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 .
DETAILED ACTION
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 9-10 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding claim 9, the limitations below use and/or:
and/or the a posteriori distribution is determined as a function
The use of “and/or” makes the claim unclear therefore, the scope of the claim is, indefinite. The use of or can include all of the limitations stated above. The and/or does not make the claim clear as to what the limitation is for controlling using a control single for the limitations specify different uses.
Regarding claim 10 the limitations below use and/or:
and/or the measurement
The use of “and/or” makes the claim unclear therefore, the scope of the claim is, indefinite. The use of or can include all of the limitations stated above. The and/or does not make the claim clear as to what the limitation is for controlling using a control single for the limitations specify different uses.
and/or the simulated measurement,
The use of “and/or” makes the claim unclear therefore, the scope of the claim is, indefinite. The use of or can include all of the limitations stated above. The and/or does not make the claim clear as to what the limitation is for controlling using a control single for the limitations specify different uses.
Claim 10 recites "the probability distribution being defined as a function of at least one hyperparameter" conflicts with claim 1 recitation “the probabilistic model being defined as a function of at least one hyperparameter” makes the claim unclear on which hyperparameter is being defined in claim.1
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.
Claim 1-13 is rejected under 35 USC § 101 because claimed invention is directed to the abstract idea without significantly more.
Regarding claim 1:
Step 1 Analysis: Is the claim to a process, machine, manufacture or composition of matter? See MPEP § 2106.03.
Claim 1 is a method claim, therefore it falls under one of four categories of statutory subject matter (machine/products/apparatus, process/method, manufactures and compositions of mater.
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “determining, as a function of the probabilistic model, an instruction for a first measurement” is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation “determining an a posteriori distribution over values for the at least one hyperparameter” is the abstract idea of a mathematical calculation. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
The limitation “for the at least one hyperparameter, determining an instruction of the first measurement” is the abstract idea of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation “determining an instruction for a second measurement as a function of the probabilistic model” is the abstract idea of a mental process and a mathematical concept that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation “determining at least one value of the at least one hyperparameter as a function of the second measurement. “Is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “computer-implemented method” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
The limitation, “providing a probabilistic model, the probabilistic model including a probability distribution including Gaussian process or Bayesian Neural Network” is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering). See MPEP §§ 2106.04(d), 2106.05(g).
The limitation “outputting the instruction for the first measurement” is an additional element that amounts to adding insignificant extra-solution activity of mere data output to the judicial exception. The claim recites generating an output in the form of instruction for an activity. See MPEP §§ 2106.04(d), 2106.05(g).
The limitation “outputting the instruction for the second measurement” is an additional element that amounts to adding insignificant extra-solution activity of mere data output to the judicial exception. The claim recites generating an output in the form of instruction for an activity. See MPEP §§ 2106.04(d), 2106.05(g).
Accordingly, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea when considered as an ordered combination and as a whole.
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “computer-implemented method” is an additional element to adding insignificant extra-solution activity to the judicial exception. See MPEP § 2106.05(g). Furthermore, the additional element is directed to a method performed by a computer, which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II).
The limitation, “providing a probabilistic model, the probabilistic model including a probability distribution including Gaussian process or Bayesian Neural Network” is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. (i.e., receiving the model and performing mere data gathering), which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II). See MPEP §§ 2106.04(d), 2106.05(g).
The limitation “outputting the instruction for the first measurement” amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP § 2106.05(g). Furthermore, the additional element is directed to produce the instruction as an output for a particular input, which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II).
The limitation “outputting the instruction for the second measurement” amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP § 2106.05(g). Furthermore, the additional element is directed to produce the instruction as an output for a particular input, which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II).
Therefore, in examining elements as recited by the limitations individually and as an ordered combination, as a whole the independent claim limitations do not recite what have the courts have identified as “significantly more”.
Similarly, claim 12-13 are rejected under 35 USC § 101 because claimed invention is directed to the abstract idea without significantly more.
Regarding claim 12
Step 1 Analysis: Is the claim to a process, machine, manufacture or composition of matter? See MPEP § 2106.03.
Claim 12 is a product claim, therefore it falls under one of four categories of statutory subject matter (machine/products/apparatus, process/method, manufactures and compositions of matter.
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “at least one processor; and at least one memory; wherein the at least one processor is configured to execute computer-readable instructions, the at least one memory being configured to store a model and computer- readable instructions upon whose execution by the at least one processor, the at least one processor performs” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or
merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “at least one processor; and at least one memory; wherein the at least one processor is configured to execute computer-readable instructions, the at least one memory being configured to store a model and computer- readable instructions upon whose execution by the at least one processor, the at least one processor performs” is an additional element that amounts to adding insignificant extra-solution activity to the judicial exception. See MPEP § 2106.05(g). Furthermore, the additional element is directed to storing and retrieving information in memory and executed by a processor which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II).
And for all other claim elements of claim 12 they are rejected using the PEG analysis of claim 1 since they are analogous claims.
Regarding claim 13,
Step 1 Analysis: Is the claim to a process, machine, manufacture or composition of matter? See MPEP § 2106.03.
Clam 13 is drawn to a computer readable medium, therefore claim 13 falls under one of four categories of statutory subject matter (machine/products/apparatus, process/method, manufactures and compositions of matter.
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “A non-transitory computer readable medium on which is stored a computer program including computer-readable instructions for machine learning, the instructions, when executed by a computer, causes the computer to perform” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its
ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “A non-transitory computer readable medium on which is stored a computer program including computer-readable instructions for machine learning, the instructions, when executed by a computer, causes the computer to perform” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
And for all other claim elements of claim 13 they are rejected using the PEG analysis of claim 1 since they are analogous claims.
Thus, considering the additional elements individually and in combination and the claims as a whole, the additional elements do not provide significantly more than the abstract idea. The claims are not eligible subject matter.
Regarding claim 2
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “checking whether the a posteriori distribution satisfies a condition is the abstract idea of a mathematical calculation, as directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
The limitation “,the at least one value for the at least one hyperparameter subsequently being determined based on the a posteriori distribution satisfying the condition” is the abstract idea of a mathematical concept ” is the abstract idea of a mathematical calculation, as directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
The limitation “further a posteriori distribution over values for the at least one hyperparameter subsequently being determined based on the a posteriori distribution not satisfing the condition.” is a abstract idea of mathematical calculation, as directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical
operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Step 2A Prong Two and Step 2B Analysis:
The claim does not recite any additional limitations that is integrated the judicial exception into a practical application. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B).
Regarding claim 3
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
The limitation “wherein the a posteriori distribution assigns values their probability measure, the condition including a first criterion that is satisfied” is a abstract idea of a mathematical concept as it involves assigning a value and checking a condition. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “when more than a specified percentage of probability measures of the distribution lie within an interval that is defined as a function of a largest probability measure of the distribution and includes the measure, and it being checked whether the first criterion is satisfied.” is a abstract idea of a mathematical concept as a claim recites mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Step 2A Prong Two and Step 2B Analysis:
The claim does not recite any additional limitations that is integrated the judicial exception into a practical application. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B).
Regarding claim 4
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “condition includes a second criterion that is satisfied when a distance including a Kullback-Leibler divergence, between the a posteriori distribution and a Gaussian distribution is smaller than a first threshold” is a abstract idea of the mathematical concept , as directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
The limitation “being checked whether the second criterion is satisfied.” is a abstract idea of the mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
Step 2A Prong Two and Step 2B Analysis:
The claim does not recite any additional limitations that is integrated the judicial exception into a practical application. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B).
Regarding claim 5
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
Claim 5 “wherein the condition includes a third criterion that is satisfied when the a posteriori distribution is unimodal, and it being checked whether the third criterion is satisfied” is a abstract idea of mathematical concept as , as directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Step 2A Prong Two and Step 2B Analysis:
The claim does not recite any additional limitations that is integrated the judicial exception into a practical application. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B).
Regarding claim 6
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
Limitation “a preceding a posteriori distribution is determined in each of a plurality of iterations preceding the iteration” is a abstract idea of mathematical concept, as, directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Limitation “condition including a fourth criterion that is satisfied when a difference including a Kullback-Leibler divergence, between a preceding a posteriori distribution”” is a abstract idea of mathematical concept as, as directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Limitation “posteriori distribution is smaller than a second threshold, and it being checked whether the fourth criterion is satisfied.” is a abstract idea of mathematical concept as, as directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Step 2A Prong Two and Step 2B Analysis:
The claim does not recite any additional limitations that is integrated the judicial exception into a practical application. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B).
Regarding claim 7
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “an entropy or a variance, of the a posteriori distribution is determined, the condition including a fifth criterion that is satisfied” is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation “characteristic is smaller than a third threshold, and it being checked whether the fifth criterion is satisfied.” is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
Step 2A Prong Two and Step 2B Analysis:
The claim does not recite any additional limitations that is integrated the judicial exception into a practical application. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B).
Regarding claim 8
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “ a preceding a posteriori distribution is determined Is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “the a posteriori distribution satisfying the condition when the a posteriori distribution and the at least one preceding a posteriori distribution satisfy the condition or the first criterion” Is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
Step 2A Prong Two Analysis: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “at least one iteration preceding the iteration,” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “at least one iteration preceding the iteration,” is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process. See MPEP §§ 2106.04(d), 2106.05(f)(2).
Regarding claim 9
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “wherein the value is determined as a function of a solution of an optimization problem that is a function of the at least one hyperparameter” is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation value being determined as a function of the solution of the optimization problem that is defined as a function of an objective function that is a function of the at least one hyperparameter, is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The Limitation “Posteriori distribution is determined as the function of a sample drawn from a set of values for at least one hyperparameter”, Is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
Step 2A Prong Two and Step 2B Analysis:
The claim does not recite any additional limitations that is integrated the judicial exception into a practical application. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception. In addition, the claim limitations do not include additional elements that are sufficient to amount to significantly more than the judicial exception (Step 2B).
Regarding claim 10
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation at “probabilistic model includes the probability distribution, the probability distribution being defined as a function of at least one hyperparameter is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation at “least one hyperparameter being determined: as a function of training data that include instructions for a measurement at a device and/or the measurement, and at least one instruction or the measurement being determined as a function of a quality measure, “Is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “the quality measure including an expected value for an entropy or a variance that is determined as a function of the probability distribution” is an abstract idea of mathematical relationship, as directed to “a mathematical relationship is a relationship between variables or numbers. A mathematical relationship may be expressed in words or using mathematical symbols”. See MPEP § 2106.04(a)(2)(I)(A).
The limitation “at least one instruction or the measurement being determined as a function of a quality measure” is an abstract idea of a of a mental process that can practically be performed in the human mind, with or without the use of a physical aid such as pen and paper (including an observation, evaluation, judgment, opinion). See MPEP § 2106.04(a)(2)(III).
The limitation “the quality measure including an expected value for an entropy or a variance that is determined as a function of the probability distribution.” is a abstract idea of mathematical concept as, directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Step 2A Prong Two: Does the claim recite additional elements that integrate the judicial exception into a practical application? See MPEP § 2106.04(d).
The limitation “training data that include instructions for a simulation of a measurement that is executable on a device and/or the simulated measurement,” is an additional element that amounts to adding insignificant extra-solution activity of mere data output to the judicial exception. The claim recites producing output in the form of instruction for an activity. See MPEP §§ 2106.04(d), 2106.05(g).
Step 2B Analysis: Does the claim recite additional elements that amount to significantly more than the judicial exception? See MPEP § 2106.05.
The limitation “training data that include instructions for a simulation of a measurement that is executable on a device and/or the simulated measurement,” is an additional element to adding insignificant extra-solution activity to the judicial exception. See MPEP § 2106.05(g). Furthermore, the additional element is directed to a method performed by a computer/or a device which the courts have recognized as well‐understood, routine, and conventional when they are claimed in a generic manner. See MPEP § 2106.05(d)(II).
Regarding Claim 11
Step 2A Prong One Analysis: Does the claim recite an abstract idea, law of nature, or natural phenomenon? See MPEP § 2106.04(II)(A)(1).
The limitation “one iteration for the at least one hyperparameter, an a posteriori distribution over values for the at least one hyperparameter is determined, at least one value of the at least one hyperparameter being determined in another iteration.” is a abstract idea of mathematical concept and mental process directed to “a claim that recites a mathematical calculation, when the claim is given its broadest reasonable interpretation in light of the specification, will be considered as falling within the "mathematical concepts" grouping. A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number”. See MPEP § 2106.04(a)(2)(I)(C).
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
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.
Claims 1, 2, 5, 8 – 13 are rejected under 35 U.S.C. 102(a)(1) and 102 (a)(2) as being anticipated Adams, et al., pre- Grant publication No. US 2014/0358831 (“Adams”).
Regarding claim 1
A computer-implemented method for machine learning, the method comprising the following steps: (Adams, paragraph “[0003] A machine learning system may be configured to use one or more machine learning techniques (e.g., classification techniques, clustering techniques, regression techniques, structured prediction techniques, etc.) and/or models (e.g., statistical models, neural networks, support vector machines, decision trees, graphical models, etc.) for processing data.[ A computer-implemented method for machine learning, the method comprising the following steps]”).
providing a probabilistic model; (Adams, paragraphs “[0011] In some embodiments, including any of the preceding embodiments, the probabilistic model [ providing a Probabilistic model] of the objective function comprises a Gaussian process or a neural network.”).
probability distribution including a Gaussian process or a Bayesian neural Network, (Adams, paragraph, “[0078] As described above, in some embodiments, Bayesian optimization techniques described herein involve generating a probabilistic model of an objective function for a particular task (e.g., an objective function relating hyper-parameters of a machine learning system to its performance). Any suitable type of probabilistic model of the objective function may be used. In some embodiments, the probabilistic model may comprise a Gaussian process, which is a stochastic process that specifies a distribution over functions. [probability distribution including a Gaussian process]”).
the probabilistic model being defined as a function of at least one hyperparameter of the Gaussian process or of the Bayesian neural network;(Adams, paragraphs,0078, “[0078] As described above, in some embodiments, Bayesian optimization techniques described herein involve generating a probabilistic model [the Probabilistic model] of an objective function for a particular task (e.g., an objective function relating hyper-parameters [defined as a function of at least one hyperparameter] of a machine learning system to its performance. Any suitable type of probabilistic model of the objective function may be used. In some embodiments, the probabilistic model may comprise a Gaussian process [of the Gaussian process] which is a stochastic process that specifies a distribution over functions.”).
in one iteration, determining, as a function of the probabilistic model, an instruction for a first measurement, and outputting the instruction for the first measurement; (Adams, paragraph 0075, 0083,0190, “[0075] Accordingly, in some embodiments, optimization using an objective function may be performed iteratively (for one or multiple iterations) by performing, at each iteration [in one iteration] acts of: identifying a point [determining] at which to evaluate the objective function using an acquisition utility function and a probabilistic model [the function of the probabilistic model] of the objective function, evaluating the objective function at the identified point, and updating the probabilistic model based on results of the evaluation.
PNG
media_image1.png
390
795
media_image1.png
Greyscale
Examiner notes: Where y is the output of the instruction for the measurement.
[and outputting the instruction for first measurement] [0190]. Accordingly, each multi-objective function f modeled by a vector-valued Gaussian process maps inputs into T outputs [corresponding to T related tasks with each of the T outputs being an output [instruction for the first measurement] for a corresponding task.”).
for the at least one hyperparameter, determining an a posteriori distribution over values for the at least one hyperparameter as a function of the first measurement
(Adams, paragraph, 0061, 0129,0130,“[0061] (i.e., the objective function maps hyper-parameter values of a machine learning system to respective values providing a measure of performance of the machine learning system)”.[0129], Let f(x) denote the objective function and the set X [at least one hyperparameter as a function of the first measurement] denote the set of points on which the objective function may be calculated.
ψ(y,y*)= 1y< y*
ψ(y,y*)= (y* -y) 1y< y*)”).
[0130] The integrated acquisition utility function may be given according to:
PNG
media_image2.png
184
534
media_image2.png
Greyscale
is the marginal predictive density obtained from the probabilistic model of the objective function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1N) is the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*) corresponds to a selection heuristic.[ for the at least one hyperparameter, determining an a posteriori distribution over values]
Examiner notes: f(x) is the objective function representing the hyperparameters x to be optimized which provides measure of performance to ml system .p (theta|{x, y}) is the posteriori distribution over f(X) which is a function of first measurement.
in another iteration, determining an instruction for a second measurement as a function of the probabilistic model, and outputting the instruction for the second measurement; (Adams, paragraph 0075,0080,”[0075] Accordingly, in some embodiments, optimization using an objective function may be performed iteratively (for one or multiple iterations) by performing, at each iteration [in another iteration] acts of: identifying [ determining ] a point [ an instruction for a second measurement] ]at which to evaluate the objective function using an acquisition utility function and a probabilistic model [as a function of the probabilistic model] of the objective function, evaluating the objective function at the identified point and updating probabilistic model based on results of the evaluation.[080]
PNG
media_image3.png
444
558
media_image3.png
Greyscale
Examiner notes: Determining a instruction maps to identifying a point for second measurement, [0080] details calculating probabilistic model of objective function using gaussian process,..Outputting of the instruction ,of second measurement maps to a Gaussian process...used to calculate an estimate of the objective function [as represented by equations 2 and 3]" of para. 0080. [outputting the instruction of second measurement]
8 determining at least one value of the at least one hyperparameter as a function of the second measurement. (Adams, 0124, 0127,0117, “[0117] second point different from the first point at which to evaluate [determining] the objective function [ at least one value of the at least one hyperparameter] and beginning evaluation of the objective function at the second point. [0061] (i.e., the objective function maps hyper-parameter values of a machine learning system to respective values providing a measure performance of the machine learning system) [0124] The probabilistic model of the objective function may be updated in any of numerous ways based on results of the new evaluation obtained at act 406. updating the probabilistic model of the objective function may comprise calculating an updated measure of uncertainty associated with the updated estimate of the objective function (e.g., calculating the predictive covariance of the probabilistic model based on any previously-obtained evaluations [a function of the second measurement.] of the objective function and results of the evaluation of the objective function at act 406).”).
Regarding claim 12
Adams teaches at least one processor; and at least one memory; wherein the at least one processor is configured to execute computer-readable instructions, (Adams paragraph “0005, fig 4] Some embodiments are directed to at least one non-transitory computer readable storage medium [at least one memory] storing processor executable instructions that, when executed [execute computer-readable instructions] by at least one computer hardware processor, [wherein the at least one processor] cause the at least one computer hardware processor to perform a method.”).
2. one memory being configured to store a model and computer- readable instructions upon whose execution by the at least one processor, the at least one processor performs: providing the model (Adams,0117, 005 “[0117] The probabilistic model of the objective function may be initialized by setting the values for one or more (e.g., all) of the parameters of the probabilistic model. [providing the model] The parameter(s) may be set to any suitable values, which in some instances may be based on any prior information available about the objective function, if any. The parameter values may be stored in memory [one memory being configured to store a model] or on any other suitable type of non-transitory computer-readable medium. [0005] Some embodiments are directed to at least one non-transitory computer readable storage medium storing processor executable instructions that, when executed [execute computer-readable instructions] by at least one computer hardware processor, cause the at least one computer hardware processor to perform a method. [at least one processor performs] The method comprises identifying, using an integrated acquisition utility function and a probabilistic model of the objective Function.”).
And for all other claim elements of claim 12, they are rejected on the same basis as independent claim 1 since the claim are analogous.
Regarding claim 13
Adams teaches A non-transitory computer readable medium on which is stored a computer program including computer-readable instructions for machine learning, the instructions, when executed by a computer, causes the computer to perform (Adams, Paragraph 0005,0117, “[0005] Some embodiments are directed to at least one non-transitory computer-readable storage medium [non-transitory computer readable medium ] storing processor executable instructions [ on which stored a computer program] that, when executed by at least one computer hardware processor,[computer] cause the at least one computer hardware processor to perform [causes the computer to perform] a method for use in connection with performing optimization using an objective function.[0028] In some embodiments, including any of the preceding embodiments, the objective function relates values of hyper-parameters of a machine learning system [ machine learning] to values providing a measure of performance of the machine learning system.”).
And for all other claim elements of claim 13 they are rejected on the same basis as independent claim 1 since the claims are analogous.
Regarding claim 2, Adams teaches the method of claim 1.
Adams further teaches checking whether the a posteriori distribution satisfies a condition, (Adams, paragraph “[0130] The integrated acquisition utility function may be given according to:
PNG
media_image2.png
184
534
media_image2.png
Greyscale
is the marginal predictive density obtained from the probabilistic model of the objective function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1N) is the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*) corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
ψ(y,y*)= 1y< y*
ψ(y,y*)= (y* -y) 1y< y*) “). [checking whether the a posteriori distribution satisfies a condition]
Examiner notes: f(x) is the objective function representing the hyperparameters to be optimized; p(theta|{x, y}) is the posteriori distribution and phi(y, y*) is the constraint that has to be satisfied (i.e., the new y* must be greater than the previous point y).
2. the at least one value for the at least one hyperparameter subsequently being determined(Adams, paragraph 0061,0057, “[0061] (i.e., the objective function maps hyper-parameter values of a machine learning system to respective values providing a measure of performance of the machine learning system)”[0057] For example, as illustrated in FIG. 1, machine learning system 102 may be configured by first manually setting hyper-parameters 104, [at least one value for the at least one hyperparameter] and subsequently learning [subsequently being determined] during training stage 110, the values of parameters 106a, based on training data 108 and hyper-parameters 104, to obtain learned parameter values 106b.”).
3. based on the a posteriori distribution satisfying the condition (Adams, paragraph ,0126,0130 “[0126] When it is determined, at decision block 410, that the objective function is to be evaluated again, process 400 returns, via the YES branch, to act 404, and acts 404-408 are repeated. [0130] The integrated acquisition utility function may be given according to:
PNG
media_image2.png
184
534
media_image2.png
Greyscale
is the marginal predictive density obtained from the probabilistic model of the objective `function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1N) [a] is the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*) corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
`ψ(y,y*)= 1y< y*
ψ(y,y*)= (y* -y) 1y< y*)*
f(x) is the objective function representing the hyperparameters to be optimized; p(theta|{x, y}) is the posteriori distribution and phi(y, y*) is the constraint that has to be satisfied (i.e., the new y* must be greater than the previous point y). [based on the a posteriori distribution satisfying the condition]”).
4. or a further a posteriori distribution over values for the at least one hyperparameter subsequently being determined based on the a posteriori distribution not satisfing the condition. (Adams, paragraph 0129,0126, 0057,0130, “[0129] Let f(x) denote the objective function [over values for the at least one hyperparameter] and the set X denote the set of points on which the objective function may be calculated. Assuming that the objective function has been evaluated N times, we have as input { xn , yn } n=1 N , where each xn represents a point at which the objective function has been evaluated and yn represents the corresponding value of the objective function (i.e., yn = f(xn)). Let p( ) denote the probabilistic model of the objective function. [0126] When it is determined, at decision block 410, that the objective function is to be evaluated again, process 400 returns, via the YES branch, to act 404, and acts 404-408 are repeated. [0057] For example, as illustrated in FIG. 1, machine learning system 102 may be configured by first manually setting hyper-parameters 104, and subsequently learning [subsequently being determined] during training stage 110, the values of parameters 106a, based on training data 108 and hyper-parameters 104, to obtain learned parameter values 106b. [0130] The integrated acquisition utility function may be given according to:
PNG
media_image2.png
184
534
media_image2.png
Greyscale
is the marginal predictive density obtained from the probabilistic model of the objective `function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1N) [a] is the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*) corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
`ψ(y,y*)= 1y< y*
ψ(y,y*)= (y* -y) 1y< y*)
f(x) is the objective function representing the hyperparameters to be optimized; p(theta|{x, y}) is the posteriori distribution [or a further a posteriori distribution] and phi(y, y*) is the constraint that has to be satisfied (i.e., the new y* must be greater than the previous point y). [based on the a posteriori distribution not satisfying the condition]”).
Regarding claim 5, Adams teaches the method of claim 2.
Adams further teaches the condition includes a third criterion (Adams, paragraph “[0023] Identifying, using the updated probabilistic model of the objective function, at least a third point at which to evaluate the objective function; and beginning evaluation [condition] of the objective function at least at the identified third point.[ third criterion].”).
2. that is satisfied when the a posteriori distribution is unimodal (Adams ,0132,136,141, “[0132]
PNG
media_image4.png
391
600
media_image4.png
Greyscale
where p(theta|x, y) is the posteriori distribution. [ that is satisfied when the a posteriori distribution]
[0136] For example, conventional Bayesian optimization techniques utilize stationary Gaussian processes for modeling objective functions. [141] For example, the non-linear one-to-one mapping of points in d-dimensional space on which an objective function is defined (e.g., the space of hyper-parameter values of a machine learning system that has d hyper-parameters) may be specified coordinate-wise as follows:
PNG
media_image5.png
156
567
media_image5.png
Greyscale
here xd is the value of x at its dth coordinate, BetaCDF refers to the cumulative distribution function (CDF) of the Beta random variable, and B(α, βd) is the normalization constant of the Beta CDF. The Beta CDF is parameterized by positive-valued ("shape") parameters αd and βd. It should be appreciated that the non-linear one-one mapping is not limited to comprising the cumulative distribution function of a Beta random variable and may instead comprise the cumulative distribution function of Kumaraswamy random variable, Gamma random variable, Poisson random variable, Binomial random variable, Gaussian random variable, [ is unimodel] or any other suitable random variable.”).
Examiner notes: Gaussian random variable is always unimodal.
3. and it being checked whether the third criterion is satisfied; (Adams 0075,125, [0075]Accordingly, in some embodiments, optimization using an objective function may be performed iteratively (for one or multiple iterations) by performing, at each iteration, acts of: identifying a point at which to evaluate the objective function using an acquisition utility function and a probabilistic model of the objective function, evaluating the objective function at the identified point, and updating the probabilistic model based on results of the evaluation.[125] As one non-limiting example, process 400 may involve performing no more than a threshold number of evaluations of the objective function and when that number of evaluations has been performed, it may be determined [ Being checked]that the objective function is not to be evaluated again. [whether the third criterion that is satisfied].”).
Regarding claim 8 Adams teaches the method of claim 3.
Adams further teaches wherein in at least one iteration preceding the iteration (Adams paragraph 0013, 0036 “[0013] identifying a plurality of points at which to evaluate the objective function; evaluating the objective function at each of the plurality of points; and identifying or approximating, based on results of the evaluating, a point at which the objective function attains a maximum value. [0036] identifying, based at least in part on a joint probabilistic model of the plurality of objective functions, a first point at which to evaluate an objective function in the plurality of objective functions; selecting, based at least in part on the joint probabilistic model, a first objective function in the plurality of objective functions to evaluate at the identified first point; and updating the joint probabilistic model based on results of the evaluation to obtain an updated joint probabilistic model. [wherein in at least one iteration preceding the iteration]”).
a preceding a posteriori distribution is determined, (Adams, paragraph 0132,
PNG
media_image6.png
354
750
media_image6.png
Greyscale
where p(theta|x, y) is the posteriori distribution. [a preceding posteriori distribution is determined]
the a posteriori distribution satisfying the condition (Adams paragraph “[0125] After the probabilistic model of the objective function is updated at act 408, process 400 proceeds to decision block 410, where it is determined whether the objective function is to be evaluated at another point. This determination may be made in any suitable way. [the a posteriori distribution satisfying the condition]”).
when the a posteriori distribution and the at least one preceding a posteriori distribution (Adams paragraph “ [0126] When it is determined, at decision block 410, that the objective function is to be evaluated again, process 400 returns, via the YES branch, to act 404, and acts 404-408 are repeated.
PNG
media_image6.png
354
750
media_image6.png
Greyscale
where p(theta|x, y) is the posteriori distribution [when the a posteriori distribution and the at least one preceding a posteriori distribution]”).
satisfy the condition or the first criterion condition (Adams, Paragraph “[0125] After the probabilistic model of the objective function is updated at act 408, process 400 proceeds to decision block 410, where it is determined whether the objective function is to be evaluated at another point. This determination may be made in any suitable way. [ satisfy the condition or the first criterion condition]”).
Regarding claim 9 Adams teaches the method of claim 2.
Adams further teaches value is determined as a function of a solution of an optimization problem (Adams, paragraph “[0153] The approximation of the integrated acquisition utility function computed via Equation 22[ value is determined as a function] may be used to identify a point which is (or is an approximation of) a point x* at which the integrated acquisition utility function attains its maximum value. [of solution of an optimization]”).
2. a function of the at least one hyperparameter, the value being determined as a function of the solution of the optimization problem that is defined as a function of an objective function that is a function of the at least one hyperparameter. (Adams 148, 149, “[148] As was discussed with reference to process 400, in some embodiments, numerical techniques may be used to identify and/or approximate the point at which the integrated acquisition utility function attains its maximum value.149] Let f(x) denote the objective function and the set X [ a function of the at least one hyperparameter ] denote the set of points on which the objective function [defined as a function of an objective function] may be calculated. Assuming that the objective function has been evaluated N times, we have as input {g(xn; Φ ) yn; for 1 ≤n≤N}, where each xn represents a point at which the objective function has been evaluated, g(xn; Φ,)represents the result of applying a non-linear bijective warping function g, having parameters Φ, to the point xn, and yn represents the corresponding value of the objective function (i.e. yn=f(xn)).2
[0153] The approximation of the integrated acquisition utility function computed via Equation 22 may be used to identify a point [determined as a function] which is (or is an approximation of) a point x* at which the integrated acquisition utility function attains its maximum value. [solution of the optimization problem] This may be done in any suitable way. For example, in some embodiments, the integrated acquisition function may be approximated according to Equation 22 on a grid of points and the point on the grid for which the objective function achieves the maximum value may be taken as the point x*. Alternatively, local exploration (e.g., based on the gradient of the warping function) may be performed around one or more points on the grid to identify the point x*. After the point x* is identified, the objective function may be evaluated at x*[ at least one hyperparameter]”).
posteriori distribution is determined as a function of a sample drawn from a set of values for the at least one hyperparameter.
(Adams, paragraph, 0132,149, “[0132],
PNG
media_image4.png
391
600
media_image4.png
Greyscale
where p(theta|x, y) is the posteriori distribution. [posteriori distribution is determined]
(Adams, Paragraph ,[0149] Let p( ) denote the probabilistic objective function [as a function of samples drawn from a set of values for the at least one hyperparameter ] that depends on a non-linear one-to-one mapping g, the probabilistic model having parameters. Theta)”).
Regarding claim 10 Adams teaches he method of claim 2.
Adams further teaches wherein the probabilistic model includes the probability distribution: (Adams, paragraph “[0086] As may be appreciated from the above examples, in some embodiments, the probabilistic model for an objective function may specify a probability distribution [the probability distribution] on a set of functions (e.g., a set of functions believed to include the objective function or another function that closely approximate the objective function)”).
2. The probability distribution being defined as a function of at least one hyperparameter, the at least one hyperparameter being determined:
(Adams, Paragraph ,0086, 0061, 0057 and fig1, “[0086] As may be appreciated from the above examples, in some embodiments, the probabilistic model for an objective function may specify a probability distribution [the probability distribution] on a set of functions (e.g., a set of functions believed to include the objective function or another function that closely approximate the objective function)[0061] (i.e., the objective function maps hyper-parameter values [being defined as the function at least one value for the at least one hyperparameter] of a machine learning system to respective values providing a measure of performance of the machine learning system) [0057] For example, as illustrated in FIG. 1, machine learning system 102 may be configured by first manually setting hyper-parameters 104, and subsequently learning [the at least one hyperparameter being determined] during training stage 110, the values of parameters 106a, based on training data 108 and hyper-parameters 104, to obtain learned parameter values 106b.”).
3. as a function of training data that include instructions for a measurement at a device.
(Adams, Paragraph ,0182, 0014, 0209, 0030 and fig1,fig 4 “[0182] teaches us In T-fold cross-validation, the data used to train [as a function of training data] a machine learning system is partitioned into T subsets, [0114, fig4] Process 400 may be performed using any suitable computing device(s) [device]comprising one or multiple computer hardware processors, as aspects of the technology described herein are not limited in this respect. [0209] Processor-executable instructions may be in many forms, such as program modules, executed by one or more computers or other devices. [0030] processor-executable instructions further cause the at least one computer hardware processor to perform: identifying a second point at which to evaluate the objective function; [include instructions for a measurement]”).
4. and at least one instruction or the measurement being determined as a function of a quality measure (Adams, Paragraph 0059, 117, 0119, “[0059] The performance of machine learning systems (e.g., the generalization performance) is sensitive to hyper-parameters and manually setting the hyper-parameters [at least one instructions] of a machine learning system to "reasonable" values (i.e., manually tuning the machine learning system), as is conventionally done, may lead to poor or sub-optimal performance of the system. [117] second point different from the first point at which to evaluate the objective function; [measurement] and beginning evaluation of the objective function at the second point. [0119] As described above, the integrated utility function may be obtained by selecting an initial acquisition utility function that depends on one or more parameters of the probabilistic model (e.g., a probability of improvement utility function, expected improvement utility function, regret minimization utility function, entropy-based utility function, etc.), and calculating the integrated utility function by integrating the initial acquisition function with respect to one or more of the probabilistic model parameters.[determined as a function of a quality measure]
5. the quality measure including an expected value for an entropy or a variance that is determined as a function of the probability distribution, (Adams, Paragraph 0087, 0088 0197, [0087] For example, in embodiments where the probabilistic model of an objective function comprises a Gaussian process, the Gaussian process may be updated (e.g., its mean and/or covariance function may be updated) based on the new evaluation(s) of the objective function. As another example, in embodiments where the probabilistic model of an objective function comprises a neural network, the neural network may be updated (e.g., probability distributions associated with the weights of a neural network may be updated) based on the new evaluation(s) of the objective function.[0088]It may be seen from FIG. 2A that the probabilistic model is more uncertain about the objective function in regions where the objective function has not been evaluated and less uncertainty around regions where the objective function has been evaluated (e.g., the region of uncertainty shrinks closer to evaluations 202, 204, and 206). That is, the uncertainty associated with the estimate of the objective function is larger in regions where the objective function has not been evaluated (e.g., the predictive variance [ the quality measure including a variance that is determined as a function of the probability distribution,] of the Gaussian process is larger in regions where the objective function has not been evaluated; the predictive variance is 0 at the points where the objective function has been evaluated since the value of the objective function at those points is known exactly). [0197] In some embodiments, the entropy-search acquisition function may be generalized to take into account the computational cost of evaluating the objective functions in the set of objective functions. The resultant acquisition function a.sub.IG(x), termed a cost-weighted entropy search acquisition utility function may be computed according to
PNG
media_image7.png
291
645
media_image7.png
Greyscale
H(p) represents entropy. [the quality measure including an expected value for an entropy]
6. as a function of training data (Adams, paragraph“[0058] As one non-limiting example, machine learning system 102 may be a machine learning system for object recognition comprising a multi-layer neural network associated with one or more hyper-parameters (e.g., one or more learning rates, one or more dropout rates, one or more weight norms, one or more hidden layer sizes, convolutional kernel size when the neural network is a convolutional neural network, pooling size, etc. [function of training data]”).
7. include instructions for a simulation of a measurement (Adams paragraph “[0113] For example, in some embodiments, Monte Carlo simulation techniques may be used to approximate the integrated acquisition utility function and/or find a point (or an approximation to the point) at which the integrated acquisition utility function attains its maximum. [simulation of a measurement]and sequential Monte Carlo techniques (e.g., particle filters). respect. [0209] Processor-executable instructions may be in many forms, such as program [instructions] modules, executed by one or more computers or other devices.”).
8. is executable on a device and/or the simulated measurement, (Adams paragraph”[0005] Some embodiments are directed to at least one non-transitory computer readable storage medium storing processor executable instructions [executable on a device] that, when executed by at least one computer hardware processor, cause the at least one computer hardware processor to perform a method. Markov chain Monte Carlo techniques (e.g., slice sampling, Gibbs sampling, Metropolis sampling, Metropolis-within-Gibbs sampling, exact sampling, simulated tempering, parallel tempering, annealed sampling, population Monte Carlo sampling, etc.) [ simulated measurement] and sequential Monte Carlo techniques (e.g., particle filters).”).
9. and at least one instruction or the measurement being determined as a function of a quality measure (Adams, Paragraph 0059, 117, 0119, [0059] The performance of machine learning systems (e.g., the generalization performance) is sensitive to hyper-parameters and manually setting the hyper-parameters [instructions] of a machine learning system to "reasonable" values (i.e., manually tuning the machine learning system), as is conventionally done, may lead to poor or sub-optimal performance of the system. [117] second point different from the first point at which to evaluate the objective function; [measurement] and beginning evaluation of the objective function at the second point. [0119] As described above, the integrated utility function may be obtained by selecting an initial acquisition utility function that depends on one or more parameters of the probabilistic model (e.g., a probability of improvement utility function, expected improvement utility function, regret minimization utility function, entropy-based utility function, etc.), and calculating the integrated utility function by integrating the initial acquisition function with respect to one or more of the probabilistic model parameters.[determined as a function of a quality measure]”).
10. the quality measure including an expected value for an entropy or a variance that is determined as a function of the probability distribution (Adams, Paragraph ,0087, 0088, 0197 “[0197] In some embodiments, the entropy-search acquisition function may be generalized to take into account the computational cost of evaluating the objective functions in the set of objective functions. The resultant acquisition function a.sub.IG(x), termed a cost-weighted entropy search acquisition utility function may be computed according to
PNG
media_image7.png
291
645
media_image7.png
Greyscale
H(p) represents entropy. [the quality measure including an expected value for an entropy] [088] It may be seen from FIG. 2A that the probabilistic model is more uncertain about the objective function in regions where the objective function has not been evaluated and less uncertainty around regions where the objective function has been evaluated (e.g., the region of uncertainty shrinks closer to evaluations 202, 204, and 206). That is, the uncertainty associated with the estimate of the objective function is larger in regions where the objective function has not been evaluated (e.g., the predictive variance [ the quality measure including a variance that is determined as a function of the probability distribution,] of the Gaussian process is larger in regions where the objective function has not been evaluated; the predictive variance is 0 at the points where the objective function has been evaluated since the value of the objective function at those points is known exactly.” ).
Regarding Claim 11, Adams teaches the method of claim 1.
Adams further teaches one iteration for the at least one hyperparameter; (Adams, paragraphs 0118, 0125, and fig 4, “ [ 0018] when the objective function relates values of hyperparameters of [at least one hyperparameter] machine learning system to its performance, a set of hyper-parameter values for which to evaluate performance of the machine learning system may be identified at act 404. [0125] it may be determined that the objective function is to be evaluated again.[iteration]”).
2. an a posteriori distribution over values for the at least one hyperparameter is determined, (Adams, paragraph,129,130 “[0129] Let f(x) denote the objective function and the set X denote the set of points on which the objective function may be calculated. Assuming that the objective function has been evaluated N times, we have as input { xn , yn } n=1 N , where each xn represents a point at which the objective function has been evaluated and yn represents the corresponding value of the objective function (i.e., yn = f(xn)). Let p( ) denote the probabilistic model of the objective function.[0130] marginal predictive density obtained from the probabilistic model of the objective function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1Nis the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*) corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
ψ(y,y*)= 1y< y*
ψ(y,y*)= (y* -y) 1y< y*) [an a posteriori distribution over values for the at least one hyperparameter is determined] ”).
Examiner notes : f(x) is the objective function representing the hyperparameters x to be optimized; p(theta|{x, y}) is the posteriori distribution over the values of hyperparameter.
3. at least one value of the at least one hyperparameter being determined in another iteration. [ Adams, paragraph 0124 , [0124] After the objective function is evaluated, at act 406, at the point identified at act 408, process 400 proceeds to act 408, where the probabilistic model of the objective function is updated based on results of the evaluation.[being determined in another iteration [ 0123] After the point at which to evaluate the objective function is identified at act 404, process 400 proceeds to act 406, where the objective function is evaluated at the identified point. For example, when the objective function relates hyper-parameter [ at least value of the at least one hyperparameter] values of a machine learning system to its performance, performance of the machine learning system configured with the hyper-parameters identified at act 404 may be evaluated at act 406. Also given in fig 4. 404,406 and 410 shows iterative process.
Examiner Remarks: Also, fig 4, ref num 410 evaluates and checks the condition, [0118] if satisfied loops back to 404 which shows there is at least one iteration for objective function. objective function relates to the value of hyperparameter.
Claim Rejections – 35 USC § 103
The following is a quotation of 35 U.S.C. 103, which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
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 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.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Adams in view of liu, et al., “T. Simulation-efficient shortest probability intervals. Stat Comput 25, 809–819 (2015) (“liu”)
Regarding claim 3, Adams teaches the method of claim 2.
Adams further teaches,
1. a posteriori distribution assigns values their probability measure, (Adams, paragraph 0150, 0153, “[0150] Initially, for each for 1 ≤ j ≤ J, draw a sample (θ(j) ,Φ(j)) according to: (θ(j), Φ(j)) ~ p(θ, Φ|{g(xn; Φ ) yn; for 1 ≤ n≤ N}),
where p(theta|x, y) is the posteriori distribution.[ a posteriori distribution] and θ, Φ are value of probability measure. [assigns values their probability measure]”).
2. the condition including a first criterion that is satisfied (Adams,[153] integrated acquisition utility function computed via Equation 22 may be used to identify a point which is (or is an approximation of) a point x* at which the integrated acquisition utility function attains its maximum value. [the condition including a first criterion that is satisfied].
3. that is defined as a function of a largest probability measure of the distribution (Adams,”[144] Accordingly, in some embodiments, a non-linear warping may be inferred based, at least in part, one or more evaluations of the objective function (e.g., the maximum a posteriori estimate of the parameters [ that is defined as a function of a largest probability measure of the distribution] of the non-linear warping given results of all evaluations may be used to determine the non-linear warping) and the probabilistic model of the objective function may be specified by using the non-linear warping. “).
Adams does not explicitly teach:
1. when more than a specified percentage of probability measures of the distribution lie within an interval.
2. and includes the measure.
liu teaches more than a specified percentage of probability measures of the distribution lie within an interval. (liu pg-1 introduction, “Bayesian inferences via posterior intervals of specified coverage (for example, 50 % and 95 %) [more than a specified percentage of] for parameters [ of probability measures]and other quantities of interest. In the modern era in which posterior distributions are computed via simulation, [of the distribution] we most commonly see central intervals: The 100(1- α)% central interval [lie within an interval] is defined by the α/2 and 1- ( α/2) quantiles.) “).
2. and includes the measure. liu, pg 2 , HPD covers the highest density part [and includes the measure] of the distribution and also the mode.
Adams and liu are both related to the same field of endeavor (i.e. posterior distribution analysis). In view of the teachings of Adams it would have been obvious for a person of ordinary skill in the art to apply the teachings of liu i.e HPD( Highest Probability density techniques) to Adams i.e posterior distribution framework before the effective filing date of the claimed invention in order to determine posterior probability and it’s shortest interval to improve probabilistic evaluation with in Bayesian optimization systems. (liu, pg-810 “The goal is to estimate the 100(1−α) % shortest probability interval for F.”).
Claim 4 and 6 are rejected under 35 U.S.C. 103 as being unpatentable over Adams, et al., in view of Blei, et al. (“Variational Inference: A Review for Statisticians”(05/2019).
Regarding claim 4, Adams teaches the method of claim 2.
Adams further teaches wherein the condition includes a second criterion. (Adams, paragraph 0022 ,153, [153] integrated acquisition utility function computed via Equation 22 may be used to identify a point which is (or is an approximation of) a point x* at which the integrated acquisition utility function attains [ wherein the condtion] its maximum value. [022] updating the probabilistic model of the objective function using results of evaluating the objective function at the first point and/or the second point [ includes second criterion] to obtain an updated probabilistic model of the objective function
2. Between posteriori distribution and a Gaussian distribution (Adams paragraph “[0011] In some embodiments, including any of the preceding embodiments, the probabilistic model of the objective function comprises a Gaussian process or a neural network. [0150] Initially, for each for 1 ≤ j ≤ J, draw a sample (θ(j) ,Φ(j)) according to: (θ(j), Φ(j))
p(θ, Φ|{g(xn; Φ ) yn; for 1 ≤ n≤ N}), [Between posteriori distribution and a Gaussian distribution] “).
Examiner remarks “teach posterior distribution with gaussian probabilistic model.
3. and it being checked whether the second criterion is satisfied. (Adams [0130] The integrated acquisition utility function may be given according to:
PNG
media_image8.png
276
789
media_image8.png
Greyscale
is the marginal predictive density obtained from the probabilistic model of the objective function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1N) is the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*)
corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
ψ(y,y*) = 1y< y*
ψ(y,y*)= (y* -y) 1y< y*) [and it being checked whether the second criterion is satisfied]
Adams does not explicitly teach:
1. that is satisfied when a distance including a Kullback-Leibler divergence , is smaller than a first threshold,
Blei teaches that is satisfied when a distance including a Kullback-Leibler divergence.
(Blei, “pg 1 abs, “[pg2] Then, we try to find the member of that family that minimizes the Kullback-Leibler (KL) divergence to the exact posterior. [that is satisfied when a distance including a Kullback-Leibler divergence.] [Pg1, abs] we review variational inference (VI), a method from machine learning that approximates probability densities through optimization. VI has been used in many applications and tends to be faster than classical methods, such as Markov chain Monte Carlo sampling. The idea behind VI is to first posit a family of densities and then to find the member of that family which is close to the target [is smaller than a first threshold]. Closeness is measured by Kullback-Leibler divergence. “).
Adams and Blei are related to the same field of endeavor i.e Bayesian Optimization and posterior distribution approximation. In view of the teachings of Adams ,it would have been obvious for a person of ordinary skill in the art to apply the teachings of Blei before the effective filing date of claimed invention to calculate distance between two different distribution in iteration and in order to measure the uncertainty and improve approximation accuracy in an optimization process of distribution. (Blei, pg-1, “a method from machine learning that approximates probability densities through optimization”.)
Regarding claim 6, Adams teaches the method of claim 2.
Adams further teaches, a preceding a posteriori distribution is determined, (Adams, paragraph 132,110,0063, [132]
PNG
media_image4.png
391
600
media_image4.png
Greyscale
where p(theta|x, y) is the posteriori distribution. [a preceding posteriori distribution is determined]
2. in each of a plurality of iterations preceding the iteration (Adams 0013, 0036 [0013] identifying a plurality of points at which to evaluate the objective function; evaluating the objective function at each of the plurality of points; and identifying or approximating, based on results of the evaluating, a point at which the objective function attains a maximum value. [0036] identifying, based at least in part on a joint probabilistic model of the plurality of objective functions, a first point at which to evaluate an objective function in the plurality of objective functions; selecting, based at least in part on the joint probabilistic model, a first objective function in the plurality of objective functions to evaluate at the identified first point; and updating the joint probabilistic model based on results of the evaluation to obtain an updated joint probabilistic model. [in each of a plurality of iterations preceding the iteration]
3. the condition including a fourth criterion (Adams, paragraph [0126], process 400 proceeds to decision block 410, where it is determined whether the objective function is to be evaluated at another point. This determination may be made in any suitable way. As one non-limiting example, process 400 may involve performing no more than a threshold number of evaluations of the objective function and when that number of evaluations has been performed, it may be determined that the objective function is not to be evaluated again (e.g., due to time and/or computational cost of performing such an evaluation). [the condition including a fourth criterion]
between a preceding a posteriori distribution and the a posteriori distribution (Adams [0011] “In some embodiments, including any of the preceding embodiments, the probabilistic model of the objective function comprises a Gaussian process or a neural network.[0150] Initially, for each for 1 ≤ j ≤ J, draw a sample (θ(j) ,Φ(j)) according to: (θ(j), Φ(j)) ~ p(θ, Φ|{g(xn; Φ ) yn; for 1 ≤ n≤ N}),
Examiner remarks teach posterior distribution in iterative process. [between a preceding a posteriori distribution and the a posteriori distribution]
5. and it being checked whether the fourth criterion is satisfied. (Adams [0130] The integrated acquisition utility function may be given according to:
PNG
media_image2.png
184
534
media_image2.png
Greyscale
is the marginal predictive density obtained from the probabilistic model of the objective function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1N) is the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*) corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
ψ(y,y*) = 1y< y*
ψ(y,y*)= (y* -y) 1y< y*) [and it being checked whether the fourth criterion is satisfied]
Adams does not explicitly teach:
1. that is satisfied when a difference including a Kullback-Leibler divergence, is smaller than a second threshold,
Blei teaches that is satisfied when a distance including a Kullback-Leibler divergence.
(David M. Blei, pg 2) Then, we try to find the member of that family that minimizes the Kullback-Leibler (KL) divergence to the exact posterior. [that is satisfied when a distance including a Kullback-Leibler divergence]. [Pg1, abs] we review variational inference (VI), a method from machine learning that approximates probability densities through optimization. VI has been used in many applications and tends to be faster than classical methods, such as Markov chain Monte Carlo sampling. The idea behind VI is to first posit a family of densities and then to find the member of that family which is close to the target [is smaller than a first threshold]. Closeness is measured by Kullback-Leibler divergence.
Adams, and Blei are related to the same field of endeavor, (i.e. probabilistic models). In view of the teachings of Adams it would have been obvious for a person of ordinary skill in the art to apply the teachings of Adams to Blei before the effective filing date of the claimed invention in order to determine the posterior distribution in iteration and calculate the KL divergence in order to measure the uncertainty and improve approximation accuracy in an optimization process of distribution. (Blei, pg-1, “a method from machine learning that approximates probability densities through optimization”.)
Claim 7, is rejected under 35 U.S.C. 103 as being unpatentable over Adams, et al., in view of Kartik et. al. patent No. US 11060885 B2 (Kartik).
Regarding Claim 7, Adams teaches the method of claim 2.
Adams further teaches a characteristic, including an entropy or a variance, (Adams, paragraph 106,197,[106] The entropy search acquisition utility function aims to select the next point at which to evaluate the objective function so at to decrease the uncertainty as to the location of the minimum of the objective function (or, equivalently, as to the location of the maximum of the objective function multiplied by negative one). To this end, the next point at which to evaluate the objective function is selected by iteratively evaluating points that will decrease the entropy of the probability distribution over the minimum of the objective function. [197]
PNG
media_image7.png
291
645
media_image7.png
Greyscale
H(p) represents entropy. [a characteristic, including an entropy] [0084] Regardless of the type of probabilistic model used for modeling the objective function, the probabilistic model may be used to obtain an estimate of the objective function and a measure of uncertainty associated with the estimate. For example, when the objective function relates values of hyper-parameters of a machine learning system to its performance, the estimate of the objective function obtained based on the probabilistic model may provide an estimate of the performance of the machine learning system for each set of hyper-parameter values and the measure of uncertainty associated with the estimate may provide a measure of uncertainty (e.g., a variance, a confidence, etc.) [a variance]
2. a posteriori distribution is determined (Adams, paragraph 0132,
PNG
media_image4.png
391
600
media_image4.png
Greyscale
where p(theta|x, y) is the posteriori distribution. [posteriori distribution is determined]
3. the condition including a fifth criterion (Adams, paragraph [0126], process 400 proceeds to decision block 410, where it is determined whether the objective function is to be evaluated at another point. This determination may be made in any suitable way. As one non-limiting example, process 400 may involve performing no more than a threshold number of evaluations of the objective function and when that number of evaluations has been performed, it may be determined that the objective function is not to be evaluated again (e.g., due to time and/or computational cost of performing such an evaluation). [the condition including a fifth criterion]
4. and it is being checked whether the fifth criterion is satisfied. Adams [0130] The integrated acquisition utility function may be given according to:
PNG
media_image2.png
184
534
media_image2.png
Greyscale
is the marginal predictive density obtained from the probabilistic model of the objective function given {xn, yn}n=1N and parameters θ of the probabilistic model, p(θ|{xn, yn}n=1N) is the likelihood of the probabilistic model given {xn, yn}n=1N, and where ψ (y, y*) corresponds to a selection heuristic. For example, the probability of improvement and expected improvement heuristics may be represented, respectively, according to:
ψ(y,y*) = 1y< y*
ψ(y,y*)= (y* -y) 1y< y*) [and it is being checked whether the fifth criterion is satisfied]
Adams does not explicitly teach:
that is satisfied when the characteristic is smaller than a third threshold
Kartik teaches that is satisfied when the characteristic smaller than a third threshold (Kartik [54] As an example of the functionality of FIG. 4, if the evaluation dataset has 500 points, the input is divided into 25 segments that include 20 points each. Then each segment is scored and each score returns a KL divergence value. From the list of 25 KL divergence values, if 90% of values (i.e., 23 in this example) has an identified divergence less than the calculated “Threshold, [ that is satisfied when the characteristic smaller than a second threshold,] the minimum window size is considered to size is considered to be 20. If it does not satisfy the 90% criteria, the number of points is doubled for each segment. Therefore, in this example, the 500 points are divided into 13 segments approximately with 40 points each and the same evaluation step is repeated.
Adams and Kartik are both related to the same field of endeavor (i.e. probabilistic models). In view of the teachings of Adams it would have been obvious for a person of ordinary skill in the art to apply the teachings of Adams to kartik before the effective filing date of the claimed invention in order to iteratively check uncertainty or spread of posterior distribution with the small threshold and to improve precision of data in optimization process. ( Kartik et. al. paragraph [45] “(i.e., the number of points used to determine the threshold) to reduce or eliminate false positives, and if the number of points is below the minimum, embodiments add “clean” points to the input points to improve anomaly detection.)”.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NIROJ KOIRALA whose telephone number is (571)270-0748. The examiner can normally be reached Monday -Friday 8am-5pm.
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, MICHAEL HUNTLEY can be reached on (303) 297-4307. 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.
/N.K./Examiner, Art Unit 2129
/MICHAEL J HUNTLEY/Supervisory Patent Examiner, Art Unit 2129
1 Appropriate correction is required." claims are examined as best understood on the 112(b) issues mentioned above.
2 Where x corresponds to hyperparameter value used in optimization and f(x) denote objective function.
Prosecution Insights