DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Arguments
101 Rejection Arguments
Applicant asserts:
Applicant argues, on page 7-8, that “The resulting claim as a whole is directed to a concrete synthetic-data-generation process, not to a mathematical relationship divorced from a practical technological implementation.”
Examiner response:
Examiner respectfully disagrees. The receiving step is interpreted as extra solution activity. In addition, the use of a processor, copula model, and training of the copula model is interpreted as merely applying the abstract idea on a generic computer. The claims do not describe the separation, applying a probability integral transform and distributional transform, identifying a first and second correlated uniform marginal, and applying a continuous inverse transform and discrete inverse transform in such a way that would differentiate the process from being performed a person could perform the steps. As stated in the MPEP 2106.04(a)(2)(III)(C) “A Claim That Requires a Computer May Still Recite a Mental Process.” Therefore, the argument is not persuasive.
Applicant asserts:
Applicant argues, on page 8-11, that “amended claim 1 integrates that concept into a practical application.”
Examiner response:
Examiner respectfully disagrees. The receiving step is interpreted as extra solution activity. In addition, the use of a processor, copula model, and training of the copula model is interpreted as merely applying the abstract idea on a generic computer. The claims do not describe the separation, applying a probability integral transform and distributional transform, identifying a first and second correlated uniform marginal, and applying a continuous inverse transform and discrete inverse transform in such a way that would differentiate the process from being performed a person could perform the steps. As stated in the MPEP 2106.04(d) “The courts have also identified limitations that did not integrate a judicial exception into a practical application… Merely reciting the words "apply it"… Adding insignificant extra-solution activity to the judicial exception” Therefore, the argument is not persuasive.
Applicant asserts:
Applicant argues, on page 11-14, that “amended claim 1 integrates that concept into a practical application.”
Examiner response:
Examiner respectfully disagrees. The receiving step is interpreted as extra solution activity. In addition, the use of a processor, copula model, and training of the copula model is interpreted as merely applying the abstract idea on a generic computer. The claims do not describe the separation, applying a probability integral transform and distributional transform, identifying a first and second correlated uniform marginal, and applying a continuous inverse transform and discrete inverse transform in such a way that would differentiate the process from being performed a person could perform the steps. As stated in the MPEP 2106.05 “Limitations that the courts have found not to be enough to qualify as "significantly more" when recited in a claim with a judicial exception Adding the words "apply it"… Adding insignificant extra-solution activity to the judicial exception” Therefore, the argument is not persuasive.
Examiner further notes that the amendments to the claims changes the scope of the invention and therefore is applied with prior art rejection.
Claim Rejections - 35 USC § 112b
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.
Claim 6 and 10 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.
Claim 6 recites the limitation "The first independent uniform marginal produced from the continuous true data" in line 11. There is insufficient antecedent basis for this limitation in the claim.
Claim 6 recites the limitation "the second independent uniform marginal produced from the discrete true data;" in line 12. There is insufficient antecedent basis for this limitation in the claim.
Claim 6 recites the limitation "the second correlated uniform marginal” in line 17. There is insufficient antecedent basis for this limitation in the claim.
Claim 6 recites the limitation " the synthetic continuous data” in line 21. There is insufficient antecedent basis for this limitation in the claim.
Claim 10 recites the limitation "The first independent uniform marginal produced from the continuous true data” in line 10. There is insufficient antecedent basis for this limitation in the claim.
Claim 10 recites the limitation "the second independent uniform marginal produced from the discrete true data;" in line 11. There is insufficient antecedent basis for this limitation in the claim.
Claim 10 recites the limitation " the synthetic discrete data. " in line 21. There is insufficient antecedent basis for this limitation in the claim.
In reference to dependent claims, dependent claims do not cure the deficiencies noted in the rejection of independent claim 6 and 10. Therefore, these claims are rejected under the same rationale as claim 6 and 10.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1,3-6,8-10 and 12-14 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
In reference to claim 1:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
“separating, by the processor, the true data into continuous data and discrete data;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could evaluate the data and separate it into continuous and discrete data.
“applying, by the processor, to the continuous true data, a probability integral transform to produce a first independent uniform marginal;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply a probability transform to continuous data to produce a first independent uniform marginal.
“applying, by the processor, to the discrete true data, a distributional transform to produce a second independent uniform marginal;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply a distributional transform to discrete data to produce a second independent uniform marginal.
“identifying, by the copula model, a first correlated uniform marginal based on the transformed continuous true data;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could identify a first correlated uniform marginal based evaluation of the transformed continuous true data.
“identifying, by the copula model, a second correlated uniform marginal based on the transformed discrete true data;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could identify a second correlated uniform marginal based evaluation of the transformed discrete true data.
“applying, by the copula model, continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data.
“applying, by the copula model, discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data.
“and generating, by the copula model, the synthetic data as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data.” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could combine the synthetic continuous and discrete data into a synthetic dataset.
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“A method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, a-true data on which to model synthetic data, wherein the true data comprises continuous true data and discrete true data;” (insignificant extra-solution activity mere data gathering MPEP 2106.05(g))
“by the processor” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training, by the processor, a copula model based on normalizing flows, to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“by the copula model” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“A method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, a-true data on which to model synthetic data, wherein the true data comprises continuous true data and discrete true data;” (well-understood, routine, conventional MPEP 2106.05(d))
“by the processor” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training, by the processor, a copula model based on normalizing flows, to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“by the copula model” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 3:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
No
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“The method of claim 1, further comprising: using an autoregressive density network with the normalizing flow.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“The method of claim 1, further comprising: using an autoregressive density network with the normalizing flow.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 4:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
No
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“The method of claim 3, wherein the autoregressive density network comprises a masked autoregressive network.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“The method of claim 3, wherein the autoregressive density network comprises a masked autoregressive network.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 5:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
“The method of claim 1, further comprising: learning, by an autoregressive density network, the normalizing flows based on a functional relationship within the continuous true data to map discrete variables to corresponding random variables.” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)).
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
No
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
No
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 6:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
“applying a distributional transform to the discrete true data to produce an independent uniform marginal;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply a distributional transform to discrete data to produce a second independent uniform marginal.
“identifying, by the copula model, a first correlated uniform marginal” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could identify a first correlated uniform marginal based evaluation of the transformed continuous true data.
“applying, by the copula model, discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data.
“and generating, by the copula model, the synthetic data as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data.” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could combine the synthetic continuous and discrete data into a synthetic dataset.
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“A method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, true data on which to model synthetic data, wherein the true data comprises discrete true data including categorical discrete data and ordinal discrete data;” (insignificant extra-solution activity mere data gathering MPEP 2106.05(g))
“by the processor” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training, by the processor, a copula model based on normalizing flows, to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“by the copula model” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“A method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, true data on which to model synthetic data, wherein the true data comprises discrete true data including categorical discrete data and ordinal discrete data;” (well-understood, routine, conventional MPEP 2106.05(d))
“by the processor” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training, by the processor, a copula model based on normalizing flows, to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“by the copula model” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 8:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
No
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“The method of claim 6, further comprising: using an autoregressive density network with the normalizing flow.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“The method of claim 6, further comprising: using an autoregressive density network with the normalizing flow.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 9:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
No
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“The method of claim 6, wherein the autoregressive density network comprises a masked autoregressive network.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“The method of claim 6, wherein the autoregressive density network comprises a masked autoregressive network.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 10:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
“applying a probability integral transform to the continuous true data to produce an independent uniform marginal;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply a probability transform to continuous data to produce a first independent uniform marginal.
“identifying, by the copula model, a first correlated uniform marginal;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could identify a first correlated uniform marginal based evaluation of the transformed continuous true data.
“applying, by the copula model, continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data;” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could apply continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data.
“and generating, by the copula model, the synthetic data as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data.” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)). For example, a person could combine the synthetic continuous and discrete data into a synthetic dataset.
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“A computer-implemented method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, a-true data on which to model synthetic data, wherein the true data comprises continuous true data and discrete true data;” (insignificant extra-solution activity mere data gathering MPEP 2106.05(g))
“by the processor” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training, by the processor, a copula model based on normalizing flows, to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“by the copula model” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“A computer-implemented method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, a-true data on which to model synthetic data, wherein the true data comprises continuous true data and discrete true data;” (well-understood, routine, conventional MPEP 2106.05(d))
“by the processor” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“training, by the processor, a copula model based on normalizing flows, to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
“by the copula model” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 12:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
No
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“The method of claim 10, further comprising: using an autoregressive density network with the normalizing flow.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“The method of claim 10, further comprising: using an autoregressive density network with the normalizing flow.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 13:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
No
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
“The method of claim 12, wherein the autoregressive density network comprises a masked autoregressive network.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
“The method of claim 12, wherein the autoregressive density network comprises a masked autoregressive network.” is merely reciting the words "apply it" (or an equivalent) with the judicial exception, or merely including instructions to implement an abstract idea on a computer, or merely using a computer as a tool to perform an abstract idea (MPEP 2106.05(f)).
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception.
In reference to claim 14:
Step 1 - Is the claim to a process, machine, manufacture or composition of matter?
Yes, the claim is directed to a process
Step 2A Prong 1 - Does the claim recite an abstract idea, law of nature, or natural phenomenon?
“The method of claim 10, further comprising: learning, by an autoregressive density network, the normalizing flows based on a functional relationship within the continuous true data to map discrete variables to corresponding random variables.” which is an abstract idea because it is directed to a mental process, an observation, evaluation, judgement, or opinion. The limitation as drafted, and under a broadest reasonable interpretation, can be performed in the human mind, or by a human using a pen and paper (MPEP 2106.04(a)(2)(Ill)(c)).
Step 2A Prong 2 - Does the claim recite additional elements that integrate the judicial exception into a practical application?
No
The claim does not include additional elements that are integrated into a practical application.
Step 2B - Does the claim recite additional elements that amount to significantly more than the judicial exception?
No
The claim does not include additional elements that are sufficient to 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.
Claim(s) 1 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over Chaitanya Joshi et al; “Generative adversarial networks (GANs) for synthetic dataset generation with binary classes” available Sep 22, 2020 (hereinafter “Joshi”) in view of Sandipan Dey; “Some Statistics Concepts: Probability integral transformations” available Nov 11, 2020 (hereinafter “Dey”) in further view of Eike Christian Brechmann et al; “COPAR– Multivariate time series modeling using the COPula AutoRegressive model” available Apr 4, 2012 (hereinafter “Brechmann”) in further view of Leonie Bossemeyer; “CM Flows- Copula Density Estimation with Normalizing Flows” published Aug 2020 (hereinafter “Bossemeyer”) in further view of Matt Bonakdarpour; “Inverse Transform Sampling” available Nov 03, 2020 (hereinafter “Bonakdarpour”) in further view of Subrata Mitra et al; US 20220164346 A1 filed Nov, 2020 (hereinafter “Mitra”)
Regarding claim 1, Joshi teaches A method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, a-true data on which to model synthetic data, wherein the true data comprises continuous true data and discrete true data; (Joshi Section 2 Paragraph 4; “For the purposes of testing the performance of our synthetic data models, we will use the US adult census dataset, which is a freely available open dataset extracted from the US census bureau database. The dataset is initially designed for a binary classification problem and the task is to predict whether a person earns over $50,000 a year. The dataset is a mixture of discrete and continuous features, including age, working status (workclass), education, marital status, race, sex, relationship and hours worked per week.” Examiner notes that a true data (US adult census dataset) is received/extracted on which to model synthetic data, wherein the true data comprises continuous true data and discrete true data)
separating, by the processor, the true data into continuous data and discrete data; (Joshi Section 5 Subsection “Numerical features” Paragraph 1; “Because of the known problems in GANs in generating discrete (categorical) data features, we will first focus on training our GAN architectures by only considering the numerical (continuous) features present in the dataset. We will then briefly look into ways of representing categorical (discrete) features as continuous features and using this transformed representation for training our GANs.” Examiner notes that the true data is separated into continuous data (first focusing on numerical (continuous) features) and discrete data (then look into categorical (discrete) features))
applying, by the processor, to the continuous true data, a [probability integral] transform [to produce a first independent uniform marginal]; (Joshi Section 5 Subsection “Numerical features” Paragraph 1; “we will first focus on training our GAN architectures by only considering the numerical (continuous) features present in the dataset… The first step involves scaling the data. As we do not want any one specific feature to dominate the training procedure of our neural network we rescale all the feature values between 0 and 1.” Examiner notes that a transform (scaling data) is applied to the continuous true data (numerical (continuous) features))
applying, by the processor, to the discrete true data, a distributional transform to produce a second independent uniform marginal; (Joshi Section 5 Subsection “Categorical features” Paragraph 1; “To deal with the categorical features in our dataset we use a relatively simple strategy recently outlined. Under this approach, discrete columns are encoded into numerical [0,1] columns using the following method. Discrete values are first sorted in descending order based on their proportion in the dataset. Then, the [0,1] interval is split into sections [ac,bc] based on the proportion of each category c. To convert a discrete value to a numerical one, we replace it with a value sampled from a Gaussian distribution centred at the midpoint of [ac,bc] and with standard deviation σ = (bc − ac)/6.” Examiner notes that a discrete transform (relatively simple strategy recently outlined to encode discrete values into numerical [0, 1]) is applied to the discrete true data (categorical features))
Joshi does not teach a probability integral transform to produce a first independent uniform marginal
However, Dey does teach a probability integral transform to produce a first independent uniform marginal (Dey attached figure shows applying a probability integral transform to continuous true data (X) to produce a first independent uniform marginal)
PNG
media_image1.png
578
681
media_image1.png
Greyscale
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi and Dey. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. One of ordinary skill would have motivation to combine Joshi and Dey to simplifies complex data and enables computer simulations by the fact that a continuous random variable transformed by its own CDF with always have a U(0,1) distribution. “the fact that a continuous random variable transformed by its own CDF with always have a U(0,1) distribution.” (Dey Paragraph 1).
Joshi in view of Dey does not teach training, by the processor, a copula model [based on normalizing flows], to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;
identifying, by the copula model, a first correlated uniform marginal based on the transformed continuous true data;
identifying, by the copula model, a second correlated uniform marginal based on the transformed discrete true data;
However, Brechmann does teach training, by the processor, a copula model [based on normalizing flows], to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data; (Brechmann Page 6 Paragraph 2; “The copula autoregressive model (COPAR) for time series {Xt}t=1,...,T and {Yt}t=1,...,T has the following components. (i) Unconditional marginal distributions… The number of different pair-copulas utilized in the COPAR model” Brechmann Page 11 Paragraph 1; “a COPAR model needs the full specification of marginal distributions” Examiner notes that a copula model (copula autoregressive model) is trained to learn a copula function (pair-copulas) from the first and second independent uniform marginal (marginal distributions))
identifying, by the copula model, a first correlated uniform marginal based on the transformed continuous true data; (Brechmann Page 7 Paragraph 1; “the R-vine structure specifies the full joint distribution of {Xt} and {Yt} so that this is mainly an issue of interpretability… However note that the fitted joint distribution may in the end actually look differently depending on the order of variables in the modeling, since different copulas can be used. In the case of Gaussian pair-copulas, this is not the case, since then the R-vine copula corresponds to a multivariate Gaussian copula, where the correlation matrix can be computed from conditional correlations as given by the R-vine copula parameters.” Examiner notes that a first correlated uniform marginal (full joint distribution) is identified/specified by the R-vine structure based on the transformed continuous true data ({Xt} and {Yt}))
identifying, by the copula model, a second correlated uniform marginal based on the transformed discrete true data; (Examiner refers to previous mapping to show that a second correlated uniform marginal (full joint distribution) is identified/specified by the R-vine structure based on the transformed continuous discrete data ({Xt} and {Yt}))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, and Brechmann. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. One of ordinary skill would have motivation to combine Joshi, Dey, and Brechmann to allow for non-linear and non-symmetric modeling of serial and between-series dependence. “which exploits the enormous flexibility of vine copula models and allows for non-linear and non-symmetric modeling of serial and between-series dependence.” (Brechmann Page 2 Paragraph 3).
Joshi in view of Dey in further view of Brechmann does not teach a copula model based on normalizing flows
However, Bossemeyer does teach a copula model based on normalizing flows (Bossemeyer Abstract; “This thesis implements and evaluates a non-parametric copula estimation technique based on Normalizing Flows, a family of generative models that have recently shown success in efficient and exact density estimation” Examiner notes thesis shows a copula model is trained based on normalizing flows to learn copula function (a non-parametric copula estimation technique based on Normalizing Flows))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, and Bossemeyer. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, and Bossemeyer to incorporate CM Flows which outperforms parametric model that assumes a false copula family. “they still achieved a JS divergence of under 0.1 for the Clayton and Frank copula dataset, which out performs a parametric model that assumes a false copula family, which achieved a JS divergence of over 0.129 for the two datasets.” (Bossemeyer Page 45 Page 2).
Joshi in view of Dey in further view of Brechmann in further view of Bossemeyer does not teach applying, by the copula model, continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data;
applying, by the copula model, discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data;
However, Bonakdarpour does teach applying, by the copula model, continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data; (Bonakdarpour Section “Continuous Distributions”; “Assume we want to generate a random variable X with cumulative distribution function (CDF) FX. The inverse transform sampling algorithm is simple: 1. Generate U∼Unif(0,1) 2. Let X=F−1X(U). Then, X will follow the distribution governed by the CDF FX, which was our desired result. Note that this algorithm works in general but is not always practical. For example, inverting FX is easy if X is an exponential random variable, but its harder if X is Normal random variable.” Examiner notes that algorithm shows applying continuous inverse transform (inverse transform sampling algorithm) on the first correlated uniform marginal (random variable U uniformly distributed on [0,1]))
applying, by the copula model, discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data; (Bonakdarpour Section “Discrete Distributions”; “Now we will consider the discrete version of the inverse transform method. Assume that X is a discrete random variable such that P(X=xi)=pi. The algorithm proceeds as follows: 1. Generate U∼Unif(0,1) 2. Determine the index k such that ∑k−1j=1pj≤U<∑kj=1pj, and return X=xk. Notice that the second step requires a search. Assume our random variable X can take on any one of K values with probabilities {p1,…,pK}. We implement the algorithm below, assuming these probabilities are stored in a vector called p.vec.” Examiner notes that algorithm shows applying discrete inverse transform (discrete version of inverse transform method) on the second correlated uniform marginal (random variable U uniformly distributed on [0,1]))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, and Bonakdarpour. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, and Bonakdarpour to efficiently generate random numbers by using the inverse of the distribution’s cumulative distribution function. “Inverse transform sampling is a method for generating random numbers from any probability distribution by using its inverse cumulative distribution F−1(x).” (Bonakdarpour Paragraph 1).
Joshi in view of Dey in further view of Brechmann in further view of Bossemeyer in further view of Bonakdarpour does not teach and generating, by the copula model, as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data.
However, Mitra does teach and generating, by the copula model, as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data. (Mitra Paragraph 0063; “n some examples, the approximate query module 144 provides each of the model input(s) (O, M) to the trained sample generation model 116 to generate synthetic data entries 142. If the query 146 has been transformed into multiple sub-queries, the approximate query module 144 can generate multiple sets of synthetic data entries 142, one for each sub-query. The approximate query module 144 further combines these multiple sets of synthetic data entries 142.” Examiner notes that a synthetic data set is generated by combining the synthetic continuous data and synthetic discrete data (combines these multiple sets of synthetic data entries))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, and Mitra. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, and Mitra to reduce the storage resources consumption and computation resource consumption. “Because the synthetic data entries are accurate and can be utilized to satisfy the search query, the need to search the original dataset is eliminated. This reduces the response time of returning the search results. It also reduces the network bandwidth consumption and computational resource consumption at the server end. Further, because the model generation process generates highly targeted samples based on the specified conditions, the size of the generated synthetic data entries are smaller than the size of the generated synthetic data entries of the existing approach. Therefore, the storage resource consumption at each client computing system is also reduced.” (Mitra Paragraph 0024).
Regarding claim 10, Joshi teaches A computer implemented method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, a-true data on which to model synthetic data, wherein the true data comprises continuous true data; (Joshi Section 2 Paragraph 4; “For the purposes of testing the performance of our synthetic data models, we will use the US adult census dataset, which is a freely available open dataset extracted from the US census bureau database. The dataset is initially designed for a binary classification problem and the task is to predict whether a person earns over $50,000 a year. The dataset is a mixture of discrete and continuous features, including age, working status (workclass), education, marital status, race, sex, relationship and hours worked per week.” Examiner notes that a true data (US adult census dataset) is received/extracted on which to model synthetic data, wherein the true data comprises continuous true data)
Applying a [probability integral] transform to the continuous true data [to produce an independent uniform marginal]; (Joshi Section 5 Subsection “Numerical features” Paragraph 1; “we will first focus on training our GAN architectures by only considering the numerical (continuous) features present in the dataset… The first step involves scaling the data. As we do not want any one specific feature to dominate the training procedure of our neural network we rescale all the feature values between 0 and 1.” Examiner notes that a transform (scaling data) is applied to the continuous true data (numerical (continuous) features))
Joshi does not teach a probability integral transform to produce a first independent uniform marginal
However, Dey does teach a probability integral transform to produce a first independent uniform marginal (Dey attached figure shows applying a probability integral transform to continuous true data (X) to produce a first independent uniform marginal)
PNG
media_image1.png
578
681
media_image1.png
Greyscale
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi and Dey. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. One of ordinary skill would have motivation to combine Joshi and Dey to simplifies complex data and enables computer simulations by the fact that a continuous random variable transformed by its own CDF with always have a U(0,1) distribution. “the fact that a continuous random variable transformed by its own CDF with always have a U(0,1) distribution.” (Dey Paragraph 1).
Joshi in view of Dey does not teach training, by the processor, a copula model [based on normalizing flows], to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;
identifying, by the copula model, a first correlated uniform marginal;
However, Brechmann does teach training, by the processor, a copula model [based on normalizing flows], to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data; (Brechmann Page 6 Paragraph 2; “The copula autoregressive model (COPAR) for time series {Xt}t=1,...,T and {Yt}t=1,...,T has the following components. (i) Unconditional marginal distributions… The number of different pair-copulas utilized in the COPAR model” Brechmann Page 11 Paragraph 1; “a COPAR model needs the full specification of marginal distributions” Examiner notes that a copula model (copula autoregressive model) is trained to learn a copula function (pair-copulas) from the first and second independent uniform marginal (marginal distributions))
identifying, by the copula model, a first correlated uniform marginal; (Brechmann Page 7 Paragraph 1; “the R-vine structure specifies the full joint distribution of {Xt} and {Yt} so that this is mainly an issue of interpretability… However note that the fitted joint distribution may in the end actually look differently depending on the order of variables in the modeling, since different copulas can be used. In the case of Gaussian pair-copulas, this is not the case, since then the R-vine copula corresponds to a multivariate Gaussian copula, where the correlation matrix can be computed from conditional correlations as given by the R-vine copula parameters.” Examiner notes that a first correlated uniform marginal (full joint distribution) is identified/specified by the R-vine structure)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, and Brechmann. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. One of ordinary skill would have motivation to combine Joshi, Dey, and Brechmann to allow for non-linear and non-symmetric modeling of serial and between-series dependence. “which exploits the enormous flexibility of vine copula models and allows for non-linear and non-symmetric modeling of serial and between-series dependence.” (Brechmann Page 2 Paragraph 3).
Joshi in view of Dey in further view of Brechmann does not teach a copula model based on normalizing flows
However, Bossemeyer does teach a copula model based on normalizing flows (Bossemeyer Abstract; “This thesis implements and evaluates a non-parametric copula estimation technique based on Normalizing Flows, a family of generative models that have recently shown success in efficient and exact density estimation” Examiner notes thesis shows a copula model is trained based on normalizing flows to learn copula function (a non-parametric copula estimation technique based on Normalizing Flows))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, and Bossemeyer. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, and Bossemeyer to incorporate CM Flows which outperforms parametric model that assumes a false copula family. “they still achieved a JS divergence of under 0.1 for the Clayton and Frank copula dataset, which out performs a parametric model that assumes a false copula family, which achieved a JS divergence of over 0.129 for the two datasets.” (Bossemeyer Page 45 Page 2).
Joshi in view of Dey in further view of Brechmann in further view of Bossemeyer does not teach applying, by the copula model, continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data;
However, Bonakdarpour does teach applying, by the copula model, continuous inverse transform sampling on the first correlated uniform marginal to generate synthetic continuous data; (Bonakdarpour Section “Continuous Distributions”; “Assume we want to generate a random variable X with cumulative distribution function (CDF) FX. The inverse transform sampling algorithm is simple: 1. Generate U∼Unif(0,1) 2. Let X=F−1X(U). Then, X will follow the distribution governed by the CDF FX, which was our desired result. Note that this algorithm works in general but is not always practical. For example, inverting FX is easy if X is an exponential random variable, but its harder if X is Normal random variable.” Examiner notes that algorithm shows applying continuous inverse transform (inverse transform sampling algorithm) on the first correlated uniform marginal (random variable U uniformly distributed on [0,1]))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, and Bonakdarpour. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, and Bonakdarpour to efficiently generate random numbers by using the inverse of the distribution’s cumulative distribution function. “Inverse transform sampling is a method for generating random numbers from any probability distribution by using its inverse cumulative distribution F−1(x).” (Bonakdarpour Paragraph 1).
Joshi in view of Dey in further view of Brechmann in further view of Bossemeyer in further view of Bonakdarpour does not teach and generating, by the copula model, the synthetic data as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data.
However, Mitra does teach and generating, by the copula model, the synthetic data as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data. (Mitra Paragraph 0063; “n some examples, the approximate query module 144 provides each of the model input(s) (O, M) to the trained sample generation model 116 to generate synthetic data entries 142. If the query 146 has been transformed into multiple sub-queries, the approximate query module 144 can generate multiple sets of synthetic data entries 142, one for each sub-query. The approximate query module 144 further combines these multiple sets of synthetic data entries 142.” Examiner notes that a synthetic data set is generated by combining the synthetic continuous data and synthetic discrete data (combines these multiple sets of synthetic data entries))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, and Mitra. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, and Mitra to reduce the storage resources consumption and computation resource consumption. “Because the synthetic data entries are accurate and can be utilized to satisfy the search query, the need to search the original dataset is eliminated. This reduces the response time of returning the search results. It also reduces the network bandwidth consumption and computational resource consumption at the server end. Further, because the model generation process generates highly targeted samples based on the specified conditions, the size of the generated synthetic data entries are smaller than the size of the generated synthetic data entries of the existing approach. Therefore, the storage resource consumption at each client computing system is also reduced.” (Mitra Paragraph 0024).
Claim(s) 3, 4, 12, and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Chaitanya Joshi et al; “Generative adversarial networks (GANs) for synthetic dataset generation with binary classes” available Sep 22, 2020 (hereinafter “Joshi”) in view of Sandipan Dey; “Some Statistics Concepts: Probability integral transformations” available Nov 11, 2020 (hereinafter “Dey”) in further view of Eike Christian Brechmann et al; “COPAR– Multivariate time series modeling using the COPula AutoRegressive model” available Apr 4, 2012 (hereinafter “Brechmann”) in further view of Leonie Bossemeyer; “CM Flows- Copula Density Estimation with Normalizing Flows” published Aug 2020 (hereinafter “Bossemeyer”) in further view of Matt Bonakdarpour; “Inverse Transform Sampling” available Nov 03, 2020 (hereinafter “Bonakdarpour”) in further view of Subrata Mitra et al; US 20220164346 A1 filed Nov, 2020 (hereinafter “Mitra”) in further view of George Papamakarios et al; “Normalizing Flows for Probabilistic Modeling and Inference” published Dec 5, 2019 (hereinafter “Papamakarios”)
Regarding claim 3, Joshi does not teach The method of claim 1, further comprising: using an autoregressive density network with the normalizing flow.
However, Papamakarios does teach The method of claim 1, further comprising: using an autoregressive density network with the normalizing flow. (Papamakarios Page 2 Paragraph 2; “In this paper we present Masked Autoregressive Flow (MAF), which is a particular implementation of the above normalizing flow that uses the Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations without the sequential loop that is typical of autoregressive models, and thus makes MAF fast to evaluate and train on parallel computing architectures such as Graphics Processing Units (GPUs).” Examiner notes that a autoregressive density network (Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations) is used with the normalizing flow (implementation of the above normalizing flow))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Papamakarios teaches the use of flows for tasks such as generative modeling, approximate inference, and supervised learning. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios to perform both sampling and density evaluation efficiently. “Yet if both sampling and density evaluation must be done often or quickly, then implementing the autoregressive flow with a coupling-based conditioner will make both operations efficient at the cost of expressive power.” (Papamakarios Page 47 Paragraph 3).
Regarding claim 4, Joshi does not teach The method of claim 3, wherein the autoregressive density network comprises a masked autoregressive network.
However, Papamakarios does teach The method of claim 3, wherein the autoregressive density network comprises a masked autoregressive network. (Papamakarios Page 2 Paragraph 2; “In this paper we present Masked Autoregressive Flow (MAF), which is a particular implementation of the above normalizing flow that uses the Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations without the sequential loop that is typical of autoregressive models, and thus makes MAF fast to evaluate and train on parallel computing architectures such as Graphics Processing Units (GPUs).” Examiner notes that a autoregressive density network comprises a masked autoregressive network (Masked Autoencoder for Distribution Estimation (MADE)))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Papamakarios teaches the use of flows for tasks such as generative modeling, approximate inference, and supervised learning. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios to perform both sampling and density evaluation efficiently. “Yet if both sampling and density evaluation must be done often or quickly, then implementing the autoregressive flow with a coupling-based conditioner will make both operations efficient at the cost of expressive power.” (Papamakarios Page 47 Paragraph 3).
Regarding claim 12, Joshi does not teach The method of claim 10, further comprising: using an autoregressive density network with the normalizing flow.
However, Papamakarios does teach The method of claim 10, further comprising: using an autoregressive density network with the normalizing flow. (Papamakarios Page 2 Paragraph 2; “In this paper we present Masked Autoregressive Flow (MAF), which is a particular implementation of the above normalizing flow that uses the Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations without the sequential loop that is typical of autoregressive models, and thus makes MAF fast to evaluate and train on parallel computing architectures such as Graphics Processing Units (GPUs).” Examiner notes that a autoregressive density network (Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations) is used with the normalizing flow (implementation of the above normalizing flow))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Papamakarios teaches the use of flows for tasks such as generative modeling, approximate inference, and supervised learning. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios to perform both sampling and density evaluation efficiently. “Yet if both sampling and density evaluation must be done often or quickly, then implementing the autoregressive flow with a coupling-based conditioner will make both operations efficient at the cost of expressive power.” (Papamakarios Page 47 Paragraph 3).
Regarding claim 13, Joshi does not teach The method of claim 12, wherein the autoregressive density network comprises a masked autoregressive network.
However, Papamakarios does teach The method of claim 3, wherein the autoregressive density network comprises a masked autoregressive network. (Papamakarios Page 2 Paragraph 2; “In this paper we present Masked Autoregressive Flow (MAF), which is a particular implementation of the above normalizing flow that uses the Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations without the sequential loop that is typical of autoregressive models, and thus makes MAF fast to evaluate and train on parallel computing architectures such as Graphics Processing Units (GPUs).” Examiner notes that a autoregressive density network comprises a masked autoregressive network (Masked Autoencoder for Distribution Estimation (MADE)))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Papamakarios teaches the use of flows for tasks such as generative modeling, approximate inference, and supervised learning. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios to perform both sampling and density evaluation efficiently. “Yet if both sampling and density evaluation must be done often or quickly, then implementing the autoregressive flow with a coupling-based conditioner will make both operations efficient at the cost of expressive power.” (Papamakarios Page 47 Paragraph 3).
Claim(s) 5 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Chaitanya Joshi et al; “Generative adversarial networks (GANs) for synthetic dataset generation with binary classes” available Sep 22, 2020 (hereinafter “Joshi”) in view of Sandipan Dey; “Some Statistics Concepts: Probability integral transformations” available Nov 11, 2020 (hereinafter “Dey”) in further view of Eike Christian Brechmann et al; “COPAR– Multivariate time series modeling using the COPula AutoRegressive model” available Apr 4, 2012 (hereinafter “Brechmann”) in further view of Leonie Bossemeyer; “CM Flows- Copula Density Estimation with Normalizing Flows” published Aug 2020 (hereinafter “Bossemeyer”) in further view of Matt Bonakdarpour; “Inverse Transform Sampling” available Nov 03, 2020 (hereinafter “Bonakdarpour”) in further view of Subrata Mitra et al; US 20220164346 A1 filed Nov, 2020 (hereinafter “Mitra”) in further view of Zachary M. Ziegler et al; “Latent Normalizing Flows for Discrete Sequences” published Jun 4, 2019 (hereinafter “Ziegler”)
Regarding claim 5, Joshi does not teach The method of claim 1, further comprising: learning, by an autoregressive density network, the normalizing flows based on a functional relationship within the continuous true data to map discrete variables to corresponding random variables.
However, Ziegler does teach The method of claim 1, further comprising: learning, by an autoregressive density network, the normalizing flows based on a functional relationship within the continuous true data to map discrete variables to corresponding random variables. (Ziegler Page 4 Paragraph 2; “Using these building blocks, we aim to develop flexible flow-based models for discrete sequences.” Ziegler Page 4 Paragraph 4; “we model the generation of a discrete sequence x1:T = {x1,...,xT} conditioned on a latent sequence z1:T made up of continuous random vectors {z1,...,zT} with zt ∈ RH andH is a hidden dimension. Define p(z1:T|T)as our prior distribution, and generate from the conditional distribution over discrete observed variables p(x1:T|z1:T,T).” Examiner notes that the normalizing flow (flow-based models) learns based on a functional relationship within the continuous true data (conditioned on a latent sequence z1:T made up of continuous random vectors) to map discrete variables to corresponding random variables (generation of a discrete sequence))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Ziegler. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Zieglar teaches a VAE-based generative model which jointly learns a normalizing flow-based distribution in the latent space and a stochastic mapping to an observed discrete space. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Ziegler to reduce the storage resources consumption and computation resource consumption. “Because the synthetic data entries are accurate and can be utilized to satisfy the search query, the need to search the original dataset is eliminated. This reduces the response time of returning the search results. It also reduces the network bandwidth consumption and computational resource consumption at the server end. Further, because the model generation process generates highly targeted samples based on the specified conditions, the size of the generated synthetic data entries are smaller than the size of the generated synthetic data entries of the existing approach. Therefore, the storage resource consumption at each client computing system is also reduced.” (Mitra Paragraph 0024).
Regarding claim 14, Joshi does not teach The method of claim 10, further comprising: learning, by an autoregressive density network, the normalizing flows based on a functional relationship within the continuous true data to map discrete variables to corresponding random variables.
However, Ziegler does teach The method of claim 10, further comprising: learning, by an autoregressive density network, the normalizing flows based on a functional relationship within the continuous true data to map discrete variables to corresponding random variables. (Ziegler Page 4 Paragraph 2; “Using these building blocks, we aim to develop flexible flow-based models for discrete sequences.” Ziegler Page 4 Paragraph 4; “we model the generation of a discrete sequence x1:T = {x1,...,xT} conditioned on a latent sequence z1:T made up of continuous random vectors {z1,...,zT} with zt ∈ RH andH is a hidden dimension. Define p(z1:T|T)as our prior distribution, and generate from the conditional distribution over discrete observed variables p(x1:T|z1:T,T).” Examiner notes that the normalizing flow (flow-based models) learns based on a functional relationship within the continuous true data (conditioned on a latent sequence z1:T made up of continuous random vectors) to map discrete variables to corresponding random variables (generation of a discrete sequence))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Ziegler. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Dey teaches probability integral transformations. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Zieglar teaches a VAE-based generative model which jointly learns a normalizing flow-based distribution in the latent space and a stochastic mapping to an observed discrete space. One of ordinary skill would have motivation to combine Joshi, Dey, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Ziegler to reduce the storage resources consumption and computation resource consumption. “Because the synthetic data entries are accurate and can be utilized to satisfy the search query, the need to search the original dataset is eliminated. This reduces the response time of returning the search results. It also reduces the network bandwidth consumption and computational resource consumption at the server end. Further, because the model generation process generates highly targeted samples based on the specified conditions, the size of the generated synthetic data entries are smaller than the size of the generated synthetic data entries of the existing approach. Therefore, the storage resource consumption at each client computing system is also reduced.” (Mitra Paragraph 0024).
Claim(s) 6 is rejected under 35 U.S.C. 103 as being unpatentable over Chaitanya Joshi et al; “Generative adversarial networks (GANs) for synthetic dataset generation with binary classes” available Sep 22, 2020 (hereinafter “Joshi”) in view of Eike Christian Brechmann et al; “COPAR– Multivariate time series modeling using the COPula AutoRegressive model” available Apr 4, 2012 (hereinafter “Brechmann”) in further view of Leonie Bossemeyer; “CM Flows- Copula Density Estimation with Normalizing Flows” published Aug 2020 (hereinafter “Bossemeyer”) in further view of Matt Bonakdarpour; “Inverse Transform Sampling” available Nov 03, 2020 (hereinafter “Bonakdarpour”) in further view of Subrata Mitra et al; US 20220164346 A1 filed Nov, 2020 (hereinafter “Mitra”)
Regarding claim 6, Joshi teaches A method for generating synthetic data using a learned copula function, comprising: receiving, by a processor executing a program in connection with a memory storing the program, a-true data on which to model synthetic data, wherein the true data comprises discrete true data including categorical discrete data and ordinal discrete data; (Joshi Section 2 Paragraph 4; “For the purposes of testing the performance of our synthetic data models, we will use the US adult census dataset, which is a freely available open dataset extracted from the US census bureau database. The dataset is initially designed for a binary classification problem and the task is to predict whether a person earns over $50,000 a year. The dataset is a mixture of discrete and continuous features, including age, working status (workclass), education, marital status, race, sex, relationship and hours worked per week.” Examiner notes that a true data (US adult census dataset) is received/extracted on which to model synthetic data, wherein the true data comprises categorical discrete data (sex) and ordinal discrete data (education))
Applying a distributional transform to the discrete true data to produce an independent uniform marginal; (Joshi Section 5 Subsection “Categorical features” Paragraph 1; “To deal with the categorical features in our dataset we use a relatively simple strategy recently outlined. Under this approach, discrete columns are encoded into numerical [0,1] columns using the following method. Discrete values are first sorted in descending order based on their proportion in the dataset. Then, the [0,1] interval is split into sections [ac,bc] based on the proportion of each category c. To convert a discrete value to a numerical one, we replace it with a value sampled from a Gaussian distribution centred at the midpoint of [ac,bc] and with standard deviation σ = (bc − ac)/6.” Examiner notes that a discrete transform (relatively simple strategy recently outlined to encode discrete values into numerical [0, 1]) is applied to the discrete true data (categorical features))
Joshi does not teach training, by the processor, a copula model [based on normalizing flows], to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data;
identifying, by the copula model, a first correlated uniform marginal
However, Brechmann does teach training, by the processor, a copula model [based on normalizing flows], to learn a copula function from the first independent uniform marginal produced from the continuous true data and the second independent uniform marginal produced from the discrete true data; (Brechmann Page 6 Paragraph 2; “The copula autoregressive model (COPAR) for time series {Xt}t=1,...,T and {Yt}t=1,...,T has the following components. (i) Unconditional marginal distributions… The number of different pair-copulas utilized in the COPAR model” Brechmann Page 11 Paragraph 1; “a COPAR model needs the full specification of marginal distributions” Examiner notes that a copula model (copula autoregressive model) is trained to learn a copula function (pair-copulas) from the first and second independent uniform marginal (marginal distributions))
identifying, by the copula model, a first correlated uniform marginal; (Brechmann Page 7 Paragraph 1; “the R-vine structure specifies the full joint distribution of {Xt} and {Yt} so that this is mainly an issue of interpretability… However note that the fitted joint distribution may in the end actually look differently depending on the order of variables in the modeling, since different copulas can be used. In the case of Gaussian pair-copulas, this is not the case, since then the R-vine copula corresponds to a multivariate Gaussian copula, where the correlation matrix can be computed from conditional correlations as given by the R-vine copula parameters.” Examiner notes that a first correlated uniform marginal (full joint distribution) is identified/specified by the R-vine structure)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi and Brechmann. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. One of ordinary skill would have motivation to combine Joshi and Brechmann to allow for non-linear and non-symmetric modeling of serial and between-series dependence. “which exploits the enormous flexibility of vine copula models and allows for non-linear and non-symmetric modeling of serial and between-series dependence.” (Brechmann Page 2 Paragraph 3).
Joshi in view of Brechmann does not teach a copula model based on normalizing flows
However, Bossemeyer does teach a copula model based on normalizing flows (Bossemeyer Abstract; “This thesis implements and evaluates a non-parametric copula estimation technique based on Normalizing Flows, a family of generative models that have recently shown success in efficient and exact density estimation” Examiner notes thesis shows a copula model is trained based on normalizing flows to learn copula function (a non-parametric copula estimation technique based on Normalizing Flows))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Brechmann, and Bossemeyer. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. One of ordinary skill would have motivation to combine Joshi, Brechmann, and Bossemeyer to incorporate CM Flows which outperforms parametric model that assumes a false copula family. “they still achieved a JS divergence of under 0.1 for the Clayton and Frank copula dataset, which out performs a parametric model that assumes a false copula family, which achieved a JS divergence of over 0.129 for the two datasets.” (Bossemeyer Page 45 Page 2).
Joshi in view of Brechmann in further view of Bossemeyer does not teach applying, by the copula model, discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data;
However, Bonakdarpour does teach applying, by the copula model, discrete inverse transform sampling on the second correlated uniform marginal to generate synthetic discrete data; (Bonakdarpour Section “Discrete Distributions”; “Now we will consider the discrete version of the inverse transform method. Assume that X is a discrete random variable such that P(X=xi)=pi. The algorithm proceeds as follows: 1. Generate U∼Unif(0,1) 2. Determine the index k such that ∑k−1j=1pj≤U<∑kj=1pj, and return X=xk. Notice that the second step requires a search. Assume our random variable X can take on any one of K values with probabilities {p1,…,pK}. We implement the algorithm below, assuming these probabilities are stored in a vector called p.vec.” Examiner notes that algorithm shows applying discrete inverse transform (discrete version of inverse transform method) on the second correlated uniform marginal (random variable U uniformly distributed on [0,1]))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Brechmann, Bossemeyer, and Bonakdarpour. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. One of ordinary skill would have motivation to combine Joshi, Brechmann, Bossemeyer, and Bonakdarpour to efficiently generate random numbers by using the inverse of the distribution’s cumulative distribution function. “Inverse transform sampling is a method for generating random numbers from any probability distribution by using its inverse cumulative distribution F−1(x).” (Bonakdarpour Paragraph 1).
Joshi in view of Brechmann in further view of Bossemeyer in further view of Bonakdarpour does not teach and generating, by the copula model, the synthetic data as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data.
However, Mitra does teach and generating, by the copula model, the synthetic data as a synthetic dataset by combining the synthetic continuous data and the synthetic discrete data. (Mitra Paragraph 0063; “n some examples, the approximate query module 144 provides each of the model input(s) (O, M) to the trained sample generation model 116 to generate synthetic data entries 142. If the query 146 has been transformed into multiple sub-queries, the approximate query module 144 can generate multiple sets of synthetic data entries 142, one for each sub-query. The approximate query module 144 further combines these multiple sets of synthetic data entries 142.” Examiner notes that a synthetic data set is generated by combining the synthetic continuous data and synthetic discrete data (combines these multiple sets of synthetic data entries))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Brechmann, Bossemeyer, Bonakdarpour, and Mitra. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. One of ordinary skill would have motivation to combine Joshi, Brechmann, Bossemeyer, Bonakdarpour, and Mitra to reduce the storage resources consumption and computation resource consumption. “Because the synthetic data entries are accurate and can be utilized to satisfy the search query, the need to search the original dataset is eliminated. This reduces the response time of returning the search results. It also reduces the network bandwidth consumption and computational resource consumption at the server end. Further, because the model generation process generates highly targeted samples based on the specified conditions, the size of the generated synthetic data entries are smaller than the size of the generated synthetic data entries of the existing approach. Therefore, the storage resource consumption at each client computing system is also reduced.” (Mitra Paragraph 0024).
Claim(s) 8 and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Chaitanya Joshi et al; “Generative adversarial networks (GANs) for synthetic dataset generation with binary classes” available Sep 22, 2020 (hereinafter “Joshi”) in view of Eike Christian Brechmann et al; “COPAR– Multivariate time series modeling using the COPula AutoRegressive model” available Apr 4, 2012 (hereinafter “Brechmann”) in further view of Leonie Bossemeyer; “CM Flows- Copula Density Estimation with Normalizing Flows” published Aug 2020 (hereinafter “Bossemeyer”) in further view of Matt Bonakdarpour; “Inverse Transform Sampling” available Nov 03, 2020 (hereinafter “Bonakdarpour”) in further view of Subrata Mitra et al; US 20220164346 A1 filed Nov, 2020 (hereinafter “Mitra”) in further view of George Papamakarios et al; “Normalizing Flows for Probabilistic Modeling and Inference” published Dec 5, 2019 (hereinafter “Papamakarios”)
Regarding claim 8, Joshi does not teach The method of claim 6, further comprising: using an autoregressive density network with the normalizing flow.
However, Papamakarios does teach The method of claim 6, further comprising: using an autoregressive density network with the normalizing flow. (Papamakarios Page 2 Paragraph 2; “In this paper we present Masked Autoregressive Flow (MAF), which is a particular implementation of the above normalizing flow that uses the Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations without the sequential loop that is typical of autoregressive models, and thus makes MAF fast to evaluate and train on parallel computing architectures such as Graphics Processing Units (GPUs).” Examiner notes that a autoregressive density network (Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations) is used with the normalizing flow (implementation of the above normalizing flow))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Papamakarios teaches the use of flows for tasks such as generative modeling, approximate inference, and supervised learning. One of ordinary skill would have motivation to combine Joshi, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios to perform both sampling and density evaluation efficiently. “Yet if both sampling and density evaluation must be done often or quickly, then implementing the autoregressive flow with a coupling-based conditioner will make both operations efficient at the cost of expressive power.” (Papamakarios Page 47 Paragraph 3).
Regarding claim 9, Joshi does not teach The method of claim 8, wherein the autoregressive density network comprises a masked autoregressive network.
However, Papamakarios does teach The method of claim 8, wherein the autoregressive density network comprises a masked autoregressive network. (Papamakarios Page 2 Paragraph 2; “In this paper we present Masked Autoregressive Flow (MAF), which is a particular implementation of the above normalizing flow that uses the Masked Autoencoder for Distribution Estimation (MADE) [9] as a building block. The use of MADE enables density evaluations without the sequential loop that is typical of autoregressive models, and thus makes MAF fast to evaluate and train on parallel computing architectures such as Graphics Processing Units (GPUs).” Examiner notes that a autoregressive density network comprises a masked autoregressive network (Masked Autoencoder for Distribution Estimation (MADE)))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Joshi, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios. Joshi teaches generative adversarial networks for synthetic dataset generation with binary classes. Brechmann teaches copula-based model which allows for non-linear and asymmetric modeling of serial as well as between-series dependencies. Bossemeyer teaches a non-parametric copula estimation technique based on Normalizing Flows. Bonakdarpour teaches Inverse transform sampling. Mitra teaches a model training system trains a sample generation model configured to generate synthetic data entries for a dataset. Papamakarios teaches the use of flows for tasks such as generative modeling, approximate inference, and supervised learning. One of ordinary skill would have motivation to combine Joshi, Brechmann, Bossemeyer, Bonakdarpour, Mitra, and Papamakarios to perform both sampling and density evaluation efficiently. “Yet if both sampling and density evaluation must be done often or quickly, then implementing the autoregressive flow with a coupling-based conditioner will make both operations efficient at the cost of expressive power.” (Papamakarios Page 47 Paragraph 3).
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DANIEL DUC TRAN whose telephone number is (571)272-6870. The examiner can normally be reached Mon-Fri 8:00-5:00 EST.
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, Viker Lamardo can be reached at (571) 270-5871. 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.
/D.D.T./Examiner, Art Unit 2147
/ERIC NILSSON/Primary Examiner, Art Unit 2151
Prosecution Insights