DETAILED ACTION
Status of Claims
This Office action is responsive to communications filed on 2025-08-08. Claim(s) 2-6, 8, 13 were cancelled. Claim(s) 23-26 were added. Claim(s) 1, 7, 9-12, and 14-26 is/are pending and are examined herein.
Claim(s) 1, 7, 9-12, and 14-26 is/are objected to.
Claim(s) 10 is/are rejected under 35 USC 112(d).
Claim(s) 7, 9, 11-12, and 18 is/are rejected under 35 USC 112(b).
Claim(s) 7 is/are rejected under 35 USC 112(a).
Claim(s) 1, 7, 9-12, and 14-26 is/are rejected under 35 USC 101.
Claim(s) 1, 7, 9, 11-12, and 21-22 is/are rejected under 35 USC 102.
Claim(s) 10 and 14-26 is/are rejected under 35 USC 103.
Notice of Pre-AIA or AIA Status
The present application, filed on or after 2013-03-16, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The attached information disclosure statement(s) (IDS), submitted on 2025-05-08, is/are in compliance with the provisions of 37 CFR 1.97. Accordingly, the attached information disclosure statement(s) is/are being considered by the examiner.
Response to Arguments
Regarding objections for informalities and rejections under 35 USC 112, the applicant’s amendments resolve concerns discussed previously but also raise new concerns. Concerns about the pending claims are described below.
Regarding rejections under 35 USC 101, the applicant’s arguments have been fully considered but they are unpersuasive:
The applicant asserts that various operations “cannot be reasonably performed mentally and require computer implementation” [remarks, page 18] but the examiner disagrees. None of the limitations in the claim are recited narrowly enough so as to require computer implementation.
The applicant attempts various arguments attempting to argue that the claims provide integrate the abstract ideas into a practical application by providing an improvement [remarks, pages 18-20] (e.g. “[t]he interplay between grouping, model structure, and evaluation defines a technical improvement in model selection workflows” [remarks, page 18], “[t]he steps form a coherent technical solution to an optimization problem in predictive modeling” [remarks, page 18] or “[these claims] implement a sequence of technically meaningful processes to optimize model structure based on model data” [remarks, page 18], “the present claims as a whole provide an improvement to this technological environment” [remarks, page 20]). However, MPEP 2106.05(a) indicates that one of the requirements of the improvements analysis is that a “judicial exception alone cannot provide the improvement”. In the present case, every limitation of the independent claims (possibly except for recitations of generic computing equipment and of obtaining data on which to operate) is an abstract idea, so any purported improvements would necessarily be provided by judicial exceptions. Consequently, the invention as claimed does not meet the requirements of the improvements analysis.
The complete 101 analysis, updated in view of the applicant’s amendments, is given below.
Regarding rejections under 35 USC 102/103 of the originally filed claims, the applicant “traverses this ground of rejection” [remarks, page 20] but provides no rationale in support of this traversal. This bare assertion of traversal fails to comply with 37 CFR 1.111(b) because it amounts to a general allegation that the claims define a patentable invention without specifically pointing out how the language of the originally filed claims patentably distinguishes them from the references. Regarding the amended claims, the applicant’s remarks have been fully considered but they are unpersuasive:
Regarding claim 1, the applicant argues that “[c]laim 1 introduces specific time-based grouping” and that “Mavrovouniotis… does not disclose these operations, especially in the context of time-indexed variable grouping” [remarks, page 21]. However, Mavrovouniotis clearly discloses time-based grouping of time-indexed variables [Mavrovouniotis, sections 2.2-2.3]. The applicant also argues that “[c]laim 1 also now incorporates the contents of cancelled claims 2-4 and 8” [remarks, page 21] but the examiner disagrees with this characterization of amended claim 1. Cancelled claims 2-4 recited limitations regarding applying groupings based on evaluation values, whereas limitations recited in amended claim 1 are about selecting architectures based on evaluation values. The latter is disclosed in Mavrovouniotis. The examiner maintains that claim 1 as amended is disclosed by Mavrovouniotis alone.
Regarding claim 14, the applicant asserts that the calculation of cross-correlation is not disclosed by Mavrovouniotis [remarks, page 21] while also admitting that “Tran discloses the use of correlation” [remarks, page 21]. This means that the “integrated pipeline in claim 14” [remarks, page 14] is in fact disclosed by Mavrovouniotis in view of Tran. The applicant's remarks do not comply with 37 CFR 1.111(c) because they do not clearly point out the patentable novelty which they think the claims present in view of the state of the art disclosed by the references cited. Further, they do not show how the amendments avoid such references.
Regarding claim 10, the applicant’s remarks [remarks, page 22] do not comply with 37 CFR 1.111(c) because they do not clearly point out the patentable novelty which they think the claims present in view of the state of the art disclosed by the references cited. Further, they do not show how the amendments avoid such references.
Regarding claim 16, the applicant asserts that “Lin discusses coefficient-based pruning but not its integration into group-based model generation and architecture selection” [remarks, page 22] but these features are disclosed by Mavrovouniotis in view of Lin. The applicant's remarks do not comply with 37 CFR 1.111(c) because they do not clearly point out the patentable novelty which they think the claims present in view of the state of the art disclosed by the references cited. Further, they do not show how the amendments avoid such references. The applicant also asserts that the claimed invention “includes dynamic candidate evaluation and selection mechanisms that are not trivially combinable” [remarks, page 22]. This remark does not appear relevant to the rejection as given since both evaluation and selection of model architecture candidates are disclosed by Mavrovouniotis alone; no combination of references is required for disclosing the combination of these two features.
The complete prior art mapping, updated in view of the applicant’s amendments, is given below.
Examiner’s Remarks
Claims are grouped as follows in this Office action:
Claim Group A: Claims 1, 21-22, and Dependents (7, 9, 11-12)
Claim Group B: Claims 14, 23-24, and Dependents (10, 15)
Claim Group C: Claims 16, 25-26, and Dependents (17-20)
Claim Objections
Claim(s) 1, 7, 9, 11-12 is/are objected to because of the following informalities:
Claim Group A
Claims 1 and 21-22 recite the plurality of candidates but this should be “the plurality of model architecture candidates” for consistency of nomenclature and proper antecedent basis. Dependent claims 7, 9, and 11-12 inherit the objection.
Claim 12 recites according to claim 1, the prediction model but this is ungrammatical; it should be “according to claim 1, wherein the prediction model”.
Claim Group B
Claims 14 and 23-24 recite the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates [emphasis added] but this should be “the evaluation values values of the prediction models
Claim Group C
Claims 16 and 25-26 recite the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates [emphasis added] but this should be “the evaluation values values of the prediction models
Claim 19 recites when the value of the second variable corresponding to a peak portion [emphasis added] but this should be “when the value of the second variable corresponds to a peak portion” for grammaticality.
Appropriate correction is required.
Claim Rejections - 35 USC 112(d)
The following is a quotation of 35 USC 112(d):
(d) REFERENCE IN DEPENDENT FORMS.—Subject to subsection (e), a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
The following is a quotation of pre-AIA 35 USC 112, fourth paragraph:
Subject to the following paragraph [i.e., the fifth paragraph of pre-AIA 35 U.S.C. 112], a claim in dependent form shall contain a reference to a claim previously set forth and then specify a further limitation of the subject matter claimed. A claim in dependent form shall be construed to incorporate by reference all the limitations of the claim to which it refers.
Claim(s) 10 is/are rejected under 35 USC 112(d) or pre-AIA 35 USC 112, 4th paragraph, as being of improper dependent form for failing to further limit the subject matter of the claim upon which it depends, or for failing to include all the limitations of the claim upon which it depends.
Claim Group B
Dependent claim 10 includes a reference to independent claim 14, but this is not a “a reference to a claim previously set forth” [emphasis added] as required by 35 USC 112(d). MPEP 608.01(n)(III) indicates that, “[a]lthough the requirements of 35 USC 112(d) are related to matters of form, non-compliance with 35 USC 112(d) renders the claim unpatentable just as non-compliance with other paragraphs of 35 USC 112 would” and that “[c]laims which are in improper dependent form for failing to further limit the subject matter of a previous claim… should be rejected under 35 USC 112(d)”. Consequently, claim 10 is rejected for being an improper dependent of claim 14.
Applicant may cancel the claim(s), amend the claim(s) to place the claim(s) in proper dependent form, rewrite the claim(s) in independent form, or present a sufficient showing that the dependent claim(s) complies with the statutory requirements.
Claim Rejections - 35 USC 112(b)
The following is a quotation of 35 USC 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 USC 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim(s) 7, 9, 11-12, and 18 is/are rejected under 35 USC 112(b) or 35 USC 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 USC 112, the applicant), regards as the invention.
Claim Group A
Claim 7 recites and the automatic generation method is configured to generate the grouping candidates by dividing the arranged plurality of first variables at different positions, and generate a plurality of grouping candidates by repeatedly changing the division positions [emphasis added] but the underlined phrases lack antecedent basis. Regarding the first phrase, the examiner suggests simply deleting the phrase “and the automatic generation method is configured to”. The scope of the remainder of this limitation is not clear to the examiner even in view of the specification (cf. 112(a) rejections). The applicant is advised to use alternate language which clearly lays out the scope of the claim and which is consistent with language used in the specification.
Claims 9 and 11-12 recite the prediction model but this has ambiguous antecedent basis since the parent claim recites a plurality of prediction models, one for each of the plurality of model architecture candidates. For the purpose of compact prosecution, the claims are interpreted broadly as encompassing any one of these prediction models.
Claim Group C
Claim 18 recites based on a generic algorithm but this phrase is syntactically ambiguous: it is unclear whether it is the generation of first variables which is “based on a genetic algorithm” or the combining of explanatory variables which is “based on a genetic algorithm” or something else. For the purpose of compact prosecution, the claim is interpreted broadly as encompassing any of these interpretations.
Claim Rejections - 35 USC 112(a)
The following is a quotation of the first paragraph of 35 USC 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 USC 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claim(s) 7 is/are rejected under 35 USC 112(a) or 35 USC 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 USC 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Claim Group A
Claim 7 recites generate the grouping candidates by dividing the arranged plurality of first variables at different positions, and generate a plurality of grouping candidates by repeatedly changing the division positions but this feature is new matter as it is not described in the originally filed specification. The specification does not, for example, describe a process of “repeatedly changing the division positions” in chronologically arranged data. MPEP 2161.01 indicates that a computer-implemented functional claim limitation may lack adequate written description when the claims define the invention in functional language specifying a desired result but the specification does not sufficiently describe how the function is performed or the result is achieved. In other words, the algorithm or steps/procedure taken to perform the function must be described with sufficient detail so that one of ordinary skill in the art would understand how the inventor intended the function to be performed. Consequently, this limitation is new matter and is rejected for inadequate written description.
Claim Rejections - 35 USC 101
35 USC 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.
Claim(s) 1, 7, 9-12, and 14-26 is/are rejected under 35 USC 101 because the claimed invention(s) is/are directed to abstract ideas without significantly more.
Claim Group A
Step 1. Claims 1, 7, 9, 11-12, and 22 fall under the statutory category of machines. Claim 21 falls under the statutory category of methods. An analysis of step 2 for each of these claims follows.
Claim 1
Step 2A Prong 1. The claim recites the following abstract ideas:
identify, among the plurality of first variables, a variable corresponding to a first time, a variable corresponding to a second time before the first time, and a variable corresponding to a third time after the first time; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can identify variables corresponding to various times. See MPEP 2106.04(a)(2)(III).)
classify the variable corresponding to the first time into a first group, classify the variable corresponding to the second time into a second group, and classify the variable corresponding to the third time into a third group, thereby grouping the plurality of first variables into a plurality of groups and generating the plurality of groups including the first variables; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can classify variables into groups. See MPEP 2106.04(a)(2)(III).)
generate, for each of a plurality of model architecture candidates, a prediction model configured to associate the first variables included in the first, second, and third groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculate an evaluation value of each of the prediction models based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determine, based on the evaluation values, a model architecture to be used from among the plurality of candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select an architecture to be used. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
An information processing apparatus, comprising: processing circuitry configured to (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
obtain first data including a plurality of first variables and a second variable; (This recites insignificant extra-solution activity. See MPEP 2106.05(g).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
An information processing apparatus, comprising: processing circuitry configured to (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
obtain first data including a plurality of first variables and a second variable; (This insignificant extra-solution activity is well-understood, routine, conventional as it is mere data transfer. See MPEP 2106.05(d)(II), “Receiving or transmitting data over a network” and/or “Storing and retrieving information in memory”.)
Claim 7
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
[The information processing apparatus according to claim 1, wherein the processing circuitry is configured to] arrange the plurality of first variables in chronological order based on the times corresponding to the plurality of first variables, (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
[and the automatic generation method is configured to] generate the grouping candidates by dividing the arranged plurality of first variables at different positions, and generate a plurality of grouping candidates by repeatedly changing the division positions. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
Claim 9
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 1, wherein] the first time corresponds to a time at when prediction is performed by the prediction model. (This recites data of a particular type or source, merely linking an abstract idea to a particular field of use. See MPEP 2106.05(h).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 1, wherein] the first time corresponds to a time at when prediction is performed by the prediction model. (This recites data of a particular type or source, merely linking an abstract idea to a particular field of use. See MPEP 2106.05(h).)
Claim 11
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 1, wherein] the prediction model is a neural network that includes an input layer, at least one intermediate layer, and an output layer, and the plurality of model architecture candidates differ in a number of nodes in the at least one intermediate layer. (This recites generic structure of neural networks. In other words, this recites merely applying (or equivalent) an abstract idea, or implementing an abstract idea on a computer, or using a computer as a tool to perform an abstract idea. See MPEP 2106.05(f).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 1, wherein] the prediction model is a neural network that includes an input layer, at least one intermediate layer, and an output layer, and the plurality of model architecture candidates differ in a number of nodes in the at least one intermediate layer. (This recites generic structure of neural networks. In other words, this recites merely applying (or equivalent) an abstract idea, or implementing an abstract idea on a computer, or using a computer as a tool to perform an abstract idea. See MPEP 2106.05(f).)
Claim 12
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 1,] the prediction model is a neural network that includes an input layer, at least one intermediate layer, and an output layer, and the plurality of model architecture candidates differ in a number of nodes in the at least one intermediate layer. (This recites generic structure of neural networks. In other words, this recites merely applying (or equivalent) an abstract idea, or implementing an abstract idea on a computer, or using a computer as a tool to perform an abstract idea. See MPEP 2106.05(f).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 1,] the prediction model is a neural network that includes an input layer, at least one intermediate layer, and an output layer, and the plurality of model architecture candidates differ in a number of nodes in the at least one intermediate layer. (This recites generic structure of neural networks. In other words, this recites merely applying (or equivalent) an abstract idea, or implementing an abstract idea on a computer, or using a computer as a tool to perform an abstract idea. See MPEP 2106.05(f).)
Claim 21
Step 2A Prong 1. The claim recites the following abstract ideas:
identifying, among the plurality of first variables, a variable corresponding to a first time, a variable corresponding to a second time before the first time, and a variable corresponding to a third time after the first time; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can identify variables corresponding to various times. See MPEP 2106.04(a)(2)(III).)
classifying the variable corresponding to the first time into a first group, classify the variable corresponding to the second time into a second group, and classify the variable corresponding to the third time into a third group, thereby grouping the plurality of first variables into a plurality of groups and generating the plurality of groups including the first variables; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can classify variables into groups. See MPEP 2106.04(a)(2)(III).)
generating, for each of a plurality of model architecture candidates, a prediction model configured to associate the first variables included in the first, second, and third groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculating an evaluation value of each of the prediction models based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determining, based on the evaluation values, a model architecture to be used from among the plurality of candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select an architecture to be used. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
An information processing method, comprising: obtaining first data including a plurality of first variables and a second variable; (This recites insignificant extra-solution activity. See MPEP 2106.05(g).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
An information processing method, comprising: obtaining first data including a plurality of first variables and a second variable; (This insignificant extra-solution activity is well-understood, routine, conventional as it is mere data transfer. See MPEP 2106.05(d)(II), “Receiving or transmitting data over a network” and/or “Storing and retrieving information in memory”.)
Claim 22
Step 2A Prong 1. The claim recites the following abstract ideas:
identifying, among the plurality of first variables, a variable corresponding to a first time, a variable corresponding to a second time before the first time, and a variable corresponding to a third time after the first time; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can identify variables corresponding to various times. See MPEP 2106.04(a)(2)(III).)
classifying the variable corresponding to the first time into a first group, classify the variable corresponding to the second time into a second group, and classify the variable corresponding to the third time into a third group, thereby grouping the plurality of first variables into a plurality of groups and generating the plurality of groups including the first variables; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can classify variables into groups. See MPEP 2106.04(a)(2)(III).)
generating, for each of a plurality of model architecture candidates, a prediction model configured to associate the first variables included in the first, second, and third groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculating an evaluation value of each of the prediction models based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determining, based on the evaluation values, a model architecture to be used from among the plurality of candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select an architecture to be used. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
obtaining first data including a plurality of first variables and a second variable; (This recites insignificant extra-solution activity. See MPEP 2106.05(g).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
obtaining first data including a plurality of first variables and a second variable; (This insignificant extra-solution activity is well-understood, routine, conventional as it is mere data transfer. See MPEP 2106.05(d)(II), “Receiving or transmitting data over a network” and/or “Storing and retrieving information in memory”.)
Claim Group B
Step 1. Claims 14, 10, 15, and 24 fall under the statutory category of machines. Claim 23 falls under the statutory category of methods. An analysis of step 2 for each of these claims follows.
Claim 14
Step 2A Prong 1. The claim recites the following abstract ideas:
calculate a cross-correlation between a plurality of explanatory variables and an objective variable, based on time-series data of the explanatory variables and time-series data of the objective variable; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. Cross-correlation is a mathematical concept, and its calculation can be performed by a human mind. See MPEP 2106.04(a)(2)(I, III).)
create first data including a plurality of first variables corresponding to a plurality of times before a prediction target time, selected from the plurality of explanatory variables based on the cross-correlation, and a second variable that includes the objective variable corresponding to the prediction target time; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
generate one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can generate grouping candidates. See MPEP 2106.04(a)(2)(III).)
generate, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculate an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determine a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select a grouping and/or architecture to be used based on their performance. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
An information processing apparatus, comprising: processing circuitry configured to: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
An information processing apparatus, comprising: processing circuitry configured to: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Claim 10
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
[The information processing apparatus according to claim 14, wherein the processing circuitry is configured to] generate the plurality of grouping candidates by randomly assigning the plurality of first variables to a plurality of groups. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
Claim 15
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
[The information processing apparatus according to claim 14, wherein the processing circuitry is configured to] calculate an autocorrelation of the objective variable, (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. Autocorrelation is a mathematical concept, and its calculation can be performed by a human mind. See MPEP 2106.04(a)(2)(I, III).)
and to include, as one of the first variables, the objective variable corresponding to a time before the prediction target time, based on the autocorrelation. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can include variables in a data set. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
Claim 23
Step 2A Prong 1. The claim recites the following abstract ideas:
An information processing method, comprising: calculating a cross-correlation between a plurality of explanatory variables and an objective variable, based on time-series data of the explanatory variables and time-series data of the objective variable; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. Cross-correlation is a mathematical concept, and its calculation can be performed by a human mind. See MPEP 2106.04(a)(2)(I, III).)
creating first data including a plurality of first variables corresponding to a plurality of times before a prediction target time, selected from the plurality of explanatory variables based on the cross-correlation, and a second variable that includes the objective variable corresponding to the prediction target time; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
generating one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can generate grouping candidates. See MPEP 2106.04(a)(2)(III).)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select a grouping and/or architecture to be used based on their performance. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
None.
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
None.
Claim 24
Step 2A Prong 1. The claim recites the following abstract ideas:
calculating a cross-correlation between a plurality of explanatory variables and an objective variable, based on time-series data of the explanatory variables and time-series data of the objective variable; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. Cross-correlation is a mathematical concept, and its calculation can be performed by a human mind. See MPEP 2106.04(a)(2)(I, III).)
creating first data including a plurality of first variables corresponding to a plurality of times before a prediction target time, selected from the plurality of explanatory variables based on the cross-correlation, and a second variable that includes the objective variable corresponding to the prediction target time; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
generating one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can generate grouping candidates. See MPEP 2106.04(a)(2)(III).)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select a grouping and/or architecture to be used based on their performance. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Claim Group C
Step 1. Claims 16-20 and 26 fall under the statutory category of machines. Claim 25 falls under the statutory category of methods. An analysis of step 2 for each of these claims follows.
Claim 16
Step 2A Prong 1. The claim recites the following abstract ideas:
perform regression of an objective variable at a prediction target time using one or more explanatory variables corresponding to a plurality of times before the prediction target time, based on time-series data of the explanatory variables and the objective variable; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can perform time series regressions. See, for example, [specification, figure 8] for support that this recites a mathematical concept. See MPEP 2106.04(a)(2)(I, III).)
calculate coefficients corresponding to each time point of the explanatory variables; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate coefficients related to time-series regression models. See MPEP 2106.04(a)(2)(I, III).)
select, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can select variables based on calculated coefficients. See MPEP 2106.04(a)(2)(III).)
create first data including the selected explanatory variables as a plurality of first variables and the objective variable at the prediction target time as a second variable; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
generate one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can generate grouping candidates. See MPEP 2106.04(a)(2)(III).)
generate, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculate an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determine a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select a grouping and/or architecture to be used based on their performance. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
An information processing apparatus, comprising: processing circuitry configured to: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
An information processing apparatus, comprising: processing circuitry configured to: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Claim 17
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
[The information processing apparatus according to claim 16, wherein the processing circuitry is configured to] generate the first variables by combining the explanatory variables corresponding to the plurality of times and at least one operator, (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
based on a genetic algorithm. (This recites a mathematical concept. See MPEP 2106.04(a)(2)(I).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
Claim 18
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
[The information processing apparatus according to 16, wherein the processing circuitry is configured to] assign a weight to the first data, based on a value of the second variable in the first data, (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can assign weights to data. See MPEP 2106.04(a)(2)(III).)
and generate the prediction model, based on the weight. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can generate prediction models based on weights. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
Claim 19
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
[The information processing apparatus according to claim 18, wherein the processing circuitry is configured to] assign a first weight to the first data when the value of the second variable corresponding to a peak portion, (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can assign weights to data. See MPEP 2106.04(a)(2)(III).)
and assign a second weight smaller than the first weight when the value of the second variable corresponds to a non-peak portion. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can assign weights to data. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
Claim 20
Step 2A Prong 1. The claim recites the following abstract ideas:
The abstract idea(s) in the parent claim(s).
[wherein the processing circuitry is configured to] determine which method to use based on the user selection. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine which method to use based on instructions. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 16, further comprising] a graphical user interface circuit configured to allow a user to select whether to use the predefined automatic generation method or the grouping method specified by the user, (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
The additional element(s) in the parent claim(s).
[The information processing apparatus according to claim 16, further comprising] a graphical user interface circuit configured to allow a user to select whether to use the predefined automatic generation method or the grouping method specified by the user, (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Claim 25
Step 2A Prong 1. The claim recites the following abstract ideas:
An information processing method, comprising: performing regression of an objective variable at a prediction target time using one or more explanatory variables corresponding to a plurality of times before the prediction target time, based on time-series data of the explanatory variables and the objective variable; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can perform time series regressions. See, for example, [specification, figure 8] for support that this recites a mathematical concept. See MPEP 2106.04(a)(2)(I, III).)
calculating coefficients corresponding to each time point of the explanatory variables; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate coefficients related to time-series regression models. See MPEP 2106.04(a)(2)(I, III).)
selecting, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can select variables based on calculated coefficients. See MPEP 2106.04(a)(2)(III).)
creating first data including the selected explanatory variables as a plurality of first variables and the objective variable at the prediction target time as a second variable; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
generate one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can generate grouping candidates. See MPEP 2106.04(a)(2)(III).)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select a grouping and/or architecture to be used based on their performance. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
None.
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
None.
Claim 26
Step 2A Prong 1. The claim recites the following abstract ideas:
performing regression of an objective variable at a prediction target time using one or more explanatory variables corresponding to a plurality of times before the prediction target time, based on time-series data of the explanatory variables and the objective variable; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can perform time series regressions. See, for example, [specification, figure 8] for support that this recites a mathematical concept. See MPEP 2106.04(a)(2)(I, III).)
calculating coefficients corresponding to each time point of the explanatory variables; (This recites a mathematical concept and/or a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate coefficients related to time-series regression models. See MPEP 2106.04(a)(2)(I, III).)
selecting, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can select variables based on calculated coefficients. See MPEP 2106.04(a)(2)(III).)
creating first data including the selected explanatory variables as a plurality of first variables and the objective variable at the prediction target time as a second variable; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. See MPEP 2106.04(a)(2)(III).)
generate one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can generate grouping candidates. See MPEP 2106.04(a)(2)(III).)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. The examiner notes that the scope of the “prediction model” of the claim encompasses models such as linear regression [specification, 0034], which are mathematical concepts that are feasible to train/generate by a human mind. See MPEP 2106.04(a)(2)(I, III).)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; (This recites a mathematical concept and a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can calculate values based on differences between a prediction and an actual value. See MPEP 2106.04(a)(2)(I, III).)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. (This recites a mental process that can be performed in the human mind or by a human using pen and paper. A human mind can determine/select a grouping and/or architecture to be used based on their performance. See MPEP 2106.04(a)(2)(III).)
Step 2A Prong 2. The claim recites the following additional elements which, considered individually and as an ordered combination, do not integrate the abstract idea into a practical application:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Step 2B. The claim recites the following additional elements which, considered individually and as an ordered combination, do not amount to significantly more than the abstract idea:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: (This recites generic computing components for performing an abstract idea. See MPEP 2106.05(f)(2).)
Claim Rejections - 35 USC 102
The following is a quotation of the appropriate paragraphs of 35 USC 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 USC 102(b)(2)(C) for any potential 35 USC 102(a)(2) prior art against the later invention.
Claim Group A
Claim(s) 1, 7, 9, 11-12, and 21-22 is/are rejected under 35 USC 102(a)(1) as being anticipated by M. L. MAVROVOUNIOTIS et al. (Hierarchical Neural Networks, published 1992; hereafter “Mavrovouniotis”).
Claim 1
Mavrovouniotis discloses:
An information processing apparatus, comprising: processing circuitry configured to ([Mavrovouniotis, section 3]: Mavrovouniotis discloses using “Macintosh II computers” to implement the methods disclosed therein [Mavrovouniotis, section 3 paragraph beginning “The software”]. Either the computer itself or a processor contained therein can fall under the broadest reasonable interpretation of an “information processing apparatus” and of “processing circuitry” as recited by the claim.)
obtain first data including a plurality of first variables and a second variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having input variables V_i(k) for varying V, i, and k and an output variable [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables map to the “plurality of first variables” of the claim, the output variable to the “second variable” of the claim, and the dataset maps to the “first data” of the claim.)
identify, among the plurality of first variables, a variable corresponding to a first time, a variable corresponding to a second time before the first time, and a variable corresponding to a third time after the first time; ([Mavrovouniotis, section 2.2]: The value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus”]. This means, for example, that any one of the V_i(1) maps to the “variable corresponding to a first time” of the claim, any one of the V_i(2) maps to the “variable corresponding to a second time before the first time” of the claim, and any one of the V_i(0) maps to the ”variable corresponding to a third time after the first time” of the claim.)
classify the variable corresponding to the first time into a first group, classify the variable corresponding to the second time into a second group, and classify the variable corresponding to the third time into a third group, thereby grouping the plurality of first variables into a plurality of groups and generating the plurality of groups including the first variables; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph]. Of these, at least the second and third strategies (namely, the one in which “[f]or any given process stream and any fixed time point, the measurements for all of the variables of the stream are related” and the one in which “[f]or any specific kind of variable… and a fixed time point, the measurements of that kind of variable for all of the streams form a cluster” [Mavrovouniotis, section 2.3 first paragraph; emphasis added]) fall under the broadest reasonable interpretation of the claim. Under either of these strategies, any one of the clusters corresponding to the fixed time point t - Δt would be the “first group” of the claim, any one of the clusters corresponding to the fixed time point t - 2Δt would be the “second group” of the claim, and any one of the clusters corresponding to the fixed time point t would be the “third group” of the claim.)
generate, for each of a plurality of model architecture candidates, a prediction model configured to associate the first variables included in the first, second, and third groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the first, second, and third groups” as recited in the claim) with predicted values of the output variable. These architectures map to the “plurality of model architecture candidates” of the claim, and for each of these architectures, the trained neural network having that architecture maps to the “prediction model” of the claim.)
calculate an evaluation value of each of the prediction models based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determine, based on the evaluation values, a model architecture to be used from among the plurality of candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1], and of these maps to the “model architecture to be used” of the claim.)
Claim 7
Mavrovouniotis discloses the elements of the parent claim(s). It also discloses:
[The information processing apparatus according to claim 1, wherein the processing circuitry is configured to] arrange the plurality of first variables in chronological order based on the times corresponding to the plurality of first variables, ([Mavrovouniotis, section 2.2]: Since V_i(k) is the value of V_i at time t - kΔt [Mavrovouniotis, section 2.2 paragraph beginning “To focus”], the variables V_i(0), V_i(1), …, V_i(n) are arranged “in chronological order based on the times corresponding to the plurality of first variables” as recited by the claim. See also: [Mavrovouniotis, section 2.4 paragraph beginning “Consider, for example, the subnet”].)
and the automatic generation method is configured to generate the grouping candidates by dividing the arranged plurality of first variables at different positions, and generate a plurality of grouping candidates by repeatedly changing the division positions. ([Mavrovouniotis, section 2.2]: The time points t - kΔt are part of a “moving window” [Mavrovouniotis, section 2.2 paragraph beginning “To focus”]. In other words, the time points t - kΔt map to the “different positions” and/or the “division positions” of the claim, and the fact that they are part of a moving window means that these time points “repeatedly chang[e]” as recited by the claim (as best understood by the examiner in view of the 112(b) rejections).)
Claim 9
Mavrovouniotis discloses the elements of the parent claim(s). It also discloses:
[The information processing apparatus according to claim 1, wherein] the first time corresponds to a time at when prediction is performed by the prediction model. ([Mavrovouniotis, section 2]: Any of the time-points t - kΔt fall under the broadest reasonable interpretation of “correspond[ing]” to a time at when prediction though a prediction model is performed (since, for example, all of the time points are part of the input data when the prediction model makes predictions).)
Claim 11
Mavrovouniotis discloses the elements of the parent claim(s). It also discloses:
[The information processing apparatus according to claim 1, wherein] the prediction model is a neural network that includes an input layer, at least one intermediate layer, and an output layer, ([Mavrovouniotis, figures 8-14]: All of the architectures in Mavrovouniotis are neural network models having an input layer, at least one intermediate layer, and an output layer.)
and the plurality of model architecture candidates differ in a number of nodes in the at least one intermediate layer. ([Mavrovouniotis, figures 8-14]: The architectures in Mavrovouniotis have differing numbers of nodes in the intermediate layer(s). For example, 8-TNN [Mavrovouniotis, figure 8] has 8 nodes in the intermediate layer, while 3-TNN [Mavrovouniotis, figure 9] has 3 nodes in the intermediate layer, while 4-HNN-p [Mavrovouniotis, figure 10] has 4 nodes in the intermediate layer, while 8-2-HNN-p [Mavrovouniotis, figure 11] has 10 nodes in the intermediate layers, and so forth.)
Claim 12
Mavrovouniotis discloses the elements of the parent claim(s). It also discloses:
[The information processing apparatus according to claim 1,] the prediction model is a neural network including an input layer, at least one intermediate layer, and an output layer, ([Mavrovouniotis, figures 8-14]: All of the architectures in Mavrovouniotis are neural network models having an input layer, at least one intermediate layer, and an output layer.)
and the plurality of model architecture candidates differ in a number of layers of the at least one intermediate layer. ([Mavrovouniotis, figures 8-14]: The architectures in Mavrovouniotis have differing numbers of nodes in the intermediate layer(s). For example, 8-TNN [Mavrovouniotis, figure 8] has 8 nodes in the intermediate layer, while 3-TNN [Mavrovouniotis, figure 9] has 3 nodes in the intermediate layer, while 4-HNN-p [Mavrovouniotis, figure 10] has 4 nodes in the intermediate layer, while 8-2-HNN-p [Mavrovouniotis, figure 11] has 10 nodes in the intermediate layers, and so forth.)
Claim 21
Mavrovouniotis discloses:
An information processing method, comprising: obtaining first data including a plurality of first variables and a second variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having input variables V_i(k) for varying V, i, and k and an output variable [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables map to the “plurality of first variables” of the claim, the output variable to the “second variable” of the claim, and the dataset maps to the “first data” of the claim.)
identifying, among the plurality of first variables, a variable corresponding to a first time, a variable corresponding to a second time before the first time, and a variable corresponding to a third time after the first time; ([Mavrovouniotis, section 2.2]: The value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus”]. This means, for example, that any one of the V_i(1) maps to the “variable corresponding to a first time” of the claim, any one of the V_i(2) maps to the “variable corresponding to a second time before the first time” of the claim, and any one of the V_i(0) maps to the ”variable corresponding to a third time after the first time” of the claim.)
classifying the variable corresponding to the first time into a first group, classify the variable corresponding to the second time into a second group, and classify the variable corresponding to the third time into a third group, thereby grouping the plurality of first variables into a plurality of groups and generating the plurality of groups including the first variables; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph]. Of these, at least the second and third strategies (namely, the one in which “[f]or any given process stream and any fixed time point, the measurements for all of the variables of the stream are related” and the one in which “[f]or any specific kind of variable… and a fixed time point, the measurements of that kind of variable for all of the streams form a cluster” [Mavrovouniotis, section 2.3 first paragraph; emphasis added]) fall under the broadest reasonable interpretation of the claim. Under either of these strategies, any one of the clusters corresponding to the fixed time point t - Δt would be the “first group” of the claim, any one of the clusters corresponding to the fixed time point t - 2Δt would be the “second group” of the claim, and any one of the clusters corresponding to the fixed time point t would be the “third group” of the claim.)
generating, for each of a plurality of model architecture candidates, a prediction model configured to associate the first variables included in the first, second, and third groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the first, second, and third groups” as recited in the claim) with predicted values of the output variable. These architectures map to the “plurality of model architecture candidates” of the claim, and for each of these architectures, the trained neural network having that architecture maps to the “prediction model” of the claim.)
calculating an evaluation value of each of the prediction models based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determining, based on the evaluation values, a model architecture to be used from among the plurality of candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1], and of these maps to the “model architecture to be used” of the claim.)
Claim 22
Mavrovouniotis discloses:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: ([Mavrovouniotis, section 3]: Mavrovouniotis discloses using “Macintosh II computers” to implement the methods disclosed therein [Mavrovouniotis, section 3 paragraph beginning “The software”]. The Macintosh II computer is the “computer” of the claim and its hard drive is the “non-transitory computer readable medium having a computer program stored therein” of the claim.)
obtaining first data including a plurality of first variables and a second variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having input variables V_i(k) for varying V, i, and k and an output variable [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables map to the “plurality of first variables” of the claim, the output variable to the “second variable” of the claim, and the dataset maps to the “first data” of the claim.)
identifying, among the plurality of first variables, a variable corresponding to a first time, a variable corresponding to a second time before the first time, and a variable corresponding to a third time after the first time; ([Mavrovouniotis, section 2.2]: The value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus”]. This means, for example, that any one of the V_i(1) maps to the “variable corresponding to a first time” of the claim, any one of the V_i(2) maps to the “variable corresponding to a second time before the first time” of the claim, and any one of the V_i(0) maps to the ”variable corresponding to a third time after the first time” of the claim.)
classifying the variable corresponding to the first time into a first group, classify the variable corresponding to the second time into a second group, and classify the variable corresponding to the third time into a third group, thereby grouping the plurality of first variables into a plurality of groups and generating the plurality of groups including the first variables; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph]. Of these, at least the second and third strategies (namely, the one in which “[f]or any given process stream and any fixed time point, the measurements for all of the variables of the stream are related” and the one in which “[f]or any specific kind of variable… and a fixed time point, the measurements of that kind of variable for all of the streams form a cluster” [Mavrovouniotis, section 2.3 first paragraph; emphasis added]) fall under the broadest reasonable interpretation of the claim. Under either of these strategies, any one of the clusters corresponding to the fixed time point t - Δt would be the “first group” of the claim, any one of the clusters corresponding to the fixed time point t - 2Δt would be the “second group” of the claim, and any one of the clusters corresponding to the fixed time point t would be the “third group” of the claim.)
generating, for each of a plurality of model architecture candidates, a prediction model configured to associate the first variables included in the first, second, and third groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the first, second, and third groups” as recited in the claim) with predicted values of the output variable. These architectures map to the “plurality of model architecture candidates” of the claim, and for each of these architectures, the trained neural network having that architecture maps to the “prediction model” of the claim.)
calculating an evaluation value of each of the prediction models based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determining, based on the evaluation values, a model architecture to be used from among the plurality of candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1], and of these maps to the “model architecture to be used” of the claim.)
Claim Rejections - 35 USC 103
The following is a quotation of 35 USC 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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 USC 102(b)(2)(C) for any potential 35 USC 102(a)(2) prior art against the later invention.
Claim Group B
Claim(s) 14-15 and 23-24 is/are rejected under 35 USC 103 as being unpatentable over Mavrovouniotis in view of Trung Duc TRAN et al. (Improving the Accuracy of Dam Inflow Predictions Using a Long Short-Term Memory Network Coupled with Wavelet Transform and Predictor Selection, published 2021-03-05; hereafter “Tran”).
Claim 14
Mavrovouniotis discloses:
An information processing apparatus, comprising: processing circuitry configured to: ([Mavrovouniotis, section 3]: Mavrovouniotis discloses using “Macintosh II computers” to implement the methods disclosed therein [Mavrovouniotis, section 3 paragraph beginning “The software”]. Either the computer itself or a processor contained therein can fall under the broadest reasonable interpretation of an “information processing apparatus” and of “processing circuitry” as recited by the claim.)
a plurality of explanatory variables and an objective variable, [based on] time-series data of the explanatory variables and time-series data of the objective variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having an output variable and input variables V_i(k) for varying V, i, and k, where the value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Moreover, while Mavrovouniotis focuses attention on networks whose output is “1 if there is a fault, and 0 if there is no fault”, it also clearly indicates that “the hierarchical approach we will discuss… is not specific to this task and could also be used, for example, within networks that carry out… prediction of future values of variables” [Mavrovouniotis, section 2.2 paragraph beginning “To focus our attention”]. In other words, Mavrovouniotis does distinctly disclose the use of hierarchical neural networks for time-series regressions. The input variables map to the “explanatory variables” and the “time-series data of the explanatory variables” of the claim, and the output variable to the “objective variable” and the “time-series data of the objective variable” of the claim.)
create first data including a plurality of first variables corresponding to a plurality of times before a prediction target time, selected from the plurality of explanatory variables [based on the cross-correlation,] and a second variable that includes the objective variable corresponding to the prediction target time; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables and the and output variable map, respectively, to the “first variables” and “second variable” of the claim. The data set used for training and testing maps to the “first data” of the claim (since it includes all of the input variables, it a fortiori includes the “plurality of first variables… selected from the plurality of explanatory variables” as recited by the claim). The examiner notes that any of the time-points t - kΔt fall under the broadest reasonable interpretation of “corresponding to a plurality of times before a prediction target time” (since, for example, all of the time points are times before the prediction is performed) as well as under the broadest reasonable interpretation of “corresponding to the prediction target time” (since, for example, all of the time points are part of the input data when the prediction model makes predictions).)
generate one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph], any of which can map to the “one or more grouping candidates” of the claim.)
generate, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the respective groups” as recited in the claim) with predicted values of the output variable. All of the architectures incorporate a way of grouping the input variables. In other words, the architectures map to both the “one or more model architecture candidates” and the “combinations of the grouping candidates and one or more model architecture candidates” of the claim, and the trained neural networks having those architectures map to the “prediction models” of the claim.)
calculate an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determine a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1]. Any of these three architectures maps to the “model architecture” of the claim, and the grouping used to generate that architecture maps to the “grouping” of the claim.)
While Mavrovouniotis discloses the use of hierarchical neural networks for time-series regressions, it might not distinctly disclose:
calculate a cross-correlation between [a plurality of explanatory variables and an objective variable,] based on [time-series data of the explanatory variables and time-series data of the objective variable] … [a plurality of first variables… selected from the plurality of explanatory variables] based on the cross-correlation,
Tran is in the field of machine learning and discloses the use of a recurrent neural network for time series forecasting [Tran, abstract and figure 1]. Moreover, Mavrovouniotis in view of Tran discloses:
calculate a cross-correlation between [a plurality of explanatory variables and an objective variable,] based on [time-series data of the explanatory variables and time-series data of the objective variable] … [a plurality of first variables… selected from the plurality of explanatory variables] based on the cross-correlation, ([Tran, abstract and section 2.2]: Tran discloses “using a ‘correlation threshold’ for partial autocorrelation and cross-correlation functions… and only variables greater than this threshold are selected as input predictors and their time lags” [Tran, abstract]. In the combination, the input predictor selection method of Tran is used to determine the input variables for the neural network. See [Tran, section 2.2] for more details about this method. In particular, it includes a consideration of a cross-correlation function (CCF) which “measures the similarity of a time series (e.g., dam inflow, y) with… candidate input variables, v = [v_1, …, v_{N_v}]” [Tran, section 2.2]. The output series y corresponds to the “objective variable” of the claim as mapped above, and the candidate input variables v = [v_1, …, v_{N_v}] correspond to the “explanatory variables” as mapped above. In other words, the CCF maps to the “cross-correlation” of the claim, and then the selection of the first variables in the input variables (i.e., the “explanatory variables” of the claim) is “based on the cross-correlation” as recited by the claim.)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural network of Mavrovouniotis with the input predictor selection method of Tran because “[m]aintaining a high correlation between inputs and outputs can guarantee the predictability of data-driven models” [Tran, section 2.2 first paragraph], so the combination would be more effective overall.
Claim 15
Mavovouniotis in view of Tran discloses the elements of the parent claim(s). It also discloses:
[The information processing apparatus according to claim 14, wherein the processing circuitry is configured to] calculate an autocorrelation of the objective variable, and to include, as one of the first variables, the objective variable corresponding to a time before the prediction target time, based on the autocorrelation. ([Tran, abstract and section 2.2]: As noted above, Tran discloses “using a ‘correlation threshold’ for partial autocorrelation and cross-correlation functions… and only variables greater than this threshold are selected as input predictors and their time lags” [Tran, abstract]. In the combination, the input predictor selection method of Tran is used to determine the input variables for the neural network. See [Tran, section 2.2] for more details about this method. In particular, it includes a consideration of a partial autocorrelation function (PACF) which “measures the linear correlation between a time series (y_t) and a lagged version of itself (y_{t+k})” [Tran, section 2.2]. In other words, the PACF maps to the “autocorrelation” of the claim. The lagged version of the output variable maps to the “objective variable corresponding to a time before the prediction target time” of the claim.)
The same motivation to combine applies.
Claim 23
Mavrovouniotis discloses:
An information processing method, comprising: … a plurality of explanatory variables and an objective variable, [based on] time-series data of the explanatory variables and time-series data of the objective variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having an output variable and input variables V_i(k) for varying V, i, and k, where the value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Moreover, while Mavrovouniotis focuses attention on networks whose output is “1 if there is a fault, and 0 if there is no fault”, it also clearly indicates that “the hierarchical approach we will discuss… is not specific to this task and could also be used, for example, within networks that carry out… prediction of future values of variables” [Mavrovouniotis, section 2.2 paragraph beginning “To focus our attention”]. In other words, Mavrovouniotis does distinctly disclose the use of hierarchical neural networks for time-series regressions. The input variables map to the “explanatory variables” and the “time-series data of the explanatory variables” of the claim, and the output variable to the “objective variable” and the “time-series data of the objective variable” of the claim.)
creating first data including a plurality of first variables corresponding to a plurality of times before a prediction target time, selected from the plurality of explanatory variables [based on the cross-correlation,] and a second variable that includes the objective variable corresponding to the prediction target time; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables and the and output variable map, respectively, to the “first variables” and “second variable” of the claim. The data set used for training and testing maps to the “first data” of the claim (since it includes all of the input variables, it a fortiori includes the “plurality of first variables… selected from the plurality of explanatory variables” as recited by the claim). The examiner notes that any of the time-points t - kΔt fall under the broadest reasonable interpretation of “corresponding to a plurality of times before a prediction target time” (since, for example, all of the time points are times before the prediction is performed) as well as under the broadest reasonable interpretation of “corresponding to the prediction target time” (since, for example, all of the time points are part of the input data when the prediction model makes predictions).)
generating one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph], any of which can map to the “one or more grouping candidates” of the claim.)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the respective groups” as recited in the claim) with predicted values of the output variable. All of the architectures incorporate a way of grouping the input variables. In other words, the architectures map to both the “one or more model architecture candidates” and the “combinations of the grouping candidates and one or more model architecture candidates” of the claim, and the trained neural networks having those architectures map to the “prediction models” of the claim.)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1]. Any of these three architectures maps to the “model architecture” of the claim, and the grouping used to generate that architecture maps to the “grouping” of the claim.)
While Mavrovouniotis discloses the use of hierarchical neural networks for time-series regressions, it might not distinctly disclose:
calculating a cross-correlation between [a plurality of explanatory variables and an objective variable,] based on [time-series data of the explanatory variables and time-series data of the objective variable] … [a plurality of first variables… selected from the plurality of explanatory variables] based on the cross-correlation,
Tran is in the field of machine learning and discloses the use of a recurrent neural network for time series forecasting [Tran, abstract and figure 1]. Moreover, Mavrovouniotis in view of Tran discloses:
calculating a cross-correlation between [a plurality of explanatory variables and an objective variable,] based on [time-series data of the explanatory variables and time-series data of the objective variable] … [a plurality of first variables… selected from the plurality of explanatory variables] based on the cross-correlation, ([Tran, abstract and section 2.2]: Tran discloses “using a ‘correlation threshold’ for partial autocorrelation and cross-correlation functions… and only variables greater than this threshold are selected as input predictors and their time lags” [Tran, abstract]. In the combination, the input predictor selection method of Tran is used to determine the input variables for the neural network. See [Tran, section 2.2] for more details about this method. In particular, it includes a consideration of a cross-correlation function (CCF) which “measures the similarity of a time series (e.g., dam inflow, y) with… candidate input variables, v = [v_1, …, v_{N_v}]” [Tran, section 2.2]. The output series y corresponds to the “objective variable” of the claim as mapped above, and the candidate input variables v = [v_1, …, v_{N_v}] correspond to the “explanatory variables” as mapped above. In other words, the CCF maps to the “cross-correlation” of the claim, and then the selection of the first variables in the input variables (i.e., the “explanatory variables” of the claim) is “based on the cross-correlation” as recited by the claim.)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural network of Mavrovouniotis with the input predictor selection method of Tran because “[m]aintaining a high correlation between inputs and outputs can guarantee the predictability of data-driven models” [Tran, section 2.2 first paragraph], so the combination would be more effective overall.
Claim 24
Mavrovouniotis discloses:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: ([Mavrovouniotis, section 3]: Mavrovouniotis discloses using “Macintosh II computers” to implement the methods disclosed therein [Mavrovouniotis, section 3 paragraph beginning “The software”]. The Macintosh II computer is the “computer” of the claim and its hard drive is the “non-transitory computer readable medium having a computer program stored therein” of the claim.)
a plurality of explanatory variables and an objective variable, [based on] time-series data of the explanatory variables and time-series data of the objective variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having an output variable and input variables V_i(k) for varying V, i, and k, where the value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Moreover, while Mavrovouniotis focuses attention on networks whose output is “1 if there is a fault, and 0 if there is no fault”, it also clearly indicates that “the hierarchical approach we will discuss… is not specific to this task and could also be used, for example, within networks that carry out… prediction of future values of variables” [Mavrovouniotis, section 2.2 paragraph beginning “To focus our attention”]. In other words, Mavrovouniotis does distinctly disclose the use of hierarchical neural networks for time-series regressions. The input variables map to the “explanatory variables” and the “time-series data of the explanatory variables” of the claim, and the output variable to the “objective variable” and the “time-series data of the objective variable” of the claim.)
creating first data including a plurality of first variables corresponding to a plurality of times before a prediction target time, selected from the plurality of explanatory variables [based on the cross-correlation,] and a second variable that includes the objective variable corresponding to the prediction target time; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables and the and output variable map, respectively, to the “first variables” and “second variable” of the claim. The data set used for training and testing maps to the “first data” of the claim (since it includes all of the input variables, it a fortiori includes the “plurality of first variables… selected from the plurality of explanatory variables” as recited by the claim). The examiner notes that any of the time-points t - kΔt fall under the broadest reasonable interpretation of “corresponding to a plurality of times before a prediction target time” (since, for example, all of the time points are times before the prediction is performed) as well as under the broadest reasonable interpretation of “corresponding to the prediction target time” (since, for example, all of the time points are part of the input data when the prediction model makes predictions).)
generating one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph], any of which can map to the “one or more grouping candidates” of the claim.)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the respective groups” as recited in the claim) with predicted values of the output variable. All of the architectures incorporate a way of grouping the input variables. In other words, the architectures map to both the “one or more model architecture candidates” and the “combinations of the grouping candidates and one or more model architecture candidates” of the claim, and the trained neural networks having those architectures map to the “prediction models” of the claim.)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1]. Any of these three architectures maps to the “model architecture” of the claim, and the grouping used to generate that architecture maps to the “grouping” of the claim.)
While Mavrovouniotis discloses the use of hierarchical neural networks for time-series regressions, it might not distinctly disclose:
calculating a cross-correlation between [a plurality of explanatory variables and an objective variable,] based on [time-series data of the explanatory variables and time-series data of the objective variable] … [a plurality of first variables… selected from the plurality of explanatory variables] based on the cross-correlation,
Tran is in the field of machine learning and discloses the use of a recurrent neural network for time series forecasting [Tran, abstract and figure 1]. Moreover, Mavrovouniotis in view of Tran discloses:
calculating a cross-correlation between [a plurality of explanatory variables and an objective variable,] based on [time-series data of the explanatory variables and time-series data of the objective variable] … [a plurality of first variables… selected from the plurality of explanatory variables] based on the cross-correlation, ([Tran, abstract and section 2.2]: Tran discloses “using a ‘correlation threshold’ for partial autocorrelation and cross-correlation functions… and only variables greater than this threshold are selected as input predictors and their time lags” [Tran, abstract]. In the combination, the input predictor selection method of Tran is used to determine the input variables for the neural network. See [Tran, section 2.2] for more details about this method. In particular, it includes a consideration of a cross-correlation function (CCF) which “measures the similarity of a time series (e.g., dam inflow, y) with… candidate input variables, v = [v_1, …, v_{N_v}]” [Tran, section 2.2]. The output series y corresponds to the “objective variable” of the claim as mapped above, and the candidate input variables v = [v_1, …, v_{N_v}] correspond to the “explanatory variables” as mapped above. In other words, the CCF maps to the “cross-correlation” of the claim, and then the selection of the first variables in the input variables (i.e., the “explanatory variables” of the claim) is “based on the cross-correlation” as recited by the claim.)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural network of Mavrovouniotis with the input predictor selection method of Tran because “[m]aintaining a high correlation between inputs and outputs can guarantee the predictability of data-driven models” [Tran, section 2.2 first paragraph], so the combination would be more effective overall.
Claim(s) 10 is/are rejected under 35 USC 103 as being unpatentable over Mavrovouniotis in view of Tran, further in view of Mohamed ALY et al. (Novel Methods for the Feature Subset Ensembles Approach, published 2006; hereafter “Aly”).
Claim 10
Mavovouniotis in view of Tran discloses the elements of the parent claim(s). It might not distinctly disclose:
[The information processing apparatus according to claim 14, wherein the processing circuitry is configured to] generate the plurality of grouping candidates by randomly assigning the plurality of first variables to a plurality of groups.
Aly is in the field of machine learning and discusses “Feature Subset Ensemble (FSE)” techniques, which “partition[…] the input features among the individual prediction models in the ensemble”, giving an example where ten features, labelled 1-10, are grouped into four subsets {1, 2, 5, 7, 9}, {1, 3, 4, 6, 8}, {2, 3, 5, 6, 7}, and {1, 2, 3, 4, 10} [Aly, section 2]. In other words, the input features of Aly correspond to the input variables of Mavrouviontis and the “first variables” of the claim, and the groups correspond to the subsets of related inputs of Mavrouviontis and the “plurality of groups” of the claim. Moreover, Mavrouviontis in view of Tran and Aly discloses:
[The information processing apparatus according to claim 14, wherein the processing circuitry is configured to] generate the plurality of grouping candidates by randomly assigning the plurality of first variables to a plurality of groups. ([Aly, section 4.4]: Aly discloses several strategies for grouping variables, including “using randomly selected feature subset for each prediction model” [Aly, section 4.4].)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural networks of Mavrovouniotis in view of Tran with the methods of decomposing the input feature set via random groupings as described in Aly because it overcomes the problem that “features can get mingled together in a way that might worsen that prediction model’s performance” [Aly, section 4.4], so the combination would be more effective overall. (Further support for the obviousness of the combination is witnessed in the prior art reference Zheng cited in the conclusion of a previous office action.)
Claim Group C
Claim(s) 16, 18, and 25-26 is/are rejected under 35 USC 103 as being unpatentable over Mavrovouniotis in view of Tsung-Nan LIN et al. (A Delay Damage Model Selection Algorithm for NARX Neural Networks, published 1997; hereafter “Lin”).
Claim 16
Mavrovouniotis discloses:
An information processing apparatus, comprising: processing circuitry configured to: ([Mavrovouniotis, section 3]: Mavrovouniotis discloses using “Macintosh II computers” to implement the methods disclosed therein [Mavrovouniotis, section 3 paragraph beginning “The software”]. Either the computer itself or a processor contained therein can fall under the broadest reasonable interpretation of an “information processing apparatus” and of “processing circuitry” as recited by the claim.)
perform regression of an objective variable at a prediction target time using one or more explanatory variables corresponding to a plurality of times before the prediction target time, based on time-series data of the explanatory variables and the objective variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having an output variable and input variables V_i(k) for varying V, i, and k, where the value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Moreover, while Mavrovouniotis focuses attention on networks whose output is “1 if there is a fault, and 0 if there is no fault”, it also clearly indicates that “the hierarchical approach we will discuss… is not specific to this task and could also be used, for example, within networks that carry out… prediction of future values of variables” [Mavrovouniotis, section 2.2 paragraph beginning “To focus our attention”]. In other words, Mavrovouniotis does distinctly disclose the use of hierarchical neural networks for time-series regressions. The output variable of the time-series regression maps to the “objective variable at a prediction target time” of the claim, and the input variables of the time-series regression maps to the “one or more explanatory variables corresponding to a plurality of times before the prediction target time” of the claim. The values of the V_i(k) for varying V, i, and k are the “time-series data of the explanatory variables and the objective variable” of the claim.)
calculate coefficients corresponding to each time point of the explanatory variables; ([Mavrovouniotis, section 3.3]: Mavrovouniotis discloses calculating “weights (also called connection strengths)” between nodes in the neural network [Mavrouvouniotis, section 3.3 first paragraph]. Since there are input nodes for each input variable at each time point, the weights associated to connections coming out of the input nodes map to the “coefficients corresponding to each time point of the explanatory variables” of the claim.)
create first data including the selected explanatory variables as a plurality of first variables and the objective variable at the prediction target time as a second variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables and the and output variable map, respectively, to the “first variables” and “second variable” of the claim. The data set used for training and testing maps to the “first data” of the claim (since it includes all of the input variables, it a fortiori includes the “selected explanatory variables” as recited by the claim).)
generate one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph], any of which can map to the “one or more grouping candidates” of the claim.)
generate, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the respective groups” as recited in the claim) with predicted values of the output variable. All of the architectures incorporate a way of grouping the input variables. In other words, the architectures map to both the “one or more model architecture candidates” and the “combinations of the grouping candidates and one or more model architecture candidates” of the claim, and the trained neural networks having those architectures map to the “prediction models” of the claim.)
calculate an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determine a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1]. Any of these three architectures maps to the “model architecture” of the claim, and the grouping used to generate that architecture maps to the “grouping” of the claim.)
Mavrovouniotis might not distinctly disclose:
select, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times;
Lin is in the field of machine learning. It discloses the use of recurrent neural networks [Lin, figure 1] to model nonlinear autoregressive models with exogenous inputs (NARX models) [Lin, section I first paragraph or section II first paragraph]. The examiner notes that the applicant’s regression model (an example of which is depicted in figure 8 of the application) is a NARX model where the exogenous input variable u [Lin, section 2 equation (1)] is taken to be the vector (X_1, X_2, …) (and u can in fact be a vector, as witnessed more explicitly in a reference to which Lin cites, namely, Leontartis as cited in the conclusion of a previous Office action). Moreover, Mavrovouniotis in view of Lin discloses:
select, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times; ([Lin, sections 1-2]: Lin discloses “a pruning-based algorithm (the delay damage algorithm) to determine the optimal memory-order of NARX… neural networks” [Lin, section 1 paragraph beginning “In this paper”] which “starts with a NARX network with enough degrees of freedom in both input and output memory or taps and then deleting those memory orders with small sensitivity measure after training. After pruning, the network is retrained” [Lin, section 1 last paragraph]. Determining the input-memory order (denoted D_u [Lin, section 2 first paragraph]) based on a trained neural network as described in Lin amounts to a selection of times corresponding to the explanatory variables “based on the coefficients” as recited by the claim (with the “coefficients” as mapped above).)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural networks of Mavrovouniotis with NARX networks as described in Lin because they are a “popular subclass of recurrent networks and have been used in many applications” [Lin, abstract] which are not only “computationally powerful in theory, but they [also] have several advantages in practice” [Lin, section I first paragraph], and with the delay damage algorithm of Lin “demonstrate[s] improved performance on both nonlinear predictions and grammatical inference tasks… [and] results in sparsely connected architectures but with long time windows that are able to model the global features of the underlying system quite efficiently” [Lin, section V paragraph beginning “In this paper”], so the combination would make for efficient and effective time series forecasting.
Claim 18
Mavrovouniotis in view of Lin discloses the elements of the parent claim(s). It also discloses:
[The information processing apparatus according to 16, wherein the processing circuitry is configured to] assign a weight to the first data, based on a value of the second variable in the first data, and generate the prediction model, based on the weight. ([Mavrovouniotis, section 3.3]: As noted above, connections between nodes in a neural network are given by “weights (also called connection strengths)” [Mavrovouniotis, section 3.3 first paragraph]. Any of these weights can be mapped to a “weight” as recited by the claim; it is “assign[ed]… to the first data” in the sense that it is determined by training based on the training examples, and the prediction model is “generate[d]… based on the weight” since the trained model uses those weights. For an alternative mapping of weights assigned to data, the applicant is invited to consult the rejection of dependent claim 19 below.)
Claim 25
Mavrovouniotis discloses:
An information processing method, comprising: performing regression of an objective variable at a prediction target time using one or more explanatory variables corresponding to a plurality of times before the prediction target time, based on time-series data of the explanatory variables and the objective variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having an output variable and input variables V_i(k) for varying V, i, and k, where the value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Moreover, while Mavrovouniotis focuses attention on networks whose output is “1 if there is a fault, and 0 if there is no fault”, it also clearly indicates that “the hierarchical approach we will discuss… is not specific to this task and could also be used, for example, within networks that carry out… prediction of future values of variables” [Mavrovouniotis, section 2.2 paragraph beginning “To focus our attention”]. In other words, Mavrovouniotis does distinctly disclose the use of hierarchical neural networks for time-series regressions. The output variable of the time-series regression maps to the “objective variable at a prediction target time” of the claim, and the input variables of the time-series regression maps to the “one or more explanatory variables corresponding to a plurality of times before the prediction target time” of the claim. The values of the V_i(k) for varying V, i, and k are the “time-series data of the explanatory variables and the objective variable” of the claim.)
calculating coefficients corresponding to each time point of the explanatory variables; ([Mavrovouniotis, section 3.3]: Mavrovouniotis discloses calculating “weights (also called connection strengths)” between nodes in the neural network [Mavrouvouniotis, section 3.3 first paragraph]. Since there are input nodes for each input variable at each time point, the weights associated to connections coming out of the input nodes map to the “coefficients corresponding to each time point of the explanatory variables” of the claim.)
creating first data including the selected explanatory variables as a plurality of first variables and the objective variable at the prediction target time as a second variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables and the and output variable map, respectively, to the “first variables” and “second variable” of the claim. The data set used for training and testing maps to the “first data” of the claim (since it includes all of the input variables, it a fortiori includes the “selected explanatory variables” as recited by the claim).)
generating one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph], any of which can map to the “one or more grouping candidates” of the claim.)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the respective groups” as recited in the claim) with predicted values of the output variable. All of the architectures incorporate a way of grouping the input variables. In other words, the architectures map to both the “one or more model architecture candidates” and the “combinations of the grouping candidates and one or more model architecture candidates” of the claim, and the trained neural networks having those architectures map to the “prediction models” of the claim.)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1]. Any of these three architectures maps to the “model architecture” of the claim, and the grouping used to generate that architecture maps to the “grouping” of the claim.)
Mavrovouniotis might not distinctly disclose:
selecting, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times;
Lin is in the field of machine learning. It discloses the use of recurrent neural networks [Lin, figure 1] to model nonlinear autoregressive models with exogenous inputs (NARX models) [Lin, section I first paragraph or section II first paragraph]. The examiner notes that the applicant’s regression model (an example of which is depicted in figure 8 of the application) is a NARX model where the exogenous input variable u [Lin, section 2 equation (1)] is taken to be the vector (X_1, X_2, …) (and u can in fact be a vector, as witnessed more explicitly in a reference to which Lin cites, namely, Leontartis as cited in the conclusion of a previous Office action). Moreover, Mavrovouniotis in view of Lin discloses:
selecting, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times; ([Lin, sections 1-2]: Lin discloses “a pruning-based algorithm (the delay damage algorithm) to determine the optimal memory-order of NARX… neural networks” [Lin, section 1 paragraph beginning “In this paper”] which “starts with a NARX network with enough degrees of freedom in both input and output memory or taps and then deleting those memory orders with small sensitivity measure after training. After pruning, the network is retrained” [Lin, section 1 last paragraph]. Determining the input-memory order (denoted D_u [Lin, section 2 first paragraph]) based on a trained neural network as described in Lin amounts to a selection of times corresponding to the explanatory variables “based on the coefficients” as recited by the claim (with the “coefficients” as mapped above).)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural networks of Mavrovouniotis with NARX networks as described in Lin because they are a “popular subclass of recurrent networks and have been used in many applications” [Lin, abstract] which are not only “computationally powerful in theory, but they [also] have several advantages in practice” [Lin, section I first paragraph], and with the delay damage algorithm of Lin “demonstrate[s] improved performance on both nonlinear predictions and grammatical inference tasks… [and] results in sparsely connected architectures but with long time windows that are able to model the global features of the underlying system quite efficiently” [Lin, section V paragraph beginning “In this paper”], so the combination would make for efficient and effective time series forecasting.
Claim 26
Mavrovouniotis discloses:
A non-transitory computer-readable medium having a computer program stored therein which causes a computer to perform processes, comprising: ([Mavrovouniotis, section 3]: Mavrovouniotis discloses using “Macintosh II computers” to implement the methods disclosed therein [Mavrovouniotis, section 3 paragraph beginning “The software”]. The Macintosh II computer is the “computer” of the claim and its hard drive is the “non-transitory computer readable medium having a computer program stored therein” of the claim.)
performing regression of an objective variable at a prediction target time using one or more explanatory variables corresponding to a plurality of times before the prediction target time, based on time-series data of the explanatory variables and the objective variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis discloses systems having an output variable and input variables V_i(k) for varying V, i, and k, where the value of an input variable V_i in Mavrovouniotis at time t - kΔt is denoted V_i(k) [Mavrovouniotis, section 2.2 paragraph beginning “To focus” and section 3 paragraph beginning “The main example”]. Moreover, while Mavrovouniotis focuses attention on networks whose output is “1 if there is a fault, and 0 if there is no fault”, it also clearly indicates that “the hierarchical approach we will discuss… is not specific to this task and could also be used, for example, within networks that carry out… prediction of future values of variables” [Mavrovouniotis, section 2.2 paragraph beginning “To focus our attention”]. In other words, Mavrovouniotis does distinctly disclose the use of hierarchical neural networks for time-series regressions. The output variable of the time-series regression maps to the “objective variable at a prediction target time” of the claim, and the input variables of the time-series regression maps to the “one or more explanatory variables corresponding to a plurality of times before the prediction target time” of the claim. The values of the V_i(k) for varying V, i, and k are the “time-series data of the explanatory variables and the objective variable” of the claim.)
calculating coefficients corresponding to each time point of the explanatory variables; ([Mavrovouniotis, section 3.3]: Mavrovouniotis discloses calculating “weights (also called connection strengths)” between nodes in the neural network [Mavrouvouniotis, section 3.3 first paragraph]. Since there are input nodes for each input variable at each time point, the weights associated to connections coming out of the input nodes map to the “coefficients corresponding to each time point of the explanatory variables” of the claim.)
creating first data including the selected explanatory variables as a plurality of first variables and the objective variable at the prediction target time as a second variable; ([Mavrovouniotis, sections 2-3]: Mavrovouniotis also discloses a dataset containing “36 training examples and 24 testing examples” [Mavrovouniotis, section 3 paragraph beginning “We used”]. The input variables and the and output variable map, respectively, to the “first variables” and “second variable” of the claim. The data set used for training and testing maps to the “first data” of the claim (since it includes all of the input variables, it a fortiori includes the “selected explanatory variables” as recited by the claim).)
generating one or more grouping candidates by dividing the plurality of first variables in the first data into a plurality of groups using a predefined automatic generation method or a grouping method specified by a user; ([Mavrovouniotis, section 2.3]: Mavrovouniotis discloses at least three ways of grouping inputs [Mavrovouniotis, section 2.3 first paragraph], any of which can map to the “one or more grouping candidates” of the claim.)
generating, for each combination of the grouping candidates and one or more model architecture candidates, a prediction model configured to associate the first variables included in the respective groups with a predicted value of the second variable; ([Mavrovouniotis, table 1 and figures 8-14]: Mavrovouniotis discloses training neural networks of at least seven architectures listed in [Mavrovouniotis, table 1] and depicted in [Mavrovouniotis, figures 8-14], all of which associate the input variables (i.e., the “first variables included in the respective groups” as recited in the claim) with predicted values of the output variable. All of the architectures incorporate a way of grouping the input variables. In other words, the architectures map to both the “one or more model architecture candidates” and the “combinations of the grouping candidates and one or more model architecture candidates” of the claim, and the trained neural networks having those architectures map to the “prediction models” of the claim.)
calculating an evaluation value for each prediction model based on a difference between the predicted value of the second variable and a value of the second variable in the first data; ([Mavrovouniotis, table 1]: Mavrovouniotis discloses counting the number of “[m]isclassified testing patterns” for each of the seven architectures considered therein. The number of misclassified testing patterns maps to the “evaluation value” of the claim since it is “based on a difference between the predicted value of the second variable and a value of the second variable in the first data” as recited by the claim.)
and determining a grouping and a model architecture to be used, based on the evaluation value corresponding to each combination of the grouping candidates and the model architecture candidates. ([Mavrovouniotis, section 3.2 and table 1]: Mavrovouniotis discusses identifying networks exhibiting the “best behavior” using the numbers of misclassified cases [Mavrovouniotis, section 3.2]. There are three architectures which minimize this number [Mavrovouniotis, table 1]. Any of these three architectures maps to the “model architecture” of the claim, and the grouping used to generate that architecture maps to the “grouping” of the claim.)
Mavrovouniotis might not distinctly disclose:
selecting, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times;
Lin is in the field of machine learning. It discloses the use of recurrent neural networks [Lin, figure 1] to model nonlinear autoregressive models with exogenous inputs (NARX models) [Lin, section I first paragraph or section II first paragraph]. The examiner notes that the applicant’s regression model (an example of which is depicted in figure 8 of the application) is a NARX model where the exogenous input variable u [Lin, section 2 equation (1)] is taken to be the vector (X_1, X_2, …) (and u can in fact be a vector, as witnessed more explicitly in a reference to which Lin cites, namely, Leontartis as cited in the conclusion of a previous Office action). Moreover, Mavrovouniotis in view of Lin discloses:
selecting, based on the calculated coefficients, a subset of the explanatory variables corresponding to the plurality of times; ([Lin, sections 1-2]: Lin discloses “a pruning-based algorithm (the delay damage algorithm) to determine the optimal memory-order of NARX… neural networks” [Lin, section 1 paragraph beginning “In this paper”] which “starts with a NARX network with enough degrees of freedom in both input and output memory or taps and then deleting those memory orders with small sensitivity measure after training. After pruning, the network is retrained” [Lin, section 1 last paragraph]. Determining the input-memory order (denoted D_u [Lin, section 2 first paragraph]) based on a trained neural network as described in Lin amounts to a selection of times corresponding to the explanatory variables “based on the coefficients” as recited by the claim (with the “coefficients” as mapped above).)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural networks of Mavrovouniotis with NARX networks as described in Lin because they are a “popular subclass of recurrent networks and have been used in many applications” [Lin, abstract] which are not only “computationally powerful in theory, but they [also] have several advantages in practice” [Lin, section I first paragraph], and with the delay damage algorithm of Lin “demonstrate[s] improved performance on both nonlinear predictions and grammatical inference tasks… [and] results in sparsely connected architectures but with long time windows that are able to model the global features of the underlying system quite efficiently” [Lin, section V paragraph beginning “In this paper”], so the combination would make for efficient and effective time series forecasting.
Claim(s) 17 is/are rejected under 35 USC 103 as being unpatentable over Mavrovouniotis in view of Lin, further in view of Devendra VERMA et al. (Use Genetic Programming for Selecting Predictor Variables and Modeling in Process Identification, published 2016; hereafter “Verma”).
Claim 17
Mavrovouniotis in view of Lin discloses the elements of the parent claim(s). It also discloses:
[The information processing apparatus according to claim 16, wherein the processing circuitry is configured to] generate the first variables by combining the explanatory variables corresponding to the plurality of times and at least one operator, ([Mavrovouniotis, sections 2-3]: As noted under the parent claim, the “first variables” of the claim (as mapped above) are obtained by “combining the explanatory variables corresponding to the plurality of times” as recited by the claim, where the “explanatory variables corresponding to the plurality of times” are as mapped under the parent claim. Any of a variety of elements fall under the broadest reasonable interpretation of the “at least one operator” of the claim. For example, as noted under the parent claim, each variable V in Mavrovouniotis is a function of both the stream i and the time-point index k, [Mavrovouniotis, section 2.2], so V regarded as such a function can map to the “at least one operator” of the claim.)
Mavrovouniotis in view of Lin might not distinctly disclose:
[generate the first variables…] based on a genetic algorithm.
Verma is in the field of time series forecasting [Verma, equation (1)]. Moreover, Mavrovouniotis in view of Lin and Verma discloses:
[generate the first variables…] based on a genetic algorithm. ([Verma, title and section 1]: Verma discloses the use of genetic programming (GP) to select predictor variables [Verma, title], indicating that “GP selects only those time delayed inputs and outputs as predictors in (1), which significantly influence the one-step ahead-output (y_{t+1})” [Verma, section 1 paragraph beginning “In the context of process identification”]. Recall that the input/predictor variables of the neural network of Mavrovouniotis are mapped to the “first variables” of the claim. In the combination, genetic programming as in Verma is used to select the input variables, i.e., to “generate the first variables” as recited by the claim.)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural networks disclosed by Mavrovouniotis in view of Lin with the use of genetic programming to select predictor variables as discussed in Verma because “automatic selection of the important predictor variables by the GP formalism is immensely beneficial in practice since it substantially reduces the computational time and efforts required in identifying the specific time-delayed inputs and outputs in (1)” [Verma, section 1 paragraph beginning “In the context of process identification”], so the combination would be more efficient overall.
Claim(s) 19 is/are rejected under 35 USC 103 as being unpatentable over Mavrovouniotis in view of Lin, further in view of Kriti KUMAR (US20170185902A1, published 2017-06-29; hereafter “Kumar”).
Claim 19
Mavrovouniotis in view Lin discloses the elements of the parent claim(s). It might not distinctly disclose:
[The information processing apparatus according to claim 18, wherein the processing circuitry is configured to] assign a first weight to the first data when the value of the second variable corresponding to a peak portion, and assign a second weight smaller than the first weight when the value of the second variable corresponds to a non-peak portion.
Kumar is in the field of time series forecasting [Kumar, abstract]. Moreover, Kumar discloses:
[The information processing apparatus according to claim 18, wherein the processing circuitry is configured to] assign a first weight to the first data when the value of the second variable corresponding to a peak portion, and assign a second weight smaller than the first weight when the value of the second variable corresponds to a non-peak portion. ([Kumar, 0021, 0044, and figure 5]: Kumar discloses that “response time values of the enterprise system beyond the critical threshold are termed as peaks” [Kumar, 0021] and also discloses “applying higher weights to time instants in the past where system response time is identified as ‘peaks’” [Kumar, 0044]. In the combination, the system response time variable (depicted on the y-axis of [Kumar, figure 5]) is taken to be the output variable in Mavrovouniotis and maps to the “second variable” of the claim. In other words, one of the time instants where the system response time was identified as having a peak maps to a “peak portion” of the claim, and a higher weight that is assigned to one of these time instants maps to the “first weight” of the claim. The examiner notes that the fact that the weight is “higher” implies that there are other, non-peak, time instants to which lower weights are applied, and one of those lower weights maps to the “second weight” of the claim, so that “the second weight [is] smaller than the first weight” as recited by the claim.)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural networks of Mavrovouniotis in view of Lin with time series forecasting methods of Kumar because they allow for “[p]redicting well in advance… so that timely interventions can be actuated” [Kumar, 0004], so the combination would be more effective overall.
Claim(s) 20 is/are rejected under 35 USC 103 as being unpatentable over Mavrovouniotis in view Lin, further in view of Michael MCCORMACK et al. (US20060149769A1, published 2006-07-06; hereafter “McCormack”)
Claim 20
Mavrovouniotis in view of Lin discloses the elements of the parent claim(s). It might not distinctly disclose:
[The information processing apparatus according to claim 16, further comprising] a graphical user interface circuit configured to allow a user to select whether to use the predefined automatic generation method or the grouping method specified by the user, wherein the processing circuitry is configured to determine which method to use based on the user selection.
McCormack is in the field of data analysis. Moreover, Mavrovouniotis in view of Lin and McCormack discloses:
[The information processing apparatus according to claim 16, further comprising] a graphical user interface circuit configured to allow a user to select whether to use the predefined automatic generation method or the grouping method specified by the user, wherein the processing circuitry is configured to determine which method to use based on the user selection. ([McCormack, 0181 and figure 15B]: McCormack discloses a user interface having a “group by submenu” [McCormack, 0181] which allows a user to choose whether to group data in a “user-designated or defined (e.g., custom)” way or one of several “predefined group bys” [McCormack, 0181; see figure 15B]. A custom group by maps to the “grouping method specified by the user” and any of the predefined group bys maps to the “predefined automatic generation method” of the claim.)
Before the effective filing date of the invention, it would have been obvious to a person of ordinary skill in the art to combine the hierarchical neural networks of Mavrovouniotis in view of Lin with the interface for selecting grouping methods disclosed by McCormack because it would allow for conveniently switching between different grouping methods.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Shishir AGRAWAL whose telephone number is (703)756-1183. The examiner can normally be reached Monday through Friday, 08:00-16:00 Pacific Time.
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, Alexey SHMATOV can be reached on +1 571-270-3428. The fax phone number for the organization where this application or proceeding is assigned is +1 571-273-8300.
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 +1 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call +1 800-786-9199 (IN USA OR CANADA) or +1 571-272-1000.
/S.A./Examiner, Art Unit 2123
/ALEXEY SHMATOV/Supervisory Patent Examiner, Art Unit 2123
Prosecution Insights