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 .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 06/11/2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Specification
The disclosure is objected to because of the following informalities:
Paragraphs [0028,0030,0034,0038,0057,0065,0068,0075,0076,0085,0098,0106] recites “IND” however, it’s not clear what IND stands for. Does IND stand for “in-distribution data” or “in-domain data” ?.
Appropriate correction is required.
Claim Objections
Claim 10 is objected to because of the following informalities:
Claim 10 line 3 recites “…the processor being configured to to perform…” should recite “…the processor being configured to perform…”.
Appropriate correction is required.
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 4-6, 10-12, and 16-18 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.
Independent claim 4 recites “generating training data that includes pseudo data and pseudo label data for the pseudo data; calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator;” however, it’s not clear how the “pseudo data” is generated. Is the pseudo data generated by the generator? Is the “pseudo data” generated by another machine learning model and then used as input for the generator? Further, if the “pseudo data” is generated by the generator, how can the “pseudo data” being used as input for the generator to calculate a weight gradient of the generator to reduce a loss of an output? Will instead the “pseudo data” generated by the generator used as input to the classifier for calculating a weight gradient of the generator to reduce a loss of an output obtained by the classifier?. For purpose of examination examiner is viewing the limitations as follow: the generator generates “training data that includes pseudo data and pseudo label data for the pseudo data” and a weight gradient of the generator is calculated to reduce a loss of an output obtained by inputting the pseudo data to the classifier.
Independent claims 10 and 16 recites similar limitations as those in independent claim 4 and thus are rejected for reasons set forth in the rejection of claim 4.
Claims 5-6, 11-12, and 17-18 are dependent on claim 4, 10 and 16, and thus are rejected for reasons set forth in the rejection of claim 4,10 and 16.
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
Claims 1-6 and 19 are a method type claim. Claims 7-12 and 20 are information processing apparatus type claims. Claims 13-18 are a non-transitory computer-readable recording medium type claim. Therefore, claims 1-20 are directed to either a process, machine, manufacture or composition of matter.
Regarding claim 1: 2A Prong 1:
generating pseudo data and pseudo label data for the pseudo data; and (mental process – of generating pseudo data and pseudo label data for the pseudo data can be performed by the human mind with the help of pen and paper. For example, a human can generate mock data and tag the mock data (e.g., evaluation & judgement )).
updating the pseudo data in a direction for reducing a loss of an output obtained (mathematical concept – of updating the pseudo data in a direction for reducing a loss of an output obtained to generate out-of-distribution data not included in a specific domain (e.g., mathematical formulas and mathematical calculations). For example, Paragraph [0062] of the instant application discloses the OOD candidate data (OOD data, or pseudo data) and [0064] teaches updates the OOD data candidates (i.e. pseudo data) using the learning rate
I
r
and the
PNG
media_image1.png
50
334
media_image1.png
Greyscale
. In addition, paragraph [0032] of the instant application teaches OOD data is generated by the following mathematical formula
PNG
media_image2.png
46
314
media_image2.png
Greyscale
).
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
A data generation method implemented by a computer, the data generation method comprising: (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
…inputting the pseudo data to a machine learning model,… (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination of generic computer functions being implemented with generic computer elements in a high level of generality to perform the disclosed abstract idea above.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
A data generation method implemented by a computer, the data generation method comprising: (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
…inputting the pseudo data to a machine learning model,… (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are mere insignificant extra solution activity in combination of generic computer functions being implemented with generic computer elements in a high level of generality to perform the disclosed abstract idea above.
Regarding claim 2: Depends on claim 1, thus the rejection of claim 1 is incorporated.2A Prong 1:
updating a classifier included in the machine learning model, using the updated pseudo data (mathematical concept – of updating a classifier included in the machine learning model, using the updated pseudo data (e.g., mathematical calculations). See for example, algorithm in Fig. 4 element P8 and [0067] of the instant application).
2A Prong 2 and 2B: None.
Regarding claim 3: Depends on claim 2, thus the rejection of claim 2 is incorporated.2A Prong 1:
repeating, a plurality of times, a process of updating the pseudo data and a process of updating the classifier (mathematical concept – of repeating, a plurality of times, a process of updating the pseudo data and a process of updating the classifier (e.g., mathematical calculations). See for example, algorithm in Fig. 4 element P10 and [0069] of the instant application).
2A Prong 2 and 2B: None.
Regarding claim 4: 2A Prong 1:
generating training data that includes pseudo data and pseudo label data for the pseudo data; (mental process – of generating pseudo data and pseudo label data for the pseudo data can be performed by the human mind with the help of pen and paper. For example, a human can generate mock data and tag the mock data (e.g., evaluation & judgement )).
calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator; and (mathematical concept – calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator (e.g., mathematical calculations). See Fig. 4 element P5 and paragraph [0063] of the instant application)
updating a parameter of the generator, based on the calculated weight gradient (mathematical concept – of updating a parameter of the generator, based on the calculated weight gradient (e.g., mathematical calculations). Paragraph [0107] of the instant application teaches using gradient descent method to update the parameter of the generator. A person skilled in the relevant art will recognize gradient descent is a mathematical optimization algorithm).
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
A machine learning method implemented by a computer, the machine learning method comprising: (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
when training a machine learning model that includes a generator that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
The additional elements as disclosed above alone or in combination do not integrate the judicial exception into practical application as they are mere insignificant extra solution activity in combination of generic computer functions being implemented with generic computer elements in a high level of generality to perform the disclosed abstract idea above.
2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
A machine learning method implemented by a computer, the machine learning method comprising: (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
when training a machine learning model that includes a generator that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
The additional elements as disclosed above in combination of the abstract idea are not sufficient to amount to significantly more than the judicial exception as they are mere insignificant extra solution activity in combination of generic computer functions being implemented with generic computer elements in a high level of generality to perform the disclosed abstract idea above.
Regarding claim 5: Depends on claim 4, thus the rejection of claim 4 is incorporated.2A Prong 1: None.
2A Prong 2 and 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
the machine learning method further comprising: updating the classifier included in the machine learning model, using the first data generated by inputting the pseudo data to the generator (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
Regarding claim 6: Depends on claim 5, thus the rejection of claim 5 is incorporated.2A Prong 1:
…repeatedly performs, a plurality of times, a process of updating the pseudo data and a process of updating the classifier (mathematical concept – of repeating, a plurality of times, a process of updating the pseudo data and a process of updating the classifier (e.g., mathematical calculations). See for example, algorithm in Fig. 4 element P10 and [0069] of the instant application ).
2A Prong 2 and 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
wherein the computer… (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
Regarding claim 7: it is rejected under the same rationality as claim 1. Claim 7 only recites the additional elements of An information processing apparatus comprising: a memory; and a processor coupled to the memory, the processor being configured to perform processing comprising:… which is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f).
Regarding claim 8: See rejection of claim 2, same rational applies.
Regarding claim 9: See rejection of claim 3, same rational applies.
Regarding claim 10: it is rejected under the same rationality as claim 4. Claim 10 only recites the additional elements of An information processing apparatus comprising: a memory; and a processor coupled to the memory, the processor being configured to to perform processing comprising:… which is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f).
Regarding claim 11: See rejection of claim 5, same rational applies.
Regarding claim 12: See rejection of claim 3, same rational applies.
Regarding claim 13: it is rejected under the same rationality as claim 1. Claim 13 only recites the additional elements of An information processing apparatus comprising: a memory; and a processor coupled to the memory, the processor being configured to to perform processing comprising:… which is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f).
Regarding claim 14: See rejection of claim 2, same rational applies.
Regarding claim 15: See rejection of claim 6, same rational applies.
Regarding claim 16: it is rejected under the same rationality as claim 4. Claim 16 only recites the additional elements of A non-transitory computer-readable recording medium storing a machine learning program for causing a computer to perform processing comprising:… which is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f).
Regarding claim 17: See rejection of claim 5, same rational applies.
Regarding claim 18: See rejection of claim 6, same rational applies.
Regarding claim 19: Depends on claim 1, thus the rejection of claim 1 is incorporated.2A Prong 1:
generating training data that includes the updated pseudo data and the pseudo label data for the pseudo data; and (mental process – of generating pseudo data and pseudo label data for the pseudo data can be performed by the human mind with the help of pen and paper. For example, a human can generate mock data and tag the mock data (e.g., evaluation & judgement )).
calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator; and (mathematical concept – calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator (e.g., mathematical calculations). See Fig. 4 element P5 and paragraph [0063] of the instant application)
updating a parameter of the generator, based on the calculated weight gradient (mathematical concept – of updating a parameter of the generator, based on the calculated weight gradient (e.g., mathematical calculations). Paragraph [0107] of the instant application teaches using gradient descent method to update the parameter of the generator. A person skilled in the relevant art will recognize gradient descent is a mathematical optimization algorithm).
2A Prong 2 and 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
when training a machine learning model that includes a generator that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
training the machine learning model by using the generated training data, the training of the machine learning model including: (This is directed to using computers or other machinery merely as a tool to perform an existing process. See MPEP 2106.05(f)).
Regarding claim 20: See rejection of claim 19, same rational applies.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
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 1-3, 7-9, and 13-14 are rejected under 35 U.S.C. 103 as being unpatentable over Moller et al. Out-of-distribution Detection and Generation using Soft Brownian Offset Sampling and Autoencoders (hereinafter Moller) and in view of Tan et al. US 2018 / 0240011 A1 (hereinafter Tan).
Regarding claim 1:
Moller teaches: A data generation method implemented by a computer, the data generation method comprising: (Moller Fig. 1 teaches a data generation method and pg. 5 sec: Experiment teaches performing experimental evaluation with a case study on the synthetic generation of OOD samples for cyclist trajectories” thus, requiring a computer to implement the method).
generating pseudo data and pseudo label data for the pseudo data; and ( Moller pg. 5, left col. sec: 4.1. Time series, para. 1 teaches generating ID dataset denoted as
X
i
d
(i.e., pseudo data) and pg. 5, right col. Para. 3 teach all of these datasets have labels (i.e., pseudo label data).
PNG
media_image3.png
242
344
media_image3.png
Greyscale
In particular, Fig. 1 above in depicts the “ID data”, that is the pseudo data from the synthetic ID baseline dataset).
inputting the pseudo data to a machine learning model, to generate out-of-distribution data not included in a specific domain (Moller Fig. 1 and pg. 5, left col. sec: 4.1. Time series, para. 2 teach the ID baseline dataset
X
i
d
(i.e., pseudo data) is used to train a variational autoencoder (i.e., machine learning model) to generate out-of-distribution data (OOD Data)).
Moller does not teach updating the pseudo data in a direction for reducing a loss of an output obtained by inputting the pseudo data to a machine learning model,…
Nonetheless, Tan teaches the following:
updating the pseudo data in a direction for reducing a loss of an output obtained by inputting the pseudo data to a machine learning model,… ( Tan [0072] lines 6-7 teaches generating random data sample (i.e., pseudo data) and random label (i.e., pseudo label data) and lines 12-24 teaches the random data sample/input data is repeatedly updated via gradient descent for “several iterations”).
Tan is also in the same field of endeavor as Moller (machine learning). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to include the functionality of updating the random data sample in a direction for reducing a loss of an output obtained by inputting the pseudo data to a machine learning model as being disclosed and taught by Tan in the system taught by Moller to yield the predictable results of facilitate the data sampling and improve the machine learning model (see Tan [0050]).
Regarding claim 2:
Moller and Tan teach The data generation method according to claim 1. Moller teaches the data generation method further comprising: updating a classifier included in the machine learning model, using the pseudo data (Moller Fig. 1 teaches a OOD Detector (i.e., classifier) is trained with the ID baseline dataset
X
i
d
(i.e., pseudo data). To be specific, Moller pg. 5, right col., para. 6 teaches the OOD Detector (i.e., classifier) is trained (i.e., updated) in a “supervised fashion”).
Moller does not explicitly teach or suggest the pseudo data is updated. Nonetheless, Tan [0072] lines 6-7 teaches generating random data sample (i.e., pseudo data) and lines 12-24 teaches the random data sample/input data is repeatedly updated via gradient descent for “several iterations”.
Regarding claim 3:
Moller and Tan teach The data generation method according to claim 2, the data generation method further comprising: Moller specially teaches repeating, a plurality of times, a process of… updating the classifier (Moller pg. 5, right col., para. 6 teaches the OOD Detector (i.e., classifier) is trained (i.e., updated) in a “supervised fashion”) and Tan teaches repeating, a plurality of times, a process of updating the pseudo data…. (Tan [0072] lines 6-7 teaches generating random data sample (i.e., pseudo data) and lines 12-24 teaches the random data sample/input data is repeatedly updated via gradient descent for “several iterations”.
Regarding claim 7: it is rejected under the same rationality as claim 1. Claim 7 only recites the additional elements of An information processing apparatus comprising: a memory; and a processor coupled to the memory, the processor being configured to perform processing comprising:.., for which Tan Fig. 13 teaches a computer device (i.e., information processing apparatus) – element 1380 with a processor - element 1382 and a memory - element 1383.
Regarding claim 8: it is an information processing apparatus claim comprising limitations similar to claim 2 , therefore it is rejected under the same rationality as claim 2.
Regarding claim 9: it is an information processing apparatus claim comprising limitations similar to claim 3 , therefore it is rejected under the same rationality as claim 3.
Regarding claim 13: it is rejected under the same rationality as claim 1. Claim 13 only recites the additional elements of A non-transitory computer-readable recording medium storing a machine learning program for causing a computer to perform processing comprising:…, for which Tan [0055] teaches “computer readable storage media ( e.g., a memory device ) encoded with software comprising computer executable instructions and when the software is executed (by the controller) it is operable to perform the operations described herein”.
Regarding claim 14: it is a non-transitory computer-readable recording medium claim comprising limitations similar to claim 2 , therefore it is rejected under the same rationality as claim 14.
Claims 4-5, 10-11 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over of Lee et al. Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples (hereinafter Lee) in view of Moller.
Regarding claim 4:
Lee teaches: A machine learning method implemented by a computer, the machine learning method comprising: ( Lee pg. 4, sec: 2.3 JOINT TRAINING METHOD OF CONFIDENT CLASSIFIER AND ADVERSARIAL GENERATOR, para. 2 “propose an alternating algorithm, which optimizes model parameters of classifier and GAN models alternatively as shown in Algorithm 1”).
when training a machine learning model that includes a generator that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, (Lee Algorithm 1 teaches training a generator
G
that generates output
G
(
z
i
)
which is then feed into a classifier
θ
such that the classifier determines if input
x
(i.e., the output of the generator) is from
P
i
n
o
r
P
o
u
t
” (see pg. 1 sec: Introduction, para. 2). See also, annotated algorithm 1 lines 5 and 12 and below:
PNG
media_image4.png
564
761
media_image4.png
Greyscale
).
generating training data that includes pseudo data… (Lee Algorithm 1 line 5 teaches training a generator
G
that generates output
G
z
i
, that is “pseudo data”) .
calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator; and (Lee pg. 4, para. 2 teaches the generator takes as input “variable z from a prior distribution” to generate output
G
(
z
)
(pseudo data) and Algorithm 1 line 6-8 teaches “update the generator
G
by descending its stochastic gradient” . A person skilled in the relevant art will recognize descending a stochastic gradient calculates a weight gradient for a model, thus this suggest a weight gradient of the generator is calculated to reduce a loss of an output obtained by inputting the pseudo data into the classifier
P
θ
y
G
z
i
as shown in line 8).
updating a parameter of the generator, based on the calculated weight gradient (Lee Algorithm 1 teaches “update the generator
G
by descending its stochastic gradient”. A person skilled in the relevant art will recognize “stochastic gradient descent” updated a models parameter by moving in the opposite direction of the calculated weight gradient, thus Algorithm 1 teaches updating a parameter of the generator, based on the calculated weight gradient).
Lee dos does not teach generating training data that includes pseudo data and pseudo label data for the pseudo data; calculating a weight gradient of the machine learning model to reduce a loss of an output obtained by inputting the pseudo data to the machine learning model.
Nonetheless, Moller teaches the following:
when training a machine learning model that that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, (Moller Fig. 1 teaches training an autoencoder (i.e., generator) that generated OOD Data (i.e., first data) to be input into a OOD Detector (i.e., classifier) that enables to detect OOD samples such as performs out of distribution determinations as to whether OOD data is included in a specific domain, e.g., ID data or OOD data (see pg. 6, right col., sec: 4.2 Image Data, para. 3)).
generating training data that includes pseudo data and pseudo label data for the pseudo data; ( Moller pg. 5, left col. sec: 4.1. Time series, para. 1 teaches generating ID training dataset (i.e., training data) that “comprises 2000 time series sampled from this model (cf. Fig. 4) and is denoted as
X
i
d
” (i.e., pseudo data) and pg. 5, right col. Para. 3 teach all of these datasets have labels (i.e., pseudo label data).
calculating a weight gradient of the [machine learning model] to reduce a loss of an output obtained by inputting the pseudo data to the [machine learning model]; and ( Moller pg. 5, left col. sec: 4.1. Time series, para. 2 continuing on left col. Para. 1 teach “Having created an ID dataset
X
i
d
is then used to train a VAE matching the identity while minimizing reconstruction loss”. Moreover, Moller pg. 7, left col., sec: 4.3. Cyclist Trajectories, para. 1 last lines 28-29 teaches VAE (machine learning model) is trained using Adam optimizer. Thus, this suggest the VAE (machine learning model) is trained using gradient based learning methods, which involves calculating the gradient of the loss function with respect to the weights of the model).
Moller is also in the same field of endeavor as Lee (machine learning). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to include the functionality of generating training data that includes pseudo data and pseudo label data for the pseudo data and calculating a weight gradient of the machine learning model to reduce a loss of an output obtained by inputting the pseudo data to the machine learning model as being disclosed and taught by Moller, in the system taught by Lee to yield the predictable results of improve training of a model and “increase its performance” ( see Moller Abstract).
Regarding claim 5:
Lee and Moler teach The machine learning method according to claim 4. Specifically Lee teaches the machine learning method further comprising: updating the classifier included in the machine learning model, using the first data generated by inputting the pseudo data to the generator (
PNG
media_image5.png
530
835
media_image5.png
Greyscale
Lee teaches the classifier is updated by descending its stochastic gradient as shown in line 11-12. In particular, line 12 teaches
P
θ
y
G
z
i
, that is the classifier
θ
is updated using the generated first data
G
z
i
that was obtained by inputting samples from prior distribution (i.e., pseudo data) into the generator. See annotated Algorithm 1 above).
Regarding claim 10: it is an information processing apparatus claim comprising limitations similar to claim 4 , therefore it is rejected under the same rationality as claim 4.
Regarding claim 11: it is an information processing apparatus claim comprising limitations similar to claim 5 , therefore it is rejected under the same rationality as claim 5.
Regarding claim 16: it is a non-transitory computer-readable recording medium claim comprising limitations similar to claim 4 , therefore it is rejected under the same rationality as claim 4.
Regarding claim 17: it is a non-transitory computer-readable recording medium claim comprising limitations similar to claim 5 , therefore it is rejected under the same rationality as claim 5.
Claims 6,12, 15 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Lee, Moller in further view of Tan.
Regarding claim 6:
Lee and Moller teach The machine learning method according to claim 5. Specifically Lee teaches wherein the computer repeatedly performs, a plurality of times, a process of updating a process of updating the classifier( Lee Algorithm 1 line 11 teaches updating the classifier and line 1 and 13 teaches the processes repeats until convergence is achieved. Thus, the classifier is repeatedly updated a plurality of times).
Lee and Moller does not teach or suggest …repeatedly performs, a plurality of times, a process of updating the pseudo data…
Nonetheless, Tan teaches the following:
…repeatedly performs, a plurality of times, a process of updating the pseudo data… ( Tan [0072] lines 6-7 teaches generating random data sample (i.e., pseudo data) and random label (i.e., pseudo label data) and lines 12-24 teaches the random data sample/input data is repeatedly updated via gradient descent for “several iterations”).
Tan is also in the same field of endeavor as Lee and Moller (machine learning). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to include the functionality of updating the random data sample in a direction for reducing a loss of an output obtained by inputting the pseudo data to a machine learning model as being disclosed and taught by Tan in the system taught by Moller to yield the predictable results of facilitate the data sampling and improve the machine learning model (see Tan [0050]).
Regarding claim 12:
Lee and Moller teach The information processing apparatus according to claim 11, Moller specially teaches the processing further comprising: …repeatedly perform, a plurality of times, a process of… updating the classifier (Moller pg. 5, right col., para. 6 teaches the OOD Detector (i.e., classifier) is trained (i.e., updated) in a “supervised fashion”) and Tan teaches causing the processor to repeatedly perform, a plurality of times, a process causing the processor to repeatedly perform, a plurality of times, a process of updating the pseudo data…. (Tan Fig. 1 element 1382 teaches a processor and [0072] lines 6-7 teaches generating random data sample (i.e., pseudo data) and lines 12-24 teaches the random data sample/input data is repeatedly updated via gradient descent for “several iterations”).
Regarding claim 15: it is an non-transitory computer-readable recording medium claim comprising limitations similar to claim 4 , therefore it is rejected under the same rationality as claim 4.
Regarding claim 18: it is a non-transitory computer-readable recording medium claim comprising limitations similar to claim 4 , therefore it is rejected under the same rationality as claim 4.
Claims 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Moller, Tan in further view of Lee.
Regarding claim 19:
Moller and Tan teach The data generation method according to claim 1. Moller specifically teaches the data generation method further comprising:
when training a machine learning model that that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, (Moller Fig. 1 teaches training an autoencoder (i.e., generator) that generated OOD Data (i.e., first data) to be input into a OOD Detector (i.e., classifier) that enables to detect OOD samples such as performs out of distribution determinations as to whether OOD data is included in a specific domain, e.g., ID data or OOD data (see pg. 6, right col., sec: 4.2 Image Data, para. 3)).
generating training data that includes the pseudo data and the pseudo label data for the pseudo data; and ( Moller pg. 5, left col. sec: 4.1. Time series, para. 1 teaches generating ID training dataset (i.e., training data) that “comprises 2000 time series sampled from this model (cf. Fig. 4) and is denoted as
X
i
d
” (i.e., pseudo data) and pg. 5, right col. Para. 3 teach all of these datasets have labels (i.e., pseudo label data).
training the machine learning model by using the generated training data, the training of the machine learning model including: (Moller Fig. 1 teaches the synthetic ID baseline dataset (i.e., training data) is used to train an autoencoder (i.e., machine learning model).
calculating a weight gradient of the [machine learning model] to reduce a loss of an output obtained by inputting the pseudo data to the [machine learning model]; and ( Moller pg. 5, left col. sec: 4.1. Time series, para. 2 continuing on left col. Para. 1 teach “Having created an ID dataset
X
i
d
is then used to train a VAE matching the identity while minimizing reconstruction loss”. Moreover, Moller pg. 7, left col., sec: 4.3. Cyclist Trajectories, para. 1 last lines 28-29 teaches VAE (machine learning model) is trained using Adam optimizer. Thus, this suggest the VAE (machine learning model) is trained using gradient based learning methods, which involves calculating the gradient of the loss function with respect to the weights of the model).
While Moller does not explicitly disclose the updated pseudo data. Nonetheless, Tan teaches:
…updated pseudo data… ( Tan [0072] lines 6-7 teaches generating random data sample (i.e., pseudo data) and random label (i.e., pseudo label data) and lines 12-24 teaches the random data sample/input data is repeatedly updated via gradient descent for “several iterations”).
Neither Moller or Tan teach or suggest when training a machine learning model that includes a generator that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator; and calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator.
However, Lee teaches the following:
when training a machine learning model that includes a generator that generates first data to be input to a classifier that performs out-of-distribution determination as to whether input data is included in a specific domain, (Lee Algorithm 1 teaches training a generator
G
that generates output
G
(
z
i
)
which is then feed into a classifier
θ
such that the classifier determines if input
x
(i.e., the output of the generator) is from
P
i
n
o
r
P
o
u
t
” (see pg. 1 sec: Introduction, para. 2). See also, annotated algorithm 1 lines 5 and 12 and below:
PNG
media_image4.png
564
761
media_image4.png
Greyscale
).
generating training data that includes pseudo data… (Lee Algorithm 1 line 5 teaches training a generator
G
that generates output
G
z
i
, that is “pseudo data”) .
calculating a weight gradient of the generator to reduce a loss of an output obtained by inputting the pseudo data to the generator; and (Lee pg. 4, para. 2 teaches the generator takes as input “variable z from a prior distribution” to generate output
G
(
z
)
(pseudo data) and Algorithm 1 line 6-8 teaches “update the generator
G
by descending its stochastic gradient” . A person skilled in the relevant art will recognize descending a stochastic gradient calculates a weight gradient for a model, thus this suggest a weight gradient of the generator is calculated to reduce a loss of an output obtained by inputting the pseudo data into the classifier
P
θ
y
G
z
i
as shown in line 8).
updating a parameter of the generator, based on the calculated weight gradient (Lee Algorithm 1 teaches “update the generator
G
by descending its stochastic gradient”. A person skilled in the relevant art will recognize “stochastic gradient descent” updated a models parameter by moving in the opposite direction of the calculated weight gradient, thus Algorithm 1 teaches updating a parameter of the generator, based on the calculated weight gradient).
Lee is also in the same field of endeavor as Moller and Tan (machine learning). Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to include the functionality of a generator, calculating a weight gradient of the generator and updating the parameter of the generator as being disclosed and taught by Lee, in the system taught by Moller and Tan to yield the predictable results of better detecting out-of-distribution without losing original classification accuracy in order to train a more confidence classifier ( see Lee pg. 2, sec: contribution lines 1-2 and 14-15).
Regarding claim 20: it is an information processing apparatus claim comprising limitations similar to those of claim 19 , therefore it is rejected under the same rational of claim 19.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to GISEL G FACCENDA whose telephone number is (703)756-1919. The examiner can normally be reached Monday - Friday 8:00 am - 4:00 pm.
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, Abdullah Al Kawsar can be reached at (571) 270-3169. 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.
/G.G.F./Examiner, Art Unit 2127
/ABDULLAH AL KAWSAR/Supervisory Patent Examiner, Art Unit 2127
Prosecution Insights