Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
The present application is being examined under the claims filed on November 15, 2023.
Claims 1-20 are pending.
Drawings
The drawings, filed November 15, 2023, are objected to because FIG. 2B contains text that is not oriented in the same direction as the view [see CFR 1.84(p)(1)]. Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all of the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. The figure or figure number of an amended drawing should not be labeled as “amended.” If a drawing figure is to be canceled, the appropriate figure must be removed from the replacement sheet, and where necessary, the remaining figures must be renumbered and appropriate changes made to the brief description of the several views of the drawings for consistency. Additional replacement sheets may be necessary to show the renumbering of the remaining figures. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 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 U.S.C. 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.
Claims 1, 3-4, 7-8, 13-16, and 18-19 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 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 U.S.C. 112, the applicant), regards as the invention.
Regarding Claim 1:
Claim 1 recites “a first class of a plurality of weak learner classes” and “a second class of a plurality of weak learner classes”. It is unclear if Applicant intends for the first and second classes to be referring to the same plurality of weak learner classes or two different plurality of weak learner classes. Examiner is interpreting the claim to be referring same plurality of weak learner classes. Applicant is advised to amend “wherein the second weak learner corresponds to a second class of a plurality of weak learner classes” as “a second class of the plurality of weak learner classes”.
Regarding Claims 3-4:
Claim 3 recites “a second class of a plurality of weak learner classes” and claim 2 recites “a first class of a plurality of weak learner classes”. It is unclear if Applicant intends for the first and second classes to be referring to the same plurality of weak learner classes or two different plurality of weak learner classes. Examiner is interpreting the claim to be referring same plurality of weak learner classes. Applicant is advised to amend “wherein the second weak learner corresponds to a second class of a plurality of weak learner classes” as “a second class of the plurality of weak learner classes”.
Claim 4 is rejected for inheriting the deficiencies of claim 3.
Regarding Claims 7-8:
Relative term The term “best-fit criteria” in claims 7 and 8 is a relative term which renders the claim indefinite. The term “best-fit criteria” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
Claim 8 is rejected to for inheriting the deficiencies of claim 7.
Regarding Claims 13-16:
Claim 13 recites “a second class of a plurality of weak learner classes” and claim 2 recites “a first class of a plurality of weak learner classes”. It is unclear if Applicant intends for the first and second classes to be referring to the same plurality of weak learner classes or two different plurality of weak learner classes. Examiner is interpreting the claim to be referring same plurality of weak learner classes. Applicant is advised to amend “wherein the second weak learner corresponds to a second class of a plurality of weak learner classes” as “a second class of the plurality of weak learner classes”.
Claims 14-16 is rejected for inheriting the deficiencies of claim 13.
Regarding Claims 18-19:
Claim 18 recites “a second class of a plurality of weak learner classes” and claim 17 recites “a first class of a plurality of weak learner classes”. It is unclear if Applicant intends for the first and second classes to be referring to the same plurality of weak learner classes or two different plurality of weak learner classes. Examiner is interpreting the claim to be referring same plurality of weak learner classes. Applicant is advised to amend “wherein the second weak learner corresponds to a second class of a plurality of weak learner classes” as “a second class of the plurality of weak learner classes”.
Claim 19 is rejected for inheriting the deficiencies of claim 18.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Claim 1 is directed to a system [machine].
Claims 2-16 are directed to a method [process].
Claims 17-20 are directed to a non-transitory computer readable medium [machine].
Regarding Claim 1:
Step 2A, Prong 1: The following limitations are directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind or with pen and paper (including an observation, evaluation, judgement, or opinion).
generating a first feature input based on the first dataset
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class
adjusting a first weight based on comparing the first output to the first validation metric
generating a second feature input based on the first weight and the first output
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class
adjusting a second weight based on comparing the second output to a second validation metric
(a) generating…a classification for data in the first dataset based on the second weight and the second output
As drafted, under their broadest reasonable interpretation (BRI), in view of the specification, the above limitations cover concepts performed in the human mind (observation, evaluation, judgement, or opinion). Given a sufficiently small set of data, nothing in the claim prohibits this process from being performed mentally or with pen and paper.
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
A system of generating artificial intelligence models featuring flexible ensembles of weak learners using broken series training, the system comprising: one or more processors; and one or more non-transitory, computer readable mediums comprising instructions recorded thereon that, when executed by the one or more processors, cause operations comprising:
inputting the first feature input into a first weak learner of a plurality of weak learners in the artificial intelligence model, wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
inputting the second feature input into a second weak learner of the plurality of weak learners, wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
(a) …from the artificial intelligence model…
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
receiving a first dataset for classification using an artificial intelligence model comprising an ensemble of weak learners
receiving a first output from the first weak learner
receiving a second output from the second weak learner
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
A system of generating artificial intelligence models featuring flexible ensembles of weak learners using broken series training, the system comprising: one or more processors; and one or more non-transitory, computer readable mediums comprising instructions recorded thereon that, when executed by the one or more processors, cause operations comprising:
inputting the first feature input into a first weak learner of a plurality of weak learners in the artificial intelligence model, wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
inputting the second feature input into a second weak learner of the plurality of weak learners, wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
(a) …from the artificial intelligence model…
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Intellectual Ventures v. Symantec, 838 F.3d 1307, 1321; 120 USPQ2d 1353, 1362 (Fed. Cir. 2016)) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
receiving a first dataset for classification using an artificial intelligence model comprising an ensemble of weak learners
receiving a first output from the first weak learner
receiving a second output from the second weak learner
Regarding Claim 2:
Step 2A, Prong 1: The following limitations are directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind or with pen and paper (including an observation, evaluation, judgement, or opinion).
generating a first feature input based on the first dataset
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class
adjusting a first weight based on comparing the first output to the first validation metric
(a) generating…a second feature input based on the first weight and the first output
generating a classification for data in the first dataset based on the second feature input
As drafted, under their broadest reasonable interpretation (BRI), in view of the specification, the above limitations cover concepts performed in the human mind (observation, evaluation, judgement, or opinion). Given a sufficiently small set of data, nothing in the claim prohibits this process from being performed mentally or with pen and paper.
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
A method of generating flexible ensembles of weak learners using broken series training, the method comprising:
inputting the first feature input into a first weak learner of a plurality of weak learners, wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
(a) …for a second weak learner of the plurality of weak learners…
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
receiving a first dataset
receiving a first output from the first weak learner
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
A method of generating flexible ensembles of weak learners using broken series training, the method comprising:
inputting the first feature input into a first weak learner of a plurality of weak learners, wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
(a) …for a second weak learner of the plurality of weak learners…
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
receiving a first dataset
receiving a first output from the first weak learner
Regarding Claim 3:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein generating the classification for data in the first dataset based on the second feature input further comprises:
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class
adjusting a second weight based on comparing the second output to the second validation metric
generating a classification for data in the first dataset based on the second weight and the second output
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
inputting the second feature input into the second weak learner of the plurality of weak learners, wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
receiving a second output from the second weak learner
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
inputting the second feature input into the second weak learner of the plurality of weak learners, wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
receiving a second output from the second weak learner
Regarding Claim 4:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein generating the classification for the data in the first dataset based on the second weight and the second output further comprises:
combining the first output and the second output
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction
Regarding Claim 5:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein comparing the first output to the first validation metric further comprises:
determining that the first weak learner has the first class of the plurality of weak learner classes
selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Regarding Claim 6:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein selecting the first validation metric from the plurality of validation metrics based on the first validation metric corresponding to the first class further comprises:
determining a data characteristic of the first dataset
filtering the plurality of validation metrics based on the data characteristic
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Regarding Claim 7:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
determining a best-fit criteria for a first round of a weak learner ensemble
selecting the first weak learner from the plurality of weak learners based on the first weak learner corresponding to the best-fit criteria
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
wherein inputting the first feature input into the first weak learner of the plurality of weak learners further comprises:
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
wherein inputting the first feature input into the first weak learner of the plurality of weak learners further comprises:
Regarding Claim 8:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein determining the best-fit criteria for the first round of the weak learner ensemble further comprises:
determining a data characteristic of the first dataset
filtering available criteria for the first round of the weak learner ensemble based on the data characteristic
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Regarding Claim 9:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
calculating a gradient of a loss function corresponding to the first output
determining a difference between a predicted value and the target value
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
wherein receiving the first output from the first weak learner further comprises:
receiving a target value for regression tasks of a weak learner ensemble
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
wherein receiving the first output from the first weak learner further comprises:
receiving a target value for regression tasks of a weak learner ensemble
Regarding Claim 10:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
calculating a step size for the weak learner ensemble based on the learning rate
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
wherein receiving the first output from the first weak learner further comprises:
receiving a learning rate for a weak learner ensemble
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
wherein receiving the first output from the first weak learner further comprises:
receiving a learning rate for a weak learner ensemble
Regarding Claim 11:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein generating the first feature input based on the first dataset further comprises:
separating the first dataset into a training dataset and a validation dataset
determining the first feature input based on the training dataset
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Regarding Claim 12:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
determining a number of rounds in a weak learner ensemble
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
generating the weak learner ensemble with the number of rounds
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
generating the weak learner ensemble with the number of rounds
Regarding Claim 13:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
determining a first format of the first output
converting the first format to a second format, wherein the second format corresponds to a second class of a plurality of weak learner classes
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
wherein receiving the first output from the first weak learner further comprises:
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
wherein receiving the first output from the first weak learner further comprises:
Regarding Claim 14:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein converting the first format to the second format further comprises:
determining a first mapping for the first output based on the first class
determining a second mapping for the first output based on the second class
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Regarding Claim 15:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein converting the first format to the second format further comprises:
determining a first probability for the first output based on the first class
determining a second probability for the first output based on the second class
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Regarding Claim 16:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein converting the first format to the second format further comprises:
determining a first extraction requirement for the first output based on the first class
applying the first extraction requirement to the first output
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Regarding Claim 17:
Step 2A, Prong 1: The following limitations are directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind or with pen and paper (including an observation, evaluation, judgement, or opinion).
generating a first feature input based on a first dataset
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class
adjusting a first weight based on comparing the first output to the first validation metric
(a) generating…a second feature input based on the first weight and the first output
As drafted, under their broadest reasonable interpretation (BRI), in view of the specification, the above limitations cover concepts performed in the human mind (observation, evaluation, judgement, or opinion). Given a sufficiently small set of data, nothing in the claim prohibits this process from being performed mentally or with pen and paper.
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
One or more non-transitory, computer readable mediums comprising instructions recorded thereon that, when executed by one or more processors, cause operations comprising:
inputting the first feature input into a first weak learner of a plurality of weak learners, wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
(a) …for a second weak learner of the plurality of weak learners…
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
receiving a first output from the first weak learner
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
One or more non-transitory, computer readable mediums comprising instructions recorded thereon that, when executed by one or more processors, cause operations comprising:
inputting the first feature input into a first weak learner of a plurality of weak learners, wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
(a) …for a second weak learner of the plurality of weak learners…
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
receiving a first output from the first weak learner
Regarding Claim 18:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
generating a classification for data in the first dataset based on the second feature input by:
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class
adjusting a second weight based on comparing the second output to the second validation metric
generating a classification for data in the first dataset based on the second weight and the second output
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
wherein the instructions further cause operations comprising
inputting the second feature input into the second weak learner of the plurality of weak learners, wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
The following additional elements are directed to insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(g)].
receiving a second output from the second weak learner
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
wherein the instructions further cause operations comprising
inputting the second feature input into the second weak learner of the plurality of weak learners, wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
The following additional elements are directed to receiving or transmitting data over a network. The courts (as per Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362) have recognized receiving or transmitting data over a network as well-understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity to the judicial exception [see MPEP 2106.05(d) II.].
receiving a second output from the second weak learner
Regarding Claim 19:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein generating the classification for the data in the first dataset based on the second weight and the second output further comprises:
combining the first output and the second output
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to integrate the judicial exception into a practical application.
generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
The following additional elements are adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea [see MPEP 2106.05(f)] and therefore fails to amount to significantly more than the judicial exception.
generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction
Regarding Claim 20:
Step 2A, Prong 1: This claim recites the same abstract ideas as in the parent claim. Additionally,
The following limitations are/remain directed to the abstract idea of a mental process [see MPEP 2106.04(a)(2) III. C.]. In particular, the claim recites mental processes that are concepts performed in the human mind (including an observation, evaluation, judgement, or opinion).
wherein comparing the first output to the first validation metric further comprises:
determining that the first weak learner has the first class of the plurality of weak learner classes
selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class
Step 2A, Prong 2: There are no additional elements in this claim that integrate the judicial exception into a practical application.
Step 2B: There are no additional elements in this claim that amount to significantly more than the judicial exception.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-4, 7-9, 11-12, and 17-19 are rejected under 35 U.S.C. 103 as being unpatentable over Barbu et al. (US 20080027887), hereinafter Barbu in view of Sabe et al. (US 20050213810), hereinafter Sabe, and further in view of Jebara et al. (US 20140029840), hereinafter Jebara.
Regarding Claim 1:
Barbu discloses:
A system of generating artificial intelligence models featuring flexible ensembles of weak learners using broken series training, the system comprising:
Barbu, [0019], “Embodiments of the invention are directed to a boosting--based method for fusing a set of classifiers that performs classification using weak learners trained on different views of the training data sampled with a shared sampling distribution.”
one or more processors; and one or more non-transitory, computer readable mediums comprising instructions recorded thereon that, when executed by the one or more processors, cause operations comprising:
Barbu, [0036], “Other embodiments of the invention include devices such as computer systems and computer-readable media having programs or applications to accomplish the exemplary methods described herein.”
receiving a first dataset for classification using an artificial intelligence model comprising an ensemble of weak learners
Barbu, [0021], “FIG. 2 describes the method of FIG. 1 in greater detail. Input 30 includes a training set S of data of N training points
X
=
{
x
1
,
x
2
…
x
N
}
. M disjoint features are available for each point
x
i
=
{
x
i
1
,
x
i
2
…
x
i
M
}
. Each member
x
i
j
in the set
x
i
is known as a view of point
x
i
. Each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example. Since M disjoint features are available for each point, there will be M training sets. The set S is represented as
S
j
=
x
1
j
,
y
1
,
x
2
j
,
y
2
…
x
N
j
,
y
N
, where
j
=
1
…
M
and
y
i
∈
+
1
,
-
1
and each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example.”
[0045], “FIGS. 3-5 show experimental results of testing the method of FIG. 3 (‘BSSD’, or ‘Boosting with Shared Sampling Distribution’) and fusion methods such as stacked generalization (stacking), semidefinite programming (SDP/SVM) and majority vote (SVM-MV) with five data sets from the MIT VisTex texture database. The size and dimensions of the five data sets are shown in the following table.”
PNG
media_image1.png
207
574
media_image1.png
Greyscale
In para. 21, Barbu teaches a training set S. Para. 45 states five data sets from a database, which is interpreted as receiving/obtaining the data set [receiving a first dataset]. Para. 19, cited above, also describes classifiers using weak learners [for classification using an artificial intelligence model comprising an ensemble of weak learners].
generating a first feature input based on the first dataset
As cited above in para. 21, Barbu teaches features for each training point of the training data set [generating a first feature input based on the first dataset].
inputting the first feature input into a first weak learner of a plurality of weak learners in the artificial intelligence model
Barbu, [0020], “In an initialization step 22, all the views for a given for a given training point are initialized with the same weight
w
1
i
=
1
/
N
. Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
[0022], “A view may be thought of as a representation of point
x
i
using disjoint feature sets. For instance, for a color image, each training point
x
i
may be thought of as a set of three views each consisting of three disjoint features obtained from the intensities of Red, Green and Blue color components. For an example where the three views are Red, Green, and Blue intensities, the views of the point
x
i
can be represented as
{
x
i
R
,
x
i
G
,
x
i
B
}
.”
In view of para. 21 above which describes the training data as input, para. 20 teaches training each classifier on each view of the training data [inputting the first feature input into a first weak learner of a plurality of weak learners in the artificial intelligence model].
receiving a first output from the first weak learner
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
Barbu teaches obtaining a weak hypothesis from the classifiers [receiving a first output from the first weak learner].
generating a second feature input based on the first weight and the first output
Barbu, [0020], “After a classifier
h
k
* with lowest error rate is selected and combination weight
a
k
* is obtained, the weights of the training data views are updated 25 forming a shared sampling distribution that will be used to sample training data at the next iteration.”
Barbu teaches sampling training data at the next iteration [generating a second feature input] based on the updated weights and classifier with a lowest error rate [based on the first weight and the first output].
inputting the second feature input into a second weak learner of the plurality of weak learners
Barbu, [0021], “FIG. 2 describes the method of FIG. 1 in greater detail. Input 30 includes a training set S of data of N training points
X
=
{
x
1
,
x
2
…
x
N
}
. M disjoint features are available for each point
x
i
=
{
x
i
1
,
x
i
2
…
x
i
M
}
. Each member
x
i
j
in the set
x
i
is known as a view of point
x
i
. Each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example. Since M disjoint features are available for each point, there will be M training sets. The set S is represented as
S
j
=
x
1
j
,
y
1
,
x
2
j
,
y
2
…
x
N
j
,
y
N
, where
j
=
1
…
M
and
y
i
∈
+
1
,
-
1
and each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example.”
[0020], “In an initialization step 22, all the views for a given for a given training point are initialized with the same weight
w
1
i
=
1
/
N
. Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
In view of para. 21 above which describes the training data as input, para. 20 teaches training each classifier on each view of the training data [inputting the second feature input into a second weak learner of the plurality of weak learners].
receiving a second output from the second weak learner
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
Barbu teaches obtaining a weak hypothesis from the classifiers [receiving a second output from the second weak learner].
generating, from the artificial intelligence model, a classification for data in the first dataset based on the second weight and the second output
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24…After a classifier
h
k
* with lowest error rate is selected and combination weight
a
k
* is obtained, the weights of the training data views are updated 25 forming a shared sampling distribution that will be used to sample training data at the next iteration. A final hypothesis or classifier is output 26 from the system.”
Barbu teaches after the weights and training views are updated [based on the second weight and the second output], a final hypothesis is output [generating, from the artificial intelligence model, a classification for data in the first dataset]
Barbu does not explicitly disclose:
wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class; adjusting a second weight based on comparing the second output to a second validation metric
However, in the same field, analogous art Sabe teaches:
wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
Sabe, [0068], “An information processing apparatus (e.g., an object detecting apparatus 5 shown in FIG. 5) according to the present invention includes upper nodes (e.g., a node 221-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1.sub.1 to 21-1.sub.100 shown in FIG. 24) that learns learning samples with a first label (e.g., labels 1 to 15 shown in FIG. 17) of a first range among learning samples classified with a plurality of labels (e.g., labels 1 to 15 shown in FIG. 17)”
Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [the first weak learner corresponds to a first class of a plurality of weak learner classes].
wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
Sabe, [0068], “first lower nodes (e.g., a node 221-1-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1-1.sub.1 to 21-1-1.sub.100) that learns learning samples with a second label (e.g., labels 1 to 5 shown in FIG. 17) of a second range based on results of classification by the upper nodes (e.g., the node 221-1 shown in FIG. 24), the second range being a part of the first range”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [the second weak learner corresponds to a second class of a plurality of weak learner classes].
Barbu, Sabe, and the instant application are analogous art because they are all directed to ensemble learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Barbu in view of Sabe do not explicitly disclose:
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class; adjusting a second weight based on comparing the second output to a second validation metric
However, in the same field, analogous art Jebara teaches:
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
Jebara, [0025], “The AdaBoost algorithm assigns a weight w.sub.i to each training example. In each step of the AdaBoost algorithm, a weak learner G.sup.s() is obtained on the weighted examples and a coefficient .alpha..sub.s is assigned to it. Thus, the AdaBoost algorithm iteratively builds .SIGMA..sub.s=1.sup.s.alpha..sub.sG.sup.s(). If a training example is correctly classified, its weight is exponentially decreased; if it is misclassified, its weight is exponentially increased. The process is repeated until a stopping criterion is met.”
Jebara teaches training a weak learner with weights. The weights are adjusted [adjusting a first weight] based on whether the training example was correctly classified [comparing the first output to the first validation metric, wherein the first validation metric is based on the first class]
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class; adjusting a second weight based on comparing the second output to a second validation metric
Jebara, [0015], “The general idea behind boosting is that it takes a group of simple, but poor classifiers (‘weak learners’), and combines them to form a single classifier with much higher accuracy (a ‘strong learner’). This is accomplished by iteratively testing classifiers on a set of training data, and weighting the data based on each classifier's efficacy (i.e. how much error there is in identification).”
As cited above in para. 25 and further in view of para. 15, the AdaBoost algorithm iteratively repeats itself and therefore corresponds to the claimed language for a second output, second validation metric, and second weight.
Barbu, Sabe, Jebara and the instant application are analogous art because they are all directed to ensemble learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu and Sabe with Jebara in order to improve the optimization of the overall classifier. “Classifiers are methods for automating the recognition of an applied signal or data, usually one with a lot of dimensions. Classifiers are used in many applications in image processing including computer vision, medical imaging (X-ray, MRI), and face detection. Boosting works by taking many simple but weak classifiers and combining them to form an improved classifier (a composite classifier or a "strong learner"). As in Adaboost, this is accomplished by iteratively testing the weak classifiers on an example set of data. After a given classifier is tested on the data, examples that were improperly identified are given a greater weight for testing with the following classifier; as the algorithm progresses there is a greater focus on identifying examples that previous classifiers have failed on. The weighting of example data is done based on the error in identification; how "wrong" the function was. This leads to an optimized overall classifier” (Jebara, [0016]).
Regarding Claim 2:
Barbu discloses:
A method of generating flexible ensembles of weak learners using broken series training, the method comprising:
Barbu, [0019], “Embodiments of the invention are directed to a boosting--based method for fusing a set of classifiers that performs classification using weak learners trained on different views of the training data sampled with a shared sampling distribution.”
receiving a first dataset
Barbu, [0021], “FIG. 2 describes the method of FIG. 1 in greater detail. Input 30 includes a training set S of data of N training points
X
=
{
x
1
,
x
2
…
x
N
}
. M disjoint features are available for each point
x
i
=
{
x
i
1
,
x
i
2
…
x
i
M
}
. Each member
x
i
j
in the set
x
i
is known as a view of point
x
i
. Each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example. Since M disjoint features are available for each point, there will be M training sets. The set S is represented as
S
j
=
x
1
j
,
y
1
,
x
2
j
,
y
2
…
x
N
j
,
y
N
, where
j
=
1
…
M
and
y
i
∈
+
1
,
-
1
and each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example.”
[0045], “FIGS. 3-5 show experimental results of testing the method of FIG. 3 (‘BSSD’, or ‘Boosting with Shared Sampling Distribution’) and fusion methods such as stacked generalization (stacking), semidefinite programming (SDP/SVM) and majority vote (SVM-MV) with five data sets from the MIT VisTex texture database. The size and dimensions of the five data sets are shown in the following table.”
PNG
media_image1.png
207
574
media_image1.png
Greyscale
In para. 21, Barbu teaches a training set S. Para. 45 states five data sets from a database, which is interpreted as receiving/obtaining the data set [receiving a first dataset].
generating a first feature input based on the first dataset
As cited above in para. 21, Barbu teaches features for each training point of the training data set [generating a first feature input based on the first dataset].
inputting the first feature input into a first weak learner of a plurality of weak learners
Barbu, [0020], “In an initialization step 22, all the views for a given for a given training point are initialized with the same weight
w
1
i
=
1
/
N
. Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
[0022], “A view may be thought of as a representation of point
x
i
using disjoint feature sets. For instance, for a color image, each training point
x
i
may be thought of as a set of three views each consisting of three disjoint features obtained from the intensities of Red, Green and Blue color components. For an example where the three views are Red, Green, and Blue intensities, the views of the point
x
i
can be represented as
{
x
i
R
,
x
i
G
,
x
i
B
}
.”
In view of para. 21 above which describes the training data as input, para. 20 teaches training each classifier on each view of the training data [inputting the first feature input into a first weak learner of a plurality of weak learners].
receiving a first output from the first weak learner
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
Barbu teaches obtaining a weak hypothesis from the classifiers [receiving a first output from the first weak learner].
generating, for a second weak learner of the plurality of weak learners, a second feature input based on the first weight and the first output
Barbu, [0020], “After a classifier
h
k
* with lowest error rate is selected and combination weight
a
k
* is obtained, the weights of the training data views are updated 25 forming a shared sampling distribution that will be used to sample training data at the next iteration.”
Barbu teaches sampling training data at the next iteration [generating, for a second weak learner of the plurality of weak learners, a second feature] based on the updated weights and classifier with a lowest error rate [based on the first weight and the first output].
generating a classification for data in the first dataset based on the second feature input
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24…After a classifier
h
k
* with lowest error rate is selected and combination weight
a
k
* is obtained, the weights of the training data views are updated 25 forming a shared sampling distribution that will be used to sample training data at the next iteration. A final hypothesis or classifier is output 26 from the system.”
Barbu teaches after the weights and training views are updated, sample training data at the next interation [based on the second feature input], a final hypothesis is output [generating a classification for data in the first dataset]
Barbu does not explicitly disclose:
wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
However, in the same field, analogous art Sabe teaches:
wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
Sabe, [0068], “An information processing apparatus (e.g., an object detecting apparatus 5 shown in FIG. 5) according to the present invention includes upper nodes (e.g., a node 221-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1.sub.1 to 21-1.sub.100 shown in FIG. 24) that learns learning samples with a first label (e.g., labels 1 to 15 shown in FIG. 17) of a first range among learning samples classified with a plurality of labels (e.g., labels 1 to 15 shown in FIG. 17)”
Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [the first weak learner corresponds to a first class of a plurality of weak learner classes].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Barbu in view of Sabe do not explicitly disclose:
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
However, in the same field, analogous art Jebara teaches:
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
Jebara, [0025], “The AdaBoost algorithm assigns a weight w.sub.i to each training example. In each step of the AdaBoost algorithm, a weak learner G.sup.s() is obtained on the weighted examples and a coefficient .alpha..sub.s is assigned to it. Thus, the AdaBoost algorithm iteratively builds .SIGMA..sub.s=1.sup.s.alpha..sub.sG.sup.s(). If a training example is correctly classified, its weight is exponentially decreased; if it is misclassified, its weight is exponentially increased. The process is repeated until a stopping criterion is met.”
Jebara teaches training a weak learner with weights. The weights are adjusted [adjusting a first weight] based on whether the training example was correctly classified [comparing the first output to the first validation metric, wherein the first validation metric is based on the first class]
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Regarding Claim 3:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, and Barbu further teaches:
wherein generating the classification for data in the first dataset based on the second feature input further comprises: inputting the second feature input into the second weak learner of the plurality of weak learners
Barbu, [0021], “FIG. 2 describes the method of FIG. 1 in greater detail. Input 30 includes a training set S of data of N training points
X
=
{
x
1
,
x
2
…
x
N
}
. M disjoint features are available for each point
x
i
=
{
x
i
1
,
x
i
2
…
x
i
M
}
. Each member
x
i
j
in the set
x
i
is known as a view of point
x
i
. Each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example. Since M disjoint features are available for each point, there will be M training sets. The set S is represented as
S
j
=
x
1
j
,
y
1
,
x
2
j
,
y
2
…
x
N
j
,
y
N
, where
j
=
1
…
M
and
y
i
∈
+
1
,
-
1
and each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example.”
[0020], “In an initialization step 22, all the views for a given for a given training point are initialized with the same weight
w
1
i
=
1
/
N
. Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
In view of para. 21 above which describes the training data as input, para. 20 teaches training each classifier on each view of the training data [inputting the second feature input into a second weak learner of the plurality of weak learners].
receiving a second output from the second weak learner
[0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
Barbu teaches obtaining a weak hypothesis from the classifiers [receiving a second output from the second weak learner].
generating a classification for data in the first dataset based on the second weight and the second output
Sabe further teaches:
wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
Sabe, [0068], “first lower nodes (e.g., a node 221-1-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1-1.sub.1 to 21-1-1.sub.100) that learns learning samples with a second label (e.g., labels 1 to 5 shown in FIG. 17) of a second range based on results of classification by the upper nodes (e.g., the node 221-1 shown in FIG. 24), the second range being a part of the first range”
Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [the second weak learner corresponds to a second class of a plurality of weak learner classes].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Jebara further teaches:
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class; adjusting a second weight based on comparing the second output to the second validation metric
Jebara, [0025], “The AdaBoost algorithm assigns a weight w.sub.i to each training example. In each step of the AdaBoost algorithm, a weak learner G.sup.s() is obtained on the weighted examples and a coefficient .alpha..sub.s is assigned to it. Thus, the AdaBoost algorithm iteratively builds .SIGMA..sub.s=1.sup.s.alpha..sub.sG.sup.s(). If a training example is correctly classified, its weight is exponentially decreased; if it is misclassified, its weight is exponentially increased. The process is repeated until a stopping criterion is met.”
[0015], “The general idea behind boosting is that it takes a group of simple, but poor classifiers (‘weak learners’), and combines them to form a single classifier with much higher accuracy (a ‘strong learner’). This is accomplished by iteratively testing classifiers on a set of training data, and weighting the data based on each classifier's efficacy (i.e. how much error there is in identification).”
Jebara teaches training a weak learner with weights. The weights are adjusted [adjusting a second weight] based on whether the training example was correctly classified [comparing the second output to the second validation metric, wherein the second validation metric is based on the second class]
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu and Sabe with Jebara in order to improve the optimization of the overall classifier. “Classifiers are methods for automating the recognition of an applied signal or data, usually one with a lot of dimensions. Classifiers are used in many applications in image processing including computer vision, medical imaging (X-ray, MRI), and face detection. Boosting works by taking many simple but weak classifiers and combining them to form an improved classifier (a composite classifier or a "strong learner"). As in Adaboost, this is accomplished by iteratively testing the weak classifiers on an example set of data. After a given classifier is tested on the data, examples that were improperly identified are given a greater weight for testing with the following classifier; as the algorithm progresses there is a greater focus on identifying examples that previous classifiers have failed on. The weighting of example data is done based on the error in identification; how "wrong" the function was. This leads to an optimized overall classifier” (Jebara, [0016]).
Regarding Claim 4:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 3, and Sabe further teaches:
wherein generating the classification for the data in the first dataset based on the second weight and the second output further comprises: combining the first output and the second output; generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction
Sabe, [0102], “The ensemble learning machine 6 performs ensemble learning according to a boosting algorithm using a plurality of weak classifiers as described above so that a strong classification can be obtained as a result. Although each of the weak classifiers is constructed very simply, having low ability of classifying face and non-face by itself, high ability of classification can be achieved by combining, for example, on the order of hundreds to thousands of weak classifiers…Although each of the weak classifiers has low classification ability by itself, a classifier having high classification ability can be obtained depending on selection and combination of weak classifiers. Thus, the ensemble learning machine 6 learns a combination of weak classifiers, i.e., selection of weak classifiers and weights on output values of the respective weak classifiers for calculating a value of weighted majority.”
Sabe teaches combining a selection of weak classifiers [combining he first output and the second output] to obtain a strong classification [generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Regarding Claim 7:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, and Barbu further teaches:
wherein inputting the first feature input into the first weak learner of the plurality of weak learners further comprises: determining a best-fit criteria for a first round of a weak learner ensemble; and
Barbu, [0035], “This method performs classifier selection based on the lowest training error among the views at each iteration. In each iteration, sampling and weight update is performed using a shared sampling distribution.”
Barbu selects their classifier based on lowest training error. The lowest training error is thus the best-fit criteria [determining a best-fit criteria for a first round of a weak learner ensemble].
selecting the first weak learner from the plurality of weak learners based on the first weak learner corresponding to the best-fit criteria
As cited above, Barbu teaches selecting their classifier based on the lowest training error [selecting the first weak learner from the plurality of weak learners based on the first weak learner corresponding to the best-fit criteria].
Regarding Claim 8:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 7, and Jebara further teaches:
wherein determining the best-fit criteria for the first round of the weak learner ensemble further comprises: determining a data characteristic of the first dataset; and filtering available criteria for the first round of the weak learner ensemble based on the data characteristic
Jebara, [0013], “Embodiments of the disclosed subject matter relate generally to classifiers, particular types of classifiers made up of multiple, potentially similar and in embodiments, the same, component classifiers that are optimized or selected responsively to training data and combined to yield a composite classifier”
[0120], “The optimizing a weak learner or selecting a classifier from a pool of classifiers can include optimizing a weak learner. The optimizing a weak learner or selecting a classifier from a pool of classifiers can include selecting an optimal classifier from a pool of classifiers. The pool of classifiers can be adapted for responding to specific features of an object or event to be classified.”
In para. 13, Jebara teaches selecting a best weak learner by either optimization or selection from a pool of classifiers [filtering available criteria for the first round of the weak learner ensemble]. Jebara further specifies the pool selection method can be adapted for responding to specific features of an object [determining a data characteristic of the first dataset…based on the data characteristic.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu and Sabe with Jebara in order to improve the optimization of the overall classifier. “Classifiers are methods for automating the recognition of an applied signal or data, usually one with a lot of dimensions. Classifiers are used in many applications in image processing including computer vision, medical imaging (X-ray, MRI), and face detection. Boosting works by taking many simple but weak classifiers and combining them to form an improved classifier (a composite classifier or a "strong learner"). As in Adaboost, this is accomplished by iteratively testing the weak classifiers on an example set of data. After a given classifier is tested on the data, examples that were improperly identified are given a greater weight for testing with the following classifier; as the algorithm progresses there is a greater focus on identifying examples that previous classifiers have failed on. The weighting of example data is done based on the error in identification; how "wrong" the function was. This leads to an optimized overall classifier” (Jebara, [0016]).
Regarding Claim 9:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, and Barbu further teaches:
wherein receiving the first output from the first weak learner further comprises: receiving a target value for regression tasks of a weak learner ensemble
calculating a gradient of a loss function corresponding to the first output
determining a difference between a predicted value and the target value
Barbu, [0027], “Next, in step 35, obtain the training error rates
ϵ
k
j
of each
h
k
j
over the distribution Wk such that
ϵ
k
j
=
P
i
-
W
k
h
k
j
x
i
j
≠
y
i
”
[0026], “weak learners
h
R
,
h
G
, and
h
B
…”
Barbu teaches obtaining error rates for each learner [calculating a gradient of a loss function corresponding to the first output], the error rate is between the output of the learner,
h
k
j
x
i
j
, with the class label,
y
i
[receiving a target value for regression tasks of a weak learner ensemble…determining a different between a predicted value and the target value].
Regarding Claim 11:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, and Sabe further teaches:
wherein generating the first feature input based on the first dataset further comprises: separating the first dataset into a training dataset and a validation dataset; and determining the first feature input based on the training dataset
Sabe, [0256], “a weak classifier or a classifier generated may be evaluated using samples that are different from learning samples, as in cross validation or the jack-knife method. In cross validation, learning samples are uniformly divided into I units, learning is performed using the units other than one unit, and results of learning is evaluated using the one unit, and this process is repeated I times.”
Sabe teaches cross validation, splitting the learning samples [separating the first dataset] into learning units [into a training dataset] and results evaluation units [validation dataset]. As previously mentioned, learning samples are used for learning [determining the first feature input based on the training dataset].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Regarding Claim 12:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, and Barbu further teaches:
further comprising: determining a number of rounds in a weak learner ensemble; and generating the weak learner ensemble with the number of rounds
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24…After a classifier
h
k
* with lowest error rate is selected and combination weight
a
k
* is obtained, the weights of the training data views are updated 25 forming a shared sampling distribution that will be used to sample training data at the next iteration. A final hypothesis or classifier is output 26 from the system.”
[0035], “This results in a cyclic behavior and the classifiers that lose for some iterations eventually have a low weighted error rate and become a winning classifier for a given iteration. This cyclic behavior continues until there is a consensus between the classifiers over which training examples are difficult to classify. In the end, all weak learners from all classifiers agree on what examples are difficult and try to achieve a low error rate.”
In para. 20, Barbu teaches a number of iterations
k
=
1
to
k
=
k
m
a
x
[determining a number of rounds in a weak learner ensemble]. Para. 35 further details the cyclic behavior until the weak classifiers all agree [generating the weak learner ensemble with the number of rounds].
Regarding Claim 17:
Barbu discloses:
One or more non-transitory, computer readable mediums comprising instructions recorded thereon that, when executed by one or more processors, cause operations comprising:
Barbu, [0036], “Other embodiments of the invention include devices such as computer systems and computer-readable media having programs or applications to accomplish the exemplary methods described herein.”
generating a first feature input based on a first dataset
Barbu, [0021], “FIG. 2 describes the method of FIG. 1 in greater detail. Input 30 includes a training set S of data of N training points
X
=
{
x
1
,
x
2
…
x
N
}
. M disjoint features are available for each point
x
i
=
{
x
i
1
,
x
i
2
…
x
i
M
}
. Each member
x
i
j
in the set
x
i
is known as a view of point
x
i
. Each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example. Since M disjoint features are available for each point, there will be M training sets. The set S is represented as
S
j
=
x
1
j
,
y
1
,
x
2
j
,
y
2
…
x
N
j
,
y
N
, where
j
=
1
…
M
and
y
i
∈
+
1
,
-
1
and each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example.”
In para. 21, Barbu teaches features for each training point of a training data set S. [generating a first feature input based on a first dataset].
inputting the first feature input into a first weak learner of a plurality of weak learners
Barbu, [0020], “In an initialization step 22, all the views for a given for a given training point are initialized with the same weight
w
1
i
=
1
/
N
. Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
[0022], “A view may be thought of as a representation of point
x
i
using disjoint feature sets. For instance, for a color image, each training point
x
i
may be thought of as a set of three views each consisting of three disjoint features obtained from the intensities of Red, Green and Blue color components. For an example where the three views are Red, Green, and Blue intensities, the views of the point
x
i
can be represented as
{
x
i
R
,
x
i
G
,
x
i
B
}
.”
In view of para. 21 above which describes the training data as input, para. 20 teaches training each classifier on each view of the training data [inputting the first feature input into a first weak learner of a plurality of weak learners].
receiving a first output from the first weak learner
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
Barbu teaches obtaining a weak hypothesis from the classifiers [receiving a first output from the first weak learner].
generating, for a second weak learner of the plurality of weak learners, a second feature input based on the first weight and the first output
Barbu, [0020], “After a classifier
h
k
* with lowest error rate is selected and combination weight
a
k
* is obtained, the weights of the training data views are updated 25 forming a shared sampling distribution that will be used to sample training data at the next iteration.”
Barbu teaches sampling training data at the next iteration [generating, for a second weak learner of the plurality of weak learners, a second feature] based on the updated weights and classifier with a lowest error rate [based on the first weight and the first output].
Barbu does not explicitly disclose:
wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
However, in the same field, analogous art Sabe teaches:
wherein the first weak learner corresponds to a first class of a plurality of weak learner classes
Sabe, [0068], “An information processing apparatus (e.g., an object detecting apparatus 5 shown in FIG. 5) according to the present invention includes upper nodes (e.g., a node 221-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1.sub.1 to 21-1.sub.100 shown in FIG. 24) that learns learning samples with a first label (e.g., labels 1 to 15 shown in FIG. 17) of a first range among learning samples classified with a plurality of labels (e.g., labels 1 to 15 shown in FIG. 17)”
Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [the first weak learner corresponds to a first class of a plurality of weak learner classes].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination
Barbu in view of Sabe do not explicitly disclose:
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
However, in the same field, analogous art Jebara teaches:
comparing the first output to a first validation metric, wherein the first validation metric is based on the first class; adjusting a first weight based on comparing the first output to the first validation metric
Jebara, [0025], “The AdaBoost algorithm assigns a weight w.sub.i to each training example. In each step of the AdaBoost algorithm, a weak learner G.sup.s() is obtained on the weighted examples and a coefficient .alpha..sub.s is assigned to it. Thus, the AdaBoost algorithm iteratively builds .SIGMA..sub.s=1.sup.s.alpha..sub.sG.sup.s(). If a training example is correctly classified, its weight is exponentially decreased; if it is misclassified, its weight is exponentially increased. The process is repeated until a stopping criterion is met.”
Jebara teaches training a weak learner with weights. The weights are adjusted [adjusting a first weight] based on whether the training example was correctly classified [comparing the first output to the first validation metric, wherein the first validation metric is based on the first class]
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu and Sabe with Jebara in order to improve the optimization of the overall classifier. “Classifiers are methods for automating the recognition of an applied signal or data, usually one with a lot of dimensions. Classifiers are used in many applications in image processing including computer vision, medical imaging (X-ray, MRI), and face detection. Boosting works by taking many simple but weak classifiers and combining them to form an improved classifier (a composite classifier or a "strong learner"). As in Adaboost, this is accomplished by iteratively testing the weak classifiers on an example set of data. After a given classifier is tested on the data, examples that were improperly identified are given a greater weight for testing with the following classifier; as the algorithm progresses there is a greater focus on identifying examples that previous classifiers have failed on. The weighting of example data is done based on the error in identification; how "wrong" the function was. This leads to an optimized overall classifier” (Jebara, [0016]).
Regarding Claim 18:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] one or more non-transitory, computer readable mediums of claim 17, and Barbu further teaches:
wherein the instructions further cause operations comprising generating a classification for data in the first dataset based on the second feature input by: inputting the second feature input into the second weak learner of the plurality of weak learners
Barbu, [0021], “FIG. 2 describes the method of FIG. 1 in greater detail. Input 30 includes a training set S of data of N training points
X
=
{
x
1
,
x
2
…
x
N
}
. M disjoint features are available for each point
x
i
=
{
x
i
1
,
x
i
2
…
x
i
M
}
. Each member
x
i
j
in the set
x
i
is known as a view of point
x
i
. Each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example. Since M disjoint features are available for each point, there will be M training sets. The set S is represented as
S
j
=
x
1
j
,
y
1
,
x
2
j
,
y
2
…
x
N
j
,
y
N
, where
j
=
1
…
M
and
y
i
∈
+
1
,
-
1
and each
x
i
j
,
y
i
pair represents the
j
t
h
view and class label of the
i
t
h
training example.”
[0020], “In an initialization step 22, all the views for a given for a given training point are initialized with the same weight
w
1
i
=
1
/
N
. Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
In view of para. 21 above which describes the training data as input, para. 20 teaches training each classifier on each view of the training data [inputting the second feature input into a second weak learner of the plurality of weak learners].
receiving a second output from the second weak learner
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24.”
Barbu teaches obtaining a weak hypothesis from the classifiers [receiving a second output from the second weak learner].
generating a classification for data in the first dataset based on the second weight and the second output
Barbu, [0020], “Next, for a number of iterations
k
=
1
to
k
=
k
m
a
x
, the classifiers are separately trained on each view on a sample of training examples based on the shared distribution of the weights 23, weak hypotheses are obtained for each view 24…After a classifier
h
k
* with lowest error rate is selected and combination weight
a
k
* is obtained, the weights of the training data views are updated 25 forming a shared sampling distribution that will be used to sample training data at the next iteration. A final hypothesis or classifier is output 26 from the system.”
Barbu teaches after the weights and training views are updated, sample training data at the next interation [based on the second feature input], a final hypothesis is output [generating a classification for data in the first dataset]
Sabe further teaches:
wherein the second weak learner corresponds to a second class of a plurality of weak learner classes
Sabe, [0068], “first lower nodes (e.g., a node 221-1-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1-1.sub.1 to 21-1-1.sub.100) that learns learning samples with a second label (e.g., labels 1 to 5 shown in FIG. 17) of a second range based on results of classification by the upper nodes (e.g., the node 221-1 shown in FIG. 24), the second range being a part of the first range”
Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [the second weak learner corresponds to a second class of a plurality of weak learner classes].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Jebara further teaches:
comparing the second output to a second validation metric, wherein the second validation metric is based on the second class; adjusting a second weight based on comparing the second output to the second validation metric
Jebara, [0025], “The AdaBoost algorithm assigns a weight w.sub.i to each training example. In each step of the AdaBoost algorithm, a weak learner G.sup.s() is obtained on the weighted examples and a coefficient .alpha..sub.s is assigned to it. Thus, the AdaBoost algorithm iteratively builds .SIGMA..sub.s=1.sup.s.alpha..sub.sG.sup.s(). If a training example is correctly classified, its weight is exponentially decreased; if it is misclassified, its weight is exponentially increased. The process is repeated until a stopping criterion is met.”
[0015], “The general idea behind boosting is that it takes a group of simple, but poor classifiers (‘weak learners’), and combines them to form a single classifier with much higher accuracy (a ‘strong learner’). This is accomplished by iteratively testing classifiers on a set of training data, and weighting the data based on each classifier's efficacy (i.e. how much error there is in identification).”
Jebara teaches training a weak learner with weights. The weights are adjusted [adjusting a second weight] based on whether the training example was correctly classified [comparing the second output to the second validation metric, wherein the second validation metric is based on the second class]
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu and Sabe with Jebara in order to improve the optimization of the overall classifier. “Classifiers are methods for automating the recognition of an applied signal or data, usually one with a lot of dimensions. Classifiers are used in many applications in image processing including computer vision, medical imaging (X-ray, MRI), and face detection. Boosting works by taking many simple but weak classifiers and combining them to form an improved classifier (a composite classifier or a "strong learner"). As in Adaboost, this is accomplished by iteratively testing the weak classifiers on an example set of data. After a given classifier is tested on the data, examples that were improperly identified are given a greater weight for testing with the following classifier; as the algorithm progresses there is a greater focus on identifying examples that previous classifiers have failed on. The weighting of example data is done based on the error in identification; how "wrong" the function was. This leads to an optimized overall classifier” (Jebara, [0016]).
Regarding Claim 19:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] one or more non-transitory, computer readable mediums of claim 18, and Sabe further teaches:
wherein generating the classification for the data in the first dataset based on the second weight and the second output further comprises: combining the first output and the second output; generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction
Sabe, [0102], “The ensemble learning machine 6 performs ensemble learning according to a boosting algorithm using a plurality of weak classifiers as described above so that a strong classification can be obtained as a result. Although each of the weak classifiers is constructed very simply, having low ability of classifying face and non-face by itself, high ability of classification can be achieved by combining, for example, on the order of hundreds to thousands of weak classifiers…Although each of the weak classifiers has low classification ability by itself, a classifier having high classification ability can be obtained depending on selection and combination of weak classifiers. Thus, the ensemble learning machine 6 learns a combination of weak classifiers, i.e., selection of weak classifiers and weights on output values of the respective weak classifiers for calculating a value of weighted majority.”
Sabe teaches combining a selection of weak classifiers [combining he first output and the second output] to obtain a strong classification [generating an ensemble prediction based on combining the first output and the second output, wherein the classification is based on the ensemble prediction].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Claims 5-6 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Barbu in view of Sabe, further in view of Jebara as applied to claims 2 and 17 above, respectively, and further in view of Bogorad et al. (US 20240378484), hereinafter Bogorad.
Regarding Claim 5:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, and Sabe further discloses:
determining that the first weak learner has the first class of the plurality of weak learner classes
Sabe, [0068], “An information processing apparatus (e.g., an object detecting apparatus 5 shown in FIG. 5) according to the present invention includes upper nodes (e.g., a node 221-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1.sub.1 to 21-1.sub.100 shown in FIG. 24) that learns learning samples with a first label (e.g., labels 1 to 15 shown in FIG. 17) of a first range among learning samples classified with a plurality of labels (e.g., labels 1 to 15 shown in FIG. 17)”
Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [determining that the first weak learner has the first class of the plurality of weak learner classes].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Barbu in view of Sabe, and further in view of Jebara do not explicitly disclose:
wherein comparing the first output to the first validation metric further comprises:
selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class
However, in the same field, analogous art Bogorad teaches:
wherein comparing the first output to the first validation metric further comprises:
selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class
Bogorad, [0030], “Stability can correspond to how stable a machine learning model is with respect to time. Stability can evaluate if a distribution of predictions from a previous machine learning model already in production is comparable to a distribution of predictions from the ensemble model. Stability can be determined by a population stability index (PSI) or a characteristic stability index (CSI), as examples.”
[0031], “Stability metrics can be domain specific. For example, in the spam detection domain, a difference of the total number of flagged entities by a previous model in production and the retrained ensemble model should be less than or equal to a threshold percentage, such as 10%. Flagged entities can correspond to an entity having a risk score that exceeds a decision threshold. As another example in the spam detection domain, a top cost-wise number of entities, such as 20, flagged by the previous model in production and the retrained ensemble model should not differ by more than a threshold percentage, such as 10%.”
Bogorad teaches domain specific stability metrics [selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class].
Barbu, Sabe, Jebara, Bogorad and the instant application are analogous art because they are all directed to ensemble learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, and Jebara with Bogorad in order to increase the robustness of the model. “If the plurality of metrics passes the validation, then the ensemble model can be pushed to production. For classification performance, the optimal weight should lay inside a range within the search interval. For stability metrics, differences between the previous model in production and the retrained ensemble model should be less than or equal to a percentage threshold” (Bogorad, [0032]).
Regarding Claim 6:
As discussed above Barbu and Sabe in view of Jebara, further in view of Bogorad teach [the] method of claim 5, and Sabe further discloses:
wherein selecting the first validation metric from the plurality of validation metrics based on the first validation metric corresponding to the first class further comprises: determining a data characteristic of the first dataset; and filtering the plurality of validation metrics based on the data characteristic
Bogorad, [0062], “The plurality of metrics can include classification performance and stability.”
[0031], “Stability metrics can be domain specific. For example, in the spam detection domain, a difference of the total number of flagged entities by a previous model in production and the retrained ensemble model should be less than or equal to a threshold percentage, such as 10%. Flagged entities can correspond to an entity having a risk score that exceeds a decision threshold. As another example in the spam detection domain, a top cost-wise number of entities, such as 20, flagged by the previous model in production and the retrained ensemble model should not differ by more than a threshold percentage, such as 10%.”
In para. 62, Bogorad teaches a plurality of metrics including classification and stability metrics. Para. 31 specifies stability metrics can be determined for domain specific data. For example, in spam detection domain [determining a data characteristic of the first dataset], the stability metric is calculated using the total number of flagged entities [filtering the plurality of validation metrics based on the data characteristic].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, and Jebara with Bogorad in order to increase the robustness of the model. “If the plurality of metrics passes the validation, then the ensemble model can be pushed to production. For classification performance, the optimal weight should lay inside a range within the search interval. For stability metrics, differences between the previous model in production and the retrained ensemble model should be less than or equal to a percentage threshold” (Bogorad, [0032]).
Regarding Claim 20:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] one or more non-transitory, computer readable mediums of claim 17, and Sabe further teaches:
determining that the first weak learner has the first class of the plurality of weak learner classes
Sabe, [0068], “An information processing apparatus (e.g., an object detecting apparatus 5 shown in FIG. 5) according to the present invention includes upper nodes (e.g., a node 221-1 shown in FIG. 24) each including a plurality of weak classifiers (e.g., weak classifiers 21-1.sub.1 to 21-1.sub.100 shown in FIG. 24) that learns learning samples with a first label (e.g., labels 1 to 15 shown in FIG. 17) of a first range among learning samples classified with a plurality of labels (e.g., labels 1 to 15 shown in FIG. 17)”
Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [determining that the first weak learner has the first class of the plurality of weak learner classes].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu with Sabe in order to improve the accuracy of classifications and the speed of the computations. “The present invention has been made in view of the situation described above, and it is an object thereof to further increase computation speed during learning and detection by reducing the amount of computation when an object of interest is detected based on ensemble learning” and “It is expected that classification performance is improved when these estimation methods (classification methods) are used in combination during ensemble learning.” (Sabe, [0013]; [0122]).
Barbu in view of Sabe, and further in view of Jebara do not explicitly disclose:
wherein comparing the first output to the first validation metric further comprises
selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class
However, in the same field, analogous art Bogorad teaches:
wherein comparing the first output to the first validation metric further comprises
selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class
Bogorad, [0030], “Stability can correspond to how stable a machine learning model is with respect to time. Stability can evaluate if a distribution of predictions from a previous machine learning model already in production is comparable to a distribution of predictions from the ensemble model. Stability can be determined by a population stability index (PSI) or a characteristic stability index (CSI), as examples.”
[0031], “Stability metrics can be domain specific. For example, in the spam detection domain, a difference of the total number of flagged entities by a previous model in production and the retrained ensemble model should be less than or equal to a threshold percentage, such as 10%. Flagged entities can correspond to an entity having a risk score that exceeds a decision threshold. As another example in the spam detection domain, a top cost-wise number of entities, such as 20, flagged by the previous model in production and the retrained ensemble model should not differ by more than a threshold percentage, such as 10%.”
Bogorad teaches domain specific stability metrics [selecting the first validation metric from a plurality of validation metrics based on the first validation metric corresponding to the first class].
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, and Jebara with Bogorad in order to increase the robustness of the model. “If the plurality of metrics passes the validation, then the ensemble model can be pushed to production. For classification performance, the optimal weight should lay inside a range within the search interval. For stability metrics, differences between the previous model in production and the retrained ensemble model should be less than or equal to a percentage threshold” (Bogorad, [0032]).
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Barbu in view of Sabe, further in view of Jebara as applied to claim 2 above, and further in view of Alfaro Suzan et al. (US 20240394568), hereinafter Suzan.
Regarding Claim 10:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, but do not explicitly disclose:
wherein receiving the first output from the first weak learner further comprises: receiving a learning rate for a weak learner ensemble; and calculating a step size for the weak learner ensemble based on the learning rate
However, in the same field, analogous art Suzan teaches:
wherein receiving the first output from the first weak learner further comprises: receiving a learning rate for a weak learner ensemble; and calculating a step size for the weak learner ensemble based on the learning rate
Suzan [0641], “Three hyperparameters have been included in the XGBoost models: ‘learning rate’, ‘n_estimators’ and ‘maximum depth’. The learning rate regulates the size of the step in each iteration…”
Suzan teaches including hyperparameters [receiving a learning rate] in XGBoost [for a weak learner ensemble]. The learning rate is said to regulate the size of the step in each iteration [calculating a step size for the weak learner ensemble based on the learning rate].
Barbu, Sabe, Jebara, Suzan and the instant application are analogous art because they are all directed to ensemble learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, and Jebara with Suzan in order to increase the robustness of the model by controlling the models rate of learning. “Three hyperparameters have been included in the XGBoost models: ‘learning rate’, ‘n_estimators’ and ‘maximum depth’. The learning rate regulates the size of the step in each iteration, impacting the rate or speed at which the model learns or adapts during training; it controls the shrinkage of each tree's contribution. The N_estimators is a hyperparameter often associated with ensemble learning methods (Random Forests and Gradient Boosting), it determines the number of trees in the ensemble” (Suzan, [0641]).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Barbu in view of Sabe, further in view of Jebara as applied to claim 2 above, and further in view of Baransky et al. (US 20240048506), hereinafter Baransky.
Regarding Claim 13:
As discussed above Barbu in view of Sabe, further in view of Jebara teach [the] method of claim 2, and Sabe further teaches:
wherein receiving the first output from the first weak learner further comprises: wherein the second format corresponds to a second class of a plurality of weak learner classes
Sabe, Abstract, “Each of the nodes has a number of weak classifiers. Each terminal node learns face images associated with one label.”
[0183]-[0187], “In this embodiment, the learning data used during classification includes the following four types of data: (A) K pairs of two pixel positions. (B) K thresholds for weak classifiers. (C) K weights for weighted majority (K confidences for weak classifiers). (D) K termination thresholds”
Sabe teaches an ensemble of weak classifiers where each node is associated with one label [wherein the second format corresponds to a second class of a plurality of weak learner classes].
Barbu in view of Sabe, further in view of Jebara do not explicitly disclose:
determining a first format of the first output; and
converting the first format to a second format
However, in the same field, analogous art Baransky teaches:
determining a first format of the first output; and
converting the first format to a second format
Baransky, [0072], “The process may then continue to block 212, wherein the process converts, based on an output of a machine learning algorithm, the primary data format to a secondary data format.”
Barbu, Sabe, Jebara, Baransky, and the instant application are analogous art because they are all directed to machine learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, and Jebara with Baransky to have accurate data between different entities. “In particular, the system and method for autonomous conversion of a resource format using machine learning is an improvement over existing solutions regarding resource format conversion,…(ii) providing a more accurate solution to problem, thus reducing the number of resources required to remedy any errors made due to a less accurate solution, (iii) removing manual input and waste from the implementation of the solution, thus improving speed and efficiency of the process and conserving computing resources” (Baransky, [0033]).
Claims 14 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Barbu and Sabe in view of Jebara, further in view of Baransky as applied to claim 13 above, and further in view of Suh et al. (US 20210271927), hereinafter Suh.
Regarding Claim 14:
As discussed above Barbu and Sabe in view of Jebara, further in view of Baransky teach [the] method of claim 13, but do not explicitly disclose:
wherein converting the first format to the second format further comprises:
determining a first mapping for the first output based on the first class
determining a second mapping for the first output based on the second class
However, in the same field, analogous art Suh teaches:
wherein converting the first format to the second format further comprises:
determining a first mapping for the first output based on the first class
determining a second mapping for the first output based on the second class
Suh, [0028], “The apparatus for the artificial neural network may sample the first format image using a first sampling scheme to generate a first feature map, and sample the second format image using a second sampling scheme to generate a second feature map (S3000).”
Suh teaches generating a first feature map and a second feature map for a first data format and a second data format [determining a first mapping for the first output based on the first class; and determining a second mapping for the first output based on the second class].
Barbu, Sabe, Jebara, Baransky, Suh and the instant application are analogous art because they are all directed to machine learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, Jebara, and Baransky with Suh in order to increase the accuracy of the models. “Artificial neural network are modeling techniques implemented in a complex network structure to emulate a human brain. The artificial neural network is often utilized for the classification and/or clustering of data by finding and/or recognizing specific patterns in data of various types, such as a still images, video images, text, sound, etc. Various studies have been conducted to develop methods to improve the extraction of feature maps from the data and/or to increase recognition of specific patterns” (Suh, [0003]).
Regarding Claim 16:
As discussed above Barbu and Sabe in view of Jebara, further in view of Baransky teach [the] method of claim 13, but do not explicitly disclose:
wherein converting the first format to the second format further comprises:
determining a first extraction requirement for the first output based on the first class
applying the first extraction requirement to the first output
However, in the same field, analogous art Suh teaches:
wherein converting the first format to the second format further comprises:
determining a first extraction requirement for the first output based on the first class
applying the first extraction requirement to the first output
Suh, [0028], “The apparatus for the artificial neural network may sample the first format image using a first sampling scheme to generate a first feature map, and sample the second format image using a second sampling scheme to generate a second feature map (S3000).”
[0003], “Various studies have been conducted to develop methods to improve the extraction of feature maps from the data and/or to increase recognition of specific patterns.”
In para. 28, Suh teaches generating a first feature map, and para. 3 specifies extraction of feature maps [determining a first extraction requirement for the first output based on the first class; and applying the first extraction requirement to the first output].
Barbu, Sabe, Jebara, Baransky, Suh and the instant application are analogous art because they are all directed to machine learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, Jebara, and Baransky with Suh in order to increase the accuracy of the models. “Artificial neural network are modeling techniques implemented in a complex network structure to emulate a human brain. The artificial neural network is often utilized for the classification and/or clustering of data by finding and/or recognizing specific patterns in data of various types, such as a still images, video images, text, sound, etc. Various studies have been conducted to develop methods to improve the extraction of feature maps from the data and/or to increase recognition of specific patterns” (Suh, [0003]).
Claim 15 is rejected under 35 U.S.C. 103 as being unpatentable over Barbu and Sabe in view of Jebara, further in view of Baransky as applied to claim 13 above, and further in view of Suh et al. (US 20210271927), hereinafter Suh.
Regarding Claim 15:
As discussed above Barbu and Sabe in view of Jebara, further in view of Baransky teach [the] method of claim 13, but do not explicitly disclose:
wherein converting the first format to the second format further comprises: determining a first probability for the first output based on the first class; and determining a second probability for the first output based on the second class
However, in the same field, analogous art Suh teaches:
wherein converting the first format to the second format further comprises: determining a first probability for the first output based on the first class; and determining a second probability for the first output based on the second class
Ma, [42], “The first model may be a conversion prediction model p(y|x), which may predict a probability of a conversion ‘y’ given an impression ‘x.’”
Ma teaches using a conversion prediction model to predict a probability y given x, [determining a first probability for the first output based on the first class; and determining a second probability for the first output based on the second class].
Barbu, Sabe, Jebara, Baransky, Suh and the instant application are analogous art because they are all directed to machine learning.
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify Barbu, Sabe, Jebara, and Baransky with Ma in order to increase the accuracy of the models. “Given these challenges, the systems and methods described herein may involve the following advantages. An impression-level model may be built to predict conversions of an impression followed by a report-level calibration to achieve significant improvement at report-level with the consideration of the data sparsity. A two-stage approach may be used. The first stage may involve predicting whether an impression will have a conversion or not (e.g., binary classification problem), followed by predicting other outcomes conditioned on conversion (e.g., regression problem). This may solve the zero-inflated regression problem. A train/test protocol may be used to perform an effective trade-off between waiting for delayed conversions and producing reports in a timely manner. Any models may continuously be trained with new data and to update predictions to improve report accuracy.” (Ma, [44]).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to STEVEN PHUNG whose telephone number is (703) 756-1499. The examiner can normally be reached Monday-Thursday: 9:00AM-4:00PM ET.
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, KAMRAN AFSHAR can be reached at (571) 272-7796. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/STEVEN PHUNG/Examiner, Art Unit 2125
/KAMRAN AFSHAR/Supervisory Patent Examiner, Art Unit 2125
Prosecution Insights