 17- page 6 line 9 describes the mathematical formula of performing quantizing input data) under step 2A prong one and does not provide any additional elements that would integrate the judicial exception into a practical application under step 2A prong two 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 under step 2B. Accordingly, the claim is not patent-eligible under 35 U.S.C 101.
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
/ANDREW CALDWELL/Supervisory Patent Examiner, Art Unit 2182
(571)272-2764
Prosecution Insights