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 .
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
Step 1:
According to the first part of the analysis, in the instant case, claims 1-18 are directed to a method, claim 19 is directed to a non-transitory computer storage media storing instructions that when executed by one or more computers cause the one or more computers to perform the method, and claim 20 is directed to a system. Thus, each of the claims falls within one of the four statutory categories (i.e. process, machine, manufacture, or composition of matter).
Regarding claim 1:
A method performed by one or more data processing apparatus for generating one or more differentiable order statistics for a vector of scores, the method comprising:
obtaining the vector of scores, wherein each position in the vector of scores is associated with a respective index from a set of indices;
obtaining a plurality of pairs of indices, wherein each pair of indices comprises a respective first index and a respective second index from the set of indices;
generating a respective swapping probability for each pair of indices based on the vector of scores;
generating, for each pair of indices, a respective soft-swapping matrix for the pair of indices as a combination of: (i) an identity matrix that, if applied to the vector of scores, would leave the vector of scores unchanged, and (ii) an exchange matrix that, if applied to the vector of scores, would swap a pair of scores indexed by the pair of indices, wherein the exchange matrix is weighted in the combination by the swapping probability for the pair of indices; and
generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices.
Step 2A Prong 1:
“obtaining the vector of scores, wherein each position in the vector of scores is associated with a respective index from a set of indices” is directed to mental step of data gathering.
“obtaining a plurality of pairs of indices, wherein each pair of indices comprises a respective first index and a respective second index from the set of indices” is directed to mental step of data gathering.
“generating a respective swapping probability for each pair of indices based on the vector of scores” is directed to math because this process is often used in machine learning to determine the likelihood of swapping elements, where the "scores" are derived from a mathematical function, and the "swapping probability" is calculated using a function like the softmax function to convert the scores into a probability distribution.
“generating, for each pair of indices, a respective soft-swapping matrix for the pair of indices as a combination of: (i) an identity matrix that, if applied to the vector of scores, would leave the vector of scores unchanged, and (ii) an exchange matrix that, if applied to the vector of scores, would swap a pair of scores indexed by the pair of indices, wherein the exchange matrix is weighted in the combination by the swapping probability for the pair of indices” is directed to math because the concept is used in specific advanced or applied mathematical contexts, such as the generation of differentiable order statistics in machine learning and optimization problems. The process of generating such a matrix for "each pair of indices" involves systematic mathematical construction to ensure that all possible reordering combinations can be represented or approximated. This relies heavily on principles from matrix theory and combinatorics.
“generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices” is directed to math because this process is a mathematical concept used in machine learning to approximate non-differentiable functions like sorting. This technique uses differentiable matrices to perform operations that are analogous to sorting, allowing for backpropagation of gradients through the process. In mathematics, order statistics are the values of the observations which are arranged in ascending or descending order. The "soft-swapping matrices" are a way of representing the sorting/ranking operation using matrices, often as a relaxation of a permutation matrix. This leverages the mathematical properties of matrices and linear algebra to define the operation.
Each limitation recites in the claim is a process that, under BRI covers performance of the limitation in the mind but for the recitation of a generic “vector of scores” which is a mere indication of the field of use. Nothing in the claim elements precludes the steps from practically being performed in the mind. Thus, the claim recites a mental process.
Further, the claim recites the step of " generating a respective swapping probability for each pair of indices based on the vector of scores; generating, for each pair of indices, a respective soft-swapping matrix for the pair of indices as a combination of: (i) an identity matrix that, if applied to the vector of scores, would leave the vector of scores unchanged, and (ii) an exchange matrix that, if applied to the vector of scores, would swap a pair of scores indexed by the pair of indices, wherein the exchange matrix is weighted in the combination by the swapping probability for the pair of indices; and generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices” which as drafted, under BRI recites a mathematical calculation. The grouping of "mathematical concepts” in the 2019 PEG includes "mathematical calculations" as an exemplar of an abstract idea. 2019 PEG Section |, 84 Fed. Reg. at 52. Thus, the recited limitation falls into the "mathematical concept" grouping of abstract ideas. This limitation also falls into the “mental process” group of abstract ideas, because the recited mathematical calculation is simple enough that it can be practically performed in the human mind, e.g., scientists and engineers have been solving the Arrhenius equation in their minds since it was first proposed in 1889.
Note that even if most humans would use a physical aid (e.g., pen and paper, a slide rule, or a calculator) to help them complete the recited calculation, the use of such physical aid does not negate the mental nature of this limitation. See October Update at Section I(C)(i) and (iii).
Additional Elements:
Step 2A Prong 2:
“A method performed by one or more data processing apparatus for generating one or more differentiable order statistics for a vector of scores” recited in the preamble does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“generating a respective swapping probability for each pair of indices based on the vector of scores” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“generating, for each pair of indices, a respective soft-swapping matrix for the pair of indices as a combination of: (i) an identity matrix that, if applied to the vector of scores, would leave the vector of scores unchanged, and (ii) an exchange matrix that, if applied to the vector of scores, would swap a pair of scores indexed by the pair of indices, wherein the exchange matrix is weighted in the combination by the swapping probability for the pair of indices” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
The claim is merely obtaining data, manipulating or analyzing the data using math and mental process, and displaying the results.
This is similar to electric power: MPEP 2106.05(h) vi. Limiting the abstract idea of collecting information, analyzing it, and displaying certain results of the collection and analysis to data related to the electric power grid, because limiting application of the abstract idea to power-grid monitoring is simply an attempt to limit the use of the abstract idea to a particular technological environment, Electric Power Group, LLC v. Alstom S.A., 830 F.3d 1350, 1354, 119 USPQ2d 1739, 1742 (Fed. Cir. 2016).
Whether the claim invokes computers or other machinery merely as a tool to perform an existing process. Use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or 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. See Affinity Labs v. DirecTV, 838 F.3d 1253, 1262, 120 USPQ2d 1201, 1207 (Fed. Cir. 2016) (cellular telephone); TLI Communications LLC v. AV Auto, LLC, 823 F.3d 607, 613, 118 USPQ2d 1744, 1748 (Fed. Cir. 2016) (computer server and telephone unit). Similarly, "claiming the improved speed or efficiency inherent with applying the abstract idea on a computer" does not integrate a judicial exception into a practical application or provide an inventive concept. Intellectual Ventures I LLC v. Capital One Bank (USA), 792 F.3d 1363, 1367, 115 USPQ2d 1636, 1639 (Fed. Cir. 2015). In contrast, a claim that purports to improve computer capabilities or to improve an existing technology may integrate a judicial exception into a practical application or provide significantly more. McRO, Inc. v. Bandai Namco Games Am. Inc., 837 F.3d 1299, 1314-15, 120 USPQ2d 1091, 1101-02 (Fed. Cir. 2016); Enfish, LLC v. Microsoft Corp., 822 F.3d 1327, 1335-36, 118 USPQ2d 1684, 1688-89 (Fed. Cir. 2016). See MPEP §§ 2106.04(d)(1) and 2106.05(a) for a discussion of improvements to the functioning of a computer or to another technology or technical field.
The claim as a whole does not meet any of the following criteria to integrate the judicial exception into a practical application:
An additional element reflects an improvement in the functioning of a computer, or an improvement to other technology or technical field;
an additional element that applies or uses a judicial exception to effect a particular treatment or prophylaxis for a disease or medical condition;
an additional element implements a judicial exception with, or uses a judicial exception in conjunction with, a particular machine or manufacture that is integral to the claim;
an additional element effects a transformation or reduction of a particular article to a different state or thing; and
an additional element applies or uses the judicial exception in some other meaningful way beyond generally linking the use of the judicial exception to a particular technological environment, such that the claim as a whole is more than a drafting effort designed to monopolize the exception.
Step 2B:
“A method performed by one or more data processing apparatus for generating one or more differentiable order statistics for a vector of scores” recited in the preamble does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“generating a respective swapping probability for each pair of indices based on the vector of scores” does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“generating, for each pair of indices, a respective soft-swapping matrix for the pair of indices as a combination of: (i) an identity matrix that, if applied to the vector of scores, would leave the vector of scores unchanged, and (ii) an exchange matrix that, if applied to the vector of scores, would swap a pair of scores indexed by the pair of indices, wherein the exchange matrix is weighted in the combination by the swapping probability for the pair of indices” does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices” does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
The claim is therefore ineligible under 35 USC 101.
Claim 19 cites one or more non-transitory computer storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations for generating one or more differentiable order statistics for a vector of scores, the operations comprising the steps as in claim 1. This amounts to nothing more than instructions to implement the abstract idea on a computer, which fails to integrate the abstract idea into a practical application. See 2019 Guidance, 84 Fed. Reg. at 55. Additionally, using instructions to implement an abstract idea on a generic computer “is not ‘enough’ to transform an abstract idea into a patent-eligible invention.” Alice, 573 U.S. at 226. Therefore, the rejection of claim 14 for the same reason discussed above with regard to the rejection of claim 1.
Claim 20 recites a system comprising: one or more computers; and one or more storage devices communicatively coupled to the one or more computers, wherein the one or more storage devices store instructions that, when executed by the one or more computers, cause the one or more computers to perform operations for generating one or more differentiable order statistics for a vector of scores, the operations comprising the steps as in claim 1. These additional elements fail to integrate the abstract idea into a practical application. These limitations are recited at a high level of generality and do not add significantly more to the judicial exception. These elements are generic computing devices that perform generic functions. Using generic computer elements to perform an abstract idea does not integrate an abstract idea into a practical application. See 2019 Guidance, 84 Fed. Reg. at 55. Moreover, “the mere recitation of a generic computer cannot transform a patent-ineligible abstract idea into a patent-eligible invention.” Alice, 573 U.S. at 223; see also FairWarninglP, LLCv. latric SysInc., 839 F.3d 1089, 1096 (Fed. Cir. 2016) (citation omitted) (“[T]he use of generic computer elements like a microprocessor or user interface do not alone transform an otherwise abstract idea into patent-eligible subject matter”).
On the record before us, we are not persuaded that the hardware of claim 20 integrates the abstract idea into a practical application. Nor are we persuaded that the additional elements are anything more than well-understood, routine, and conventional so as to impart subject matter eligibility to claim 20.
Regrading claim 2, “wherein obtaining the vector of scores comprises generating each score in the vector of scores as an output of a machine learning model having a plurality of machine learning model parameters” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 3, “training the machine learning model parameters using an objective function that depends on the differentiable order statistics, comprising: determining gradients of the differentiable order statistics with respect to the machine learning model parameters; and updating values of the machine learning model parameters using the gradients of the differentiable order statistics with respect to the machine learning model parameters” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 4, “wherein the differentiable order statistics include a top-K order statistic that defines whether a target index is included in a set of indices of the K highest scores from the vector of scores” is directed to math because it provides a mathematical and computational framework to make "inclusion in the top-K set" a differentiable concept, allowing for its integration into calculus-based optimization problems.
Regarding claim 5, “wherein the machine learning model comprises a neural network” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 6, “wherein the respective swapping probability for each pair of indices is strictly greater than zero and strictly less than one” is directed to math.
Regarding claim 7, “wherein for each pair of indices, the soft-swapping matrix for the pair of indices is generated as: (1−ρ)I+ρE, wherein ρ is the swapping probability, I is the identity matrix, and E is the exchange matrix” is directed to math.
Regarding claim 8, “wherein the plurality of pairs of indices are associated with a sequential ordering, wherein the swapping probabilities for the pairs of indices are sequentially generated starting from a first pair of indices in the sequential ordering of the pairs of indices, and wherein for each pair of indices, generating the swapping probability for the pair of indices comprises: determining the swapping probability for the pair of indices based on a pair of scores indexed by the pair of indices in a current copy of the vector of scores; and applying a swapping operator associated with the pair of indices to the current copy of the vector of scores, wherein the swapping operator is from a sequence of swapping operators that define a sorting network” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 9, “wherein applying the swapping operator associated with the pair of indices to the current copy of the vector of scores comprises: evaluating an inequality between the pair of scores indexed by the pair of indices in the current copy of the vector of scores; and in response to determining that the inequality is satisfied, swapping the pair of scores indexed by the pair of indices in the current copy of the vector of scores” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 10, “wherein generating the swapping probability for a final pair of indices comprises applying a final swapping operator to the current copy of the vector of scores, and wherein after applying the final swapping operator, the scores in the current copy of the vector of scores are sorted” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 11, “wherein for each pair of indices, determining the swapping probability for the pair of indices based on the pair of scores indexed by the pair of indices in a current copy of the vector of scores, comprises: determining a difference between the pair of scores indexed by the pair of indices in the current copy of the vector of scores; determining a product of an inverse of a dispersion factor and the difference between the pair of scores indexed by the pair of indices in the current copy of the vector of scores; and determining the swapping probability by applying a sigmoid function to a result of the product” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 12, “obtaining a sorted version of the vector of scores; and wherein for each pair of indices, determining the swapping probability for the pair of indices based on the pair of scores indexed by the pair of indices in the current copy of the vector of scores, comprises: determining an error between: (i) the pair of scores indexed by the pair of indices in the current copy of the vector of scores, and (ii) a pair of scores indexed by the pair of indices in the sorted version of the vector of scores; and determining the swapping probability based on the error” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 13, “wherein for each pair of indices, determining the error between: (i) the pair of scores indexed by the pair of indices in the current copy of the vector of scores, and (ii) the pair of scores indexed by the pair of indices in the sorted version of the vector of scores, comprises: determining a cost matrix [C.sub.ij].sub.i,j=1,2, where C.sub.ij=h(y.sub.i−x.sub.j), [x.sub.1, x.sub.2] are the pair of scores indexed by the pair of indices in the current copy of the vector of scores, [y.sub.1, y.sub.2] are the pair of scores indexed by the pair of indices in the sorted version of the vector of scores, and h(⋅) is a convex and non-negative function” is directed to math.
Regarding claim 14, “wherein for each pair of indices, determining the swapping probability based on the error comprises determining the swapping probability λ as:
PNG
media_image1.png
93
314
media_image1.png
Greyscale
where ϵ is a dispersion factor” is directed to math.
Regarding claim 15, “wherein generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices comprises: generating a matrix product of the soft-swapping matrices; and generating the one or more differentiable order statistics based on the matrix product of the soft-swapping matrices” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 16, wherein the one or more differentiable order statistics comprise a sorted version of the vector of scores, and wherein generating the sorted version of the vector of scores comprises: determining a product of: (i) the matrix product of the soft-swapping matrices, and (ii) the vector of scores does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h))..
Regarding claim 17, “wherein the one or more differentiable order statistics comprise a maximum score in the vector of scores, and wherein generating the maximum score in the vector of scores comprises: determining a product of: (i) a transpose of a final column of an identity matrix, (ii) the matrix product of the soft-swapping matrices, and (iii) the vector of scores” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Regarding claim 18, “wherein the one or more differentiable order statistics comprise a minimum score in the vector of scores, and wherein generating the minimum sore in the vector of scores comprises: determining a product of: (i) a transpose of a first column of an identity matrix, (ii) the matrix product of the soft-swapping matrices, and (iii) the vector of scores” does not integrate the judicial exception into a practical application. It does not amount to significantly more than the judicial exception in the claim. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Hence the claims 1-20 are treated as ineligible subject matter under 35 U.S.C. § 101.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1-8, 15-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Blondel et al. ("Fast Differentiable Sorting and Ranking") in view of Anderson et al. (US 2020/0372099 A1).
Regarding claims 1, 19, and 20, Blondel et al. disclose a method performed by one or more data processing apparatus for generating one or more differentiable order statistics for a vector of scores (abstract: "differentiable sorting and ranking operators”, "ranking operator, often used for order statistics and ranking metrics", page 1, right column, 2nd paragraph: "input vector with its values"), the method comprising:
obtaining the vector of scores, wherein each position in the vector of scores is associated with a respective index from a set of indices (page 1, right column, 3rd paragraph, "positions, or ranks, of the input values in the sorted vector'');
obtaining a plurality of pairs of indices, wherein each pair of indices comprises a respective first index and a respective second index from the set of indices; generating a respective swapping probability for each pair of indices based on the vector of scores (page 1, right column, 3rd paragraph, "a simple method based on comparing pairwise distances between values", said distances corresponding to swapping probabilities for pairs of indexes; page 6, 1st paragraph of section 6.1, bullet point starting with ''All-pairs" and specifying the probability function for each pair);
generating, for each pair of indices, a respective soft-swapping matrix for the pair of indices as a combination of: (i) an identity matrix that, if applied to the vector of scores, would leave the vector of scores unchanged (page 6, "Proposition 4"); and
generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices (abstract, "order statistics", page 6, left column, last paragraph, "Jacobians of our soft operators").
Blondel et al. is silent on teaching an alternative calculation for the soft-swapping matrix.
Anderson et al. teach an alternative calculation for the soft-swapping matrix (para. [0102]: data element sorting/swapping such as matrix transposition).
Therefore, it would have been obvious to one of ordinary skill in the art at the time of the invention was made to incorporate an alternative calculation for the soft-swapping matrix of Anderson et al. with the method of Blondel et al. for the purposes of providing a methods and apparatus for vector sorting (Anderson et al., para. [0004]).
Regarding claims 2-5, Blondel et al. disclose wherein obtaining the vector of scores comprises generating each score in the vector of scores as an output of a machine learning model having a plurality of machine learning model parameters; training the machine learning model parameters using an objective function that depends on the differentiable order statistics, comprising: determining gradients of the differentiable order statistics with respect to the machine learning model parameters; and updating values of the machine learning model parameters using the gradients of the differentiable order statistics with respect to the machine learning model parameters; wherein the differentiable order statistics include a top-K order statistic that defines whether a target index is included in a set of indices of the K highest scores from the vector of scores; wherein the machine learning model comprises a neural network (see at least page 1, 1st and 2nd paragraphs of chapter 1, "deep learning architectures", "trained end to-end using gradient backpropagation", "many operations commonly used in computer
programming remain poorly differentiable or downright pathological, limiting the set of architectures for which a gradient can be computed. We focus in this paper on two such operations"; page 3, left column, 2nd paragraph, "use within a neural network trained with backpropagation'').
Regarding claim 6, Blondel et al. disclose the claimed invention except wherein the respective swapping probability for each pair of indices is strictly greater than zero and strictly less than one. It would have been an obvious matter of design choice to the respective swapping probability for each pair of indices is strictly greater than zero and strictly less than one, since such a modification would have involved a mere change in the size of a component. A change in size is generally recognized as being within the level of ordinary skill in the art. In re Rose, 105 USPQ 2)1 (CCPA 1955).
Regrading clam 7, although Blondel et al. is silent on teaching an alternative calculation for the soft-swapping matrix, wherein for each pair of indices, the soft-swapping matrix for the pair of indices is generated as: (1−ρ)I+ρE, wherein ρ is the swapping probability, I is the identity matrix, and E is the exchange matrix. However, implementing such an alternative has no further technical effect is a mere choice of a mathematical algorithm not having a credible inventive contribution over Grover et al. or any other prior art providing relaxed/ soft swapping/permutation matrices. It would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to provide an alternative calculation for the soft-swapping matrix because the method processing of Grover et al. is the same.
Regarding claim 15, Blondel et al. teach wherein generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices comprises: generating a matrix product of the soft-swapping matrices; and generating the one or more differentiable order statistics based on the matrix product of the soft-swapping matrices (abstract, "order statistics", page 6, left column, last paragraph, "Jacobians of our soft operators").
Regarding claims 16-18, Blondel et al. teach wherein the one or more differentiable order statistics comprise a sorted version of the vector of scores, and wherein generating the sorted version of the vector of scores comprises: determining a product of: (i) the matrix product of the soft-swapping matrices, and (ii) the vector of scores, wherein the one or more differentiable order statistics comprise a maximum score in the vector of scores, and wherein generating the maximum score in the vector of scores comprises: determining a product of: (i) a transpose of a final column of an identity matrix, (ii) the matrix product of the soft-swapping matrices, and (iii) the vector of scores, wherein the one or more differentiable order statistics comprise a minimum score in the vector of scores, and wherein generating the minimum sore in the vector of scores comprises: determining a product of: (i) a transpose of a first column of an identity matrix, (ii) the matrix product of the soft-swapping matrices, and (iii) the vector of scores. (page 1, right column, 3rd paragraph, "positions, or ranks, of the input values in the sorted vector'', page 6, "Proposition 4").
Claim(s) 1-8, 19-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Grover et al. ( "Stochastic Optimization of Sorting Networks via Continuous Relaxations") in view of Anderson et al. (US 2020/0372099 A1).
Regarding claims 1, 19, and 20, Grover et al. disclose a method performed by one or more data processing apparatus for generating one or more differentiable order statistics for a vector of scores (abstract and see at least Fig. 5 on page 9, "top-k"}, the method comprising:
obtaining the vector of scores, wherein each position in the vector of scores is associated with a respective index from a set of indices (page 2, chapter 2, 1st and 2nd paragraphs, "the input vectors= [9, 1, 5, 2)1", "unique indices {1, 2, ... , n}");
obtaining a plurality of pairs of indices, wherein each pair of indices comprises a respective first index and a respective second index from the set of indices; generating a respective swapping probability for each pair of indices based on the vector of scores; (page 5, equation(4));
generating, for each pair of indices, a respective soft-swapping matrix for the pair of indices (page 3, chapter 3, 3rd paragraph, "use such a relaxation to replace the permutation matrix Pz in Eq. 2 with an approximation"; page 5, equation (5)); and
generating the one or more differentiable order statistics for the vector of scores using the soft-swapping matrices (see at least Fig. 5 on page 9, "top-k'').
Grover et al. is silent on teaching an alternative calculation for the soft-swapping matrix.
Anderson et al. teach an alternative calculation for the soft-swapping matrix (para. [0102]: data element sorting/swapping such as matrix transposition).
Therefore, it would have been obvious to one of ordinary skill in the art at the time of the invention was made to incorporate an alternative calculation for the soft-swapping matrix of Anderson et al. with the method of Grover et al. for the purposes of providing a methods and apparatus for vector sorting (Anderson et al., para. [0004]).
Regarding claims 2-5, Grover et al. disclose wherein obtaining the vector of scores comprises generating each score in the vector of scores as an output of a machine learning model having a plurality of machine learning model parameters; training the machine learning model parameters using an objective function that depends on the differentiable order statistics, comprising: determining gradients of the differentiable order statistics with respect to the machine learning model parameters; and updating values of the machine learning model parameters using the gradients of the differentiable order statistics with respect to the machine learning model parameters; wherein the differentiable order statistics include a top-K order statistic that defines whether a target index is included in a set of indices of the K highest scores from the vector of scores; wherein the machine learning model comprises a neural network (page 1, 2nd paragraph, ''permutation matrices to evaluate learning objectives"; page 3, 1st paragraph of chapter 3, "Our goal is to optimize training objectives involving a sort operator with gradient-based methods"; paragraph bridging pages 3 and 4, "evaluating the learning objective in the forward pass and the relaxed operator is used in the
backward pass for gradient estimation"; see at least fig 4 and 5 for neural networks).
Regarding claim 6, Grover et al. disclose the claimed invention except wherein the respective swapping probability for each pair of indices is strictly greater than zero and strictly less than one. It would have been an obvious matter of design choice to the respective swapping probability for each pair of indices is strictly greater than zero and strictly less than one, since such a modification would have involved a mere change in the size of a component. A change in size is generally recognized as being within the level of ordinary skill in the art. In re Rose, 105 USPQ 2)1 (CCPA 1955).
Regrading clam 7, although Grover et al. is silent on teaching an alternative calculation for the soft-swapping matrix, wherein for each pair of indices, the soft-swapping matrix for the pair of indices is generated as: (1−ρ)I+ρE, wherein ρ is the swapping probability, I is the identity matrix, and E is the exchange matrix. However, implementing such an alternative has no further technical effect is a mere choice of a mathematical algorithm not having a credible inventive contribution over Grover et al. or any other prior art providing relaxed/ soft swapping/permutation matrices. It would have been obvious to one of ordinary skills in the art before the effective filing date of the claimed invention to provide an alternative calculation for the soft-swapping matrix because the method processing of Grover et al. is the same.
Allowable Subject Matter
Claims 8-14 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims and amended to overcome the rejection(s) under 35 U.S.C. 101, set forth in this Office action.
The following is a statement of reasons for the indication of allowable subject matter:
Regarding claim 8, none of the prior art of record teaches or suggests wherein the plurality of pairs of indices are associated with a sequential ordering, wherein the swapping probabilities for the pairs of indices are sequentially generated starting from a first pair of indices in the sequential ordering of the pairs of indices, and wherein for each pair of indices, generating the swapping probability for the pair of indices comprises: determining the swapping probability for the pair of indices based on a pair of scores indexed by the pair of indices in a current copy of the vector of scores; and applying a swapping operator associated with the pair of indices to the current copy of the vector of scores, wherein the swapping operator is from a sequence of swapping operators that define a sorting network. It is these limitations as they are claimed in the combination with other limitations of claim, which have not been found, taught or suggested in the prior art of record, that make these claims allowable over the prior art.
Other Prior Art
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure:
Refaat et al. (US 11,657,268) disclose a method of training a neural network having a plurality of network parameters and configured to receive a network input and to process the network input in accordance with the network parameters to generate a network output that assigns a respective score to each of a plurality of locations in the network input, the method comprising: obtaining a training input and a corresponding ground truth output that assigns a respective ground truth score for each of the plurality of locations in the training input, wherein the plurality of locations in the training input include corresponding locations of one or more agents that are in an environment within a vicinity of a vehicle, and wherein the respective ground truth scores include a respective importance score for each of the one or more agents; processing the training input using the neural network and in accordance with current values of the network parameters to generate a training output that assigns a respective training score to each of the plurality of locations in the training input; computing a loss for the training input, comprising: selecting a plurality of candidate locations from the plurality of locations; setting to zero the training scores for any location in the selected candidate locations that has a ground truth score below a threshold value; for each of a plurality of pairs of locations in the selected candidate locations: computing a pair-wise loss for the pair based on the ground truth scores and the training scores for the pair of locations; and combining the pair-wise losses for the pairs of locations to compute the loss for the training input; and determining an update to the current values of the parameters by determining a gradient of the loss with respect to the network parameters.
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOHN H LE whose telephone number is (571)272-2275. The examiner can normally be reached on Monday-Friday from 7:00am – 3:30pm Eastern Time.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Shelby A. Turner can be reached on (571) 272-6334. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/JOHN H LE/Primary Examiner, Art Unit 2857