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 .
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.
Claim 2-3, 11, 16-17 rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claim 2 recites the limitation "the initial candidate hyperparameter" in line 2 of the claim. There is insufficient antecedent basis for this limitation in the claim. For the purpose of examination, the term will be interpreted as referring to “the initial hyperparameter” recited in claim 1.
Claim 3 recites the limitation "the initial candidate hyperparameter" in line 2 of the claim. There is insufficient antecedent basis for this limitation in the claim. For the purpose of examination, the term will be interpreted as referring to “the initial hyperparameter” recited in claim 1.
Claim 11 recites the limitation "the other proxy model" in line 3 of the claim. There is insufficient antecedent basis for this limitation in the claim. For the purpose of examination, the term will be interpreted as referring to “another trained second proxy model” recited in claim 10.
Claim 16 recites the limitation "the initial candidate hyperparameter" in line 8 of the claim. There is insufficient antecedent basis for this limitation in the claim. For the purpose of examination, the term will be interpreted as referring to “the initial hyperparameter” recited in claim 1.
Claim 17 recites the limitation "the respective initial hyperparameter" in line 1 of the claim. There is insufficient antecedent basis for this limitation in the claim. For the purpose of examination, the term will be interpreted as referring to “the initial hyperparameter” recited 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.
Claims 1-20 rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
MPEP 2106 (III) sets out steps for evaluating whether a claim is drawn to patent-eligible subject matter. The analysis of claims 1-20, in accordance with these steps, follows.
Step 1 Analysis:
Claims 1-18 are directed to method (processes). Claim 19 is directed to a computer system (machine). Claim 20 is directed to a computer program product (article of manufacture). Therefore, claims 1- 20 fall into one of four statutory categories (i.e., process, machine, article of manufacture).
As to claim 1,
Step 2A Prong 1: this claim recites the following abstract ideas:
determining an initial hyperparameter; (the limitation describes selecting a parameter value, which is an evaluation and judgment activity that can be performed as a mental process in the human mind.)
calculating a first matching score based on the trained first evaluation model and the trained first proxy model; (the limitation describes calculating score based on results which is a mental process implemented using a pen and paper)
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
Receiving a target dataset; (this limitation describes data collection/receiving, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i).)
training a first proxy model based on the target dataset and the initial hyperparameter, resulting in a trained first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEР 2106.05(d)(II)(i))
sampling a first synthetic dataset based on the trained first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEР 2106.05(d)(II)(i))
training a first evaluation model based on the first synthetic dataset and a first candidate hyperparameter, resulting in a trained first evaluation model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEР 2106.05(d)(II)(i))
training a second proxy model based on the target dataset, the first candidate hyperparameter and the first matching score, resulting in a trained second proxy model; and (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEР 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 2,
Step 2A Prong 1: this claim recites the following abstract ideas:
calculating a further first matching score based on the trained further first evaluation model and the trained further first proxy model; (the limitation describes calculating a score based on results which is a mental process implemented using a pen and paper)
computing a first aggregate score for the first candidate hyperparameter based on the first matching score and the further first matching score, wherein the second proxy model is trained based on the first aggregate score; (the limitation describes aggregating numerical scores to produce a combined score, which is a mathematical calculation that can be performed as a mental process using a pen and paper)
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
training a further first proxy model based on the target dataset and the initial candidate hyperparameter, resulting in a trained further first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
sampling a further first synthetic dataset based on the trained further first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a further first evaluation model based on the first candidate hyperparameter and the further first synthetic dataset, resulting in a trained further first evaluation model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 3,
Step 2A Prong 1: this claim recites the following abstract ideas:
wherein the first proxy model is trained based on the target dataset, the initial candidate hyperparameter, and a first random seed, wherein the further first proxy model is trained based on the target dataset, the initial candidate hyperparameter, and a further first random seed; (the limitation describes selecting and using random seed values as inputs to the training process, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
As to claim 4,
Step 2A Prong 1: this claim recites the following abstract ideas:
calculating a second matching score based on the trained second evaluation model and the trained second proxy model; (the limitation describes calculating a score based on results which is a mental process implemented using a pen and paper)
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
sampling a second synthetic dataset based on the trained second proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a second evaluation model based on the second synthetic dataset and a second candidate hyperparameter, resulting in a trained second evaluation model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a third proxy model based on the target dataset, the second candidate hyperparameter and the second matching score, resulting in a trained third proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 5,
Step 2A Prong 1: this claim recites the following abstract ideas:
calculating a third matching score based on the trained third evaluation model and the trained third proxy model; (the limitation describes calculating a score based on results which is a mental process implemented using a pen and paper)
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
sampling a third synthetic dataset based on the trained third proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a third evaluation model based on the third synthetic dataset and a third candidate hyperparameter, resulting in a trained third evaluation model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a fourth proxy model based on the target dataset, the third candidate hyperparameter and the third matching score, resulting in a trained fourth proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 6,
Step 2A Prong 1: This claim does not recite an additional abstract idea, but the claim depends on claim 1.
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
wherein the first proxy model, the second proxy model and the first evaluation model each have a generative neural network architecture; (the limitation describes using a generic type of neural network architecture to implement the models, which is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 7,
Step 2A Prong 1: this claim recites the following abstract ideas:
determining using the trained first proxy model a first predicted target dataset property; (the limitation describes determining a property value based on model output, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
determining using the trained first evaluation model a first predicted synthetic dataset property, wherein the first matching score is computed between the first predicted target dataset property and the first synthetic dataset property; (the limitation describes determining a property value and computing a comparison score between two properties, which is a mathematical calculation that can be performed as a mental process using a pen and paper. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
wherein the first synthetic dataset is sampled based on the first predicted target dataset property; (this limitation describes data collection/receiving, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i).)
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 8,
Step 2A Prong 1: this claim recites the following abstract ideas:
wherein the first predicted target dataset property comprises a first predicted target dataset causal property, wherein the first predicted synthetic dataset property comprises a first predicted synthetic dataset causal property; (the limitation further describes the type of property being determined as a causal property, which narrows the data being evaluated but remains an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
As to claim 9,
Step 2A Prong 1: This claim does not recite an additional abstract idea, but the claim depends on claim 8.
wherein the first predicted target dataset causal property and the first predicted synthetic dataset causal property are embodied in respective causal graphs sampled from the trained first proxy model and the trained first evaluation model respectively. (the limitation describes using a generic type of neural network architecture to implement the models, which is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 10,
Step 2A Prong 1: this claim recites the following abstract ideas:
Step 2A Prong 1: This claim does not recite an additional abstract idea, but the claim depends on claim 1.
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
wherein the first proxy model and the second proxy model are causal models; (the limitation describes the type of models as causal models, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
sampling a causal graph from the trained second proxy model or another trained proxy model derived from the trained second proxy model via additional sampling and training operations; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 11,
Step 2A Prong 1: this claim recites the following abstract ideas:
determining a treatment action based on the causal graph sampled from the trained second proxy model or the other proxy model; (the limitation describes making a decision based on graph data, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
performing the treatment action on a physical or logical system; (this limitation describes applying the result of the abstract idea to a system recited at a high level of generality, which amounts to insignificant post-solution activity that is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i). The claim does not specify any particular physical system, any particular treatment action, or any particular manner of performing the action, and therefore does not impose meaningful limits on the abstract idea.)
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 12,
Step 2A Prong 1: this claim recites the following abstract ideas:
selecting the first candidate hyperparameter from the respective first candidate hyperparameters based on respective first matching scores computed based on the multiple trained first evaluation models and the trained first proxy model; (the limitation describes selecting a parameter based on comparing scores, which is an evaluation and judgment activity that can be performed as a mental process in the human mind.)
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
wherein the first evaluation model is one of multiple first evaluation models trained based on respective first candidate hyperparameters, resulting in multiple trained first evaluation models; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
fitting the second proxy model to the target dataset using the first candidate hyperparameter as selected based on respective first matching scores, resulting in the trained second proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 13,
Step 2A Prong 1: this claim recites the following abstract ideas:
wherein selecting the first candidate hyperparameter comprises selecting a subset of the respective first candidate hyperparameters based on the respective first matching scores, the subset comprising the first candidate hyperparameter and an additional first candidate hyperparameter; (the limitation describes selecting a subset of parameters based on comparing scores, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
determining a first likelihood of the trained first proxy model with respect to the target dataset; (the limitation describes determining a likelihood value, which is a mental process that can be performed using a pen and paper. See MPEP 2106.04(a)(2))
determining a second likelihood of the trained second proxy model with respect to the target dataset; (the limitation describes determining a likelihood value, which is a mental process that can be performed using a pen and paper. See MPEP 2106.04(a)(2))
determining an additional second likelihood of the trained additional second proxy model with respect to the target dataset; (the limitation describes determining a likelihood value, which is a mental process that can be performed using a pen and paper. See MPEP 2106.04(a)(2))
selecting the trained second proxy model from a set comprising the trained second proxy model and the trained additional second proxy model based on the first likelihood, the second likelihood and the additional second likelihood; (the limitation describes selecting a model based on comparing likelihood values, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
fitting an additional second proxy model to the target dataset using the additional first candidate hyperparameter, resulting in a trained additional second proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 14,
Step 2A Prong 1: this claim recites the following abstract ideas:
wherein the respective first candidate hyperparameters are generated randomly; (the limitation describes generating parameter values randomly, which is an evaluation and judgment activity that can be performed as a mental process in the human mind, such as by drawing numbers from a hat. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
As to claim 15,
Step 2A Prong 1: This claim does not recite an additional abstract idea, but the claim depends on claim 12.
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
wherein the respective first candidate hyperparameters are generated via a Bayesian search based on the respective first matching scores. (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 16,
Step 2A Prong 1: this claim recites the following abstract ideas:
computing an initial matching score based on each trained initial evaluation model and the initial proxy model; (the limitation describes calculating a score based on results which is a mental process implemented using a pen and paper. See MPEP 2106.04(a)(2))
selecting the initial candidate hyperparameter from the respective initial hyperparameters based on the initial matching score computed for each trained initial evaluation model; (the limitation describes selecting a parameter based on comparing scores, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
generating an initial proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
sampling an initial synthetic dataset based on the initial proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training multiple initial evaluation models based on the initial synthetic dataset and respective initial hyperparameters; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional element does not integrate the judicial exception into practical application and does not amount to significantly more than the Judicial exception.
As to claim 17,
Step 2A Prong 1: this claim recites the following abstract ideas:
wherein the respective initial hyperparameters are randomly generated; (the limitation describes generating parameter values randomly, which is an evaluation and judgment activity that can be performed as a mental process in the human mind, such as by drawing numbers from a hat.)
or determined via tuning of a likelihood of the target dataset with respect to an initial proxy model; (the limitation describes determining and optimizing a likelihood value with respect to data and a model, which is an evaluation and judgment activity that can be performed as a mental process in the human mind)
Step 2A Prong 2 and 2B:
The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception under step 2B. Thus, the judicial exception is not integrated into a practical application (see MPEP 2106.04(d) I.), failing step 2A prong 2. The claim is ineligible.
As to claim 18,
Step 2A Prong 1: This claim does not recite an additional abstract idea, but the claim depends on claim 1.
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
comprising, based on the trained second proxy model: tuning, adapting, modifying or replacing an industrial machine; (this limitation describes applying the result of the abstract idea to a machine recited at a high level of generality, which amounts to insignificant post-solution activity that is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i). The claim does not specify any particular industrial machine, any particular manner of tuning, adapting, modifying, or replacing, and therefore does not impose meaningful limits on the abstract idea.)
performing a maintenance or repair action performed on a machine; (this limitation describes applying the result of the abstract idea to a machine recited at a high level of generality, which amounts to insignificant post-solution activity that is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
generating image data, audio data, text or other content; (this limitation describes generating data output, which amounts to insignificant post-solution activity that is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
performing a security mitigation action; (this limitation describes applying the result of the abstract idea at a high level of generality, which amounts to insignificant post-solution activity that is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional elements do not integrate the judicial exception into practical application and do not amount to significantly more than the Judicial exception.
As to claim 19,
Step 2A Prong 1: this claim recites the following abstract ideas:
determining an initial hyperparameter; (the limitation describes selecting a parameter value, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
calculating a first matching score based on the trained first evaluation model and the trained first proxy model; (the limitation describes calculating a score based on results which is a mental process implemented using a pen and paper. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
at least one memory configured to store computer-readable instructions; and at least one hardware processor coupled to the at least one memory; (this limitation describes generic computer components recited at a high level of generality, which amounts to mere instruction to apply the abstract idea on a generic computer, see MPEP 2106.05(d)(II)(i))
receiving a target dataset; (this limitation describes data collection/receiving, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a first proxy model based on the target dataset and the initial hyperparameter, resulting in a trained first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
sampling a first synthetic dataset based on the trained first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a first evaluation model based on the first synthetic dataset and a first candidate hyperparameter, resulting in a trained first evaluation model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a second proxy model based on the target dataset, the first candidate hyperparameter and the first matching score, resulting in a trained second proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional elements do not integrate the judicial exception into practical application and do not amount to significantly more than the Judicial exception.
As to claim 20,
Step 1 Analysis:
Step 2A Prong 1: this claim recites the following abstract ideas:
determining an initial hyperparameter; (the limitation describes selecting a parameter value, which is an evaluation and judgment activity that can be performed as a mental process in the human mind. See MPEP 2106.04(a)(2))
calculating a first matching score based on the trained first evaluation model and the trained first proxy model; (the limitation describes calculating a score based on results which is a mental process implemented using a pen and paper. See MPEP 2106.04(a)(2))
Step 2A Prong 2 and 2B: the claim recited the following additional elements:
Computer-readable storage media embodying computer readable instructions, the computer-readable instructions configured upon execution on at least one hardware processor to cause the at least one hardware processor to perform operations; (this limitation describes generic computer components recited at a high level of generality, which amounts to mere instruction to apply the abstract idea on a generic computer, see MPEP 2106.05(d)(II)(i))
receiving a target dataset; (this limitation describes data collection/receiving, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a first proxy model based on the target dataset and the initial hyperparameter, resulting in a trained first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
sampling a first synthetic dataset based on the trained first proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a first evaluation model based on the first synthetic dataset and a first candidate hyperparameter, resulting in a trained first evaluation model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
training a second proxy model based on the target dataset, the first candidate hyperparameter and the first matching score, resulting in a trained second proxy model; (This limitation is directed to mere instruction to apply the abstract idea on a generic computer to process, which is a well-understood, routine, conventional activity, see MPEP 2106.05(d)(II)(i))
The additional elements do not integrate the judicial exception into practical application and do not amount to significantly more than the Judicial exception.
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.
Claim(s) 1-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Geffner et al. (Deep End-to-end Causal Inference, June 2022) in view of Watson et al. (US 20200012891 A1) and Moharrer et al. (US 20200380378 A1) and further in view of Tee et al. (US 20190156229 A1).
As to claim 1, Geffner teaches a computer-implemented method, comprising:
receiving a target dataset; see Geffner section [3.1] (“We aim to fit θ, the parameters of our non-linear ANM, using observational data.”);
training a first proxy model based on the target dataset and the initial hyperparameter, resulting in a trained first proxy model; see Geffner section [3.1] (“We choose qφ(G) to be the product of independent Bernoulli distributions for each potential directed edge in G. We parametrize edge existence and edge orientation separately, using the ENCO parametrization [36]. The SEM parameters θ and variational parameters φ are trained by maximizing the ELBO.”);
sampling a first synthetic dataset based on the trained first proxy model; see Geffner section [3.1] (“After training, we can use the model learnt by DECI to simulate new samples x from pθ(x|G).”); and
training a first evaluation model based on the first synthetic dataset and a first candidate hyperparameter, resulting in a trained first evaluation model; see Geffner section [B.2] (“For CATE estimation, we need to train a separate surrogate predictor per graph samples.”).
Geffner does not explicitly teach “calculating a first matching score based on the trained first evaluation model and the trained first proxy model.”
However, Watson teaches calculating a first matching score based on the trained first evaluation model and the trained first proxy model; see Watson paragraph [0029] (“The results of the application of separate evaluation data 115 to synthetic data model and original data model can be compared to one another using various exemplary comparison procedures. For example, as discussed below, an integrated variants analysis procedure, or an analysis of variance (“ANOVA”) procedure”).
Geffner-Watson do not explicitly teach “training a second proxy model based on the target dataset, the first candidate hyperparameter and the first matching score, resulting in a trained second proxy model.”
However, Moharrer teaches training a second proxy model based on the target dataset, the first candidate hyperparameter and the first matching score, resulting in a trained second proxy model; see Moharrer paragraph [0025] (“a regressor is trained to predict an estimated quality score based on a given dataset and a given hyperparameters configuration.”).
Geffner-Watson-Moharrer do not explicitly teach “determining an initial hyperparameter.”
However, Tee teaches determining an initial hyperparameter; see Tee paragraph [0066] (“The hyperparameters may be a hyperparameter of a machine learning model, or any type of model. Accordingly, a hyperparameter, as referred to herein, generally relates to a variable that can affect a single optimization trial of an optimization work request.”).
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the invention to modify the invention of Geffner to calculate a first matching score based on the trained first evaluation model and the trained first proxy model, as taught by Watson, and to train a second proxy model based on the target dataset, the first candidate hyperparameter and the first matching score, as taught by Moharrer and Tee, because Geffner teaches a deep learning-based end-to-end causal inference framework in which observational data is used to learn causal relationships and a trained generative model is used to simulate new samples for treatment-effect estimation (see Geffner sections 1, 3, 3.1 and 3.3); Watson teaches that generating a synthetic dataset that can successfully be used to train a model is difficult and that many synthetic datasets are not suitable for model training, and therefore teaches evaluating a synthetic dataset by training models using original and synthetic datasets and comparing results from the training of the models, including by ANOVA, threshold/error comparison, or statistical correlation procedures (see Watson paragraphs [0005]-[0008]); Moharrer teaches that hyperparameter optimization is essential because default hyperparameters often do not result in best-performing models, but also teaches that tuning requires repeatedly training and validating models with different hyperparameter values, which is time-consuming, resource intensive, exponentially hard, and slow, and therefore teaches training a regressor to predict an estimated quality score based on a given dataset and a given hyperparameter configuration so that promising hyperparameter configurations may be selected for actual training (see Moharrer paragraphs [0002]-[0007], [0024]-[0025], and [0042]-[0047]); and Tee teaches that poorly optimized hyperparameters adversely affect prediction accuracy and computational speed, increase required processing and storage workload, and that an intelligent optimization platform can tune hyperparameters with fewer evaluations to save computational resources while improving model performance (see Tee paragraphs [0034]-[0039]). Thus, because Geffner already generates synthetic data from a trained proxy model, Watson provides the reason to evaluate such synthetic data by comparing model-training results, and Moharrer and Tee provide the reason to use quality scores and tuned hyperparameters to improve subsequent model training while reducing computational burden, the combination of Geffner, Watson, Moharrer, and Tee teaches training a first proxy model based on a target dataset and an initial hyperparameter, sampling a first synthetic dataset from the trained first proxy model, training a first evaluation model using the first synthetic dataset and a first candidate hyperparameter, calculating a first matching score by comparing the trained first evaluation model and the trained first proxy model, and training a second proxy model based on the target dataset, the first candidate hyperparameter, and the first matching score.
As to claim 2, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, comprising:
training a further first proxy model based on the target dataset and the initial candidate hyperparameter, resulting in a trained further first proxy model; (see Geffner section [3.1] “We consider two possible models for the distribution of z. 1) A simple Gaussian … 2) A flow”)
sampling a further first synthetic dataset based on the trained further first proxy model; (see Geffner section [3.3] “ATE. After training, we can use the model learnt by DECI to simulate new samples x from p(xjG).”)
training a further first evaluation model based on the first candidate hyperparameter and the further first synthetic dataset, resulting in a trained further first evaluation model; (see Geffner section [B.2] “For CATE estimation, we need to train a separate surrogate predictor per graph samples.”)
computing a first aggregate score for the first candidate hyperparameter based on the first matching score and the further first matching score, wherein the second proxy model is trained based on the first aggregate score. (see Geffner section [4.1] “We run DECI using two models for exogenous noise: a Gaussian with learnable variance (identified as DECI-G) and a spline flow (DECI-S).”, and see Geffner figure 4 “the DECI results by are connected with with soft lines. The figure shows mean results across five different random seeds.”)
Geffner does not explicitly teach “calculating a further first matching score based on the trained further first evaluation model and the trained further first proxy model;”
However, Watson teaches
calculating a further first matching score based on the trained further first evaluation model and the trained further first proxy model; (see Watson paragraph [0029] “The results of the application of separate evaluation data 115 to synthetic data model and original data model can be compared to one another using various exemplary comparison procedures. For example, as discussed below, an integrated variants analysis procedure, or an analysis of variance (“ANOVA”) procedure”)
As to claim 3, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 2, wherein the first proxy model is trained based on the target dataset, the initial candidate hyperparameter, and a first random seed, wherein the further first proxy model is trained based on the target dataset, the initial candidate hyperparameter, and a further first random seed. (see Geffner section [3.1] “We consider two possible models for the distribution of z. 1) A simple Gaussian … are learnt. 2) A flow …”, and see Geffner section [4.1] “Figure 4: … The figure shows mean results across five different random seeds.”)
As to claim 4, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, comprising:
sampling a second synthetic dataset based on the trained second proxy model; (see Geffner section [3.3] “After training, we can use the model learnt by DECI to simulate new samples x from p(xjG).”)
training a second evaluation model based on the second synthetic dataset and a second candidate hyperparameter, resulting in a trained second evaluation model; (see Geffner section [B.2] “we need to train a separate surrogate predictor per graph samples. We draw 10 different graph samples and 10000 (xC; xY ) pair samples for each graph. We use these to train the surrogate models.”)
training a third proxy model based on the target dataset, the second candidate hyperparameter and the second matching score, resulting in a trained third proxy model. (see Geffner section [B.4] “We then marginalise the graphs using Monte Carlo: ”, and see Moharrer paragragraph [0025] “a regressor is trained to predict an estimated quality score based on a given dataset and a given hyperparameters configuration.”)
Geffner does not explicitly teach “calculating a second matching score based on the trained second evaluation model and the trained second proxy model; and”
However, Watson teaches
calculating a second matching score based on the trained second evaluation model and the trained second proxy model; and (see Watson paragraph [0029] “The results of the application of separate evaluation data 115 to synthetic data model and original data model can be compared to one another using various exemplary comparison procedures.”)
As to claim 5, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 4, comprising:
sampling a third synthetic dataset based on the trained third proxy model; (see Geffner section [3.3] “we can use the model learnt by DECI to simulate new samples x from p(xjG).”,
training a third evaluation model based on the third synthetic dataset and a third candidate hyperparameter, resulting in a trained third evaluation model; (see Geffner section [B.2] “For CATE estimation, we need to train a separate surrogate predictor per graph samples. We draw 10 different graph samples and 10000 (xC; xY ) pair samples for each graph. We use these to train the surrogate models.”)
training a fourth proxy model based on the target dataset, the third candidate hyperparameter and the third matching score, resulting in a trained fourth proxy model. (see Geffner section [B.4] “We then marginalise the graphs using Monte Carlo: ”, and see Moharrer paragragraph [0025] “a regressor is trained to predict an estimated quality score based on a given dataset and a given hyperparameters configuration.”)
Geffner does not explicitly teach “calculating a third matching score based on the trained third evaluation model and the trained third proxy model; and”
However, Watson teaches
calculating a third matching score based on the trained third evaluation model and the trained third proxy model; and (see Watson claim [3] “comparing first results from the training of the first model to second results”)
As to claim 6, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, wherein the first proxy model, the second proxy model and the first evaluation model each have a generative neural network architecture. ( see Geffner section [1] “DECI is an autoregressive-flow based non-linear additive noise SEM capable of learning complex nonlinear relationships between variables and non-Gaussian exogenous noise distributions.”, and see Geffner section [3.1] “We choose the learnable bijections k_1 to be a rational quadratic splines”)
As to claim 7, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, comprising:
determining using the trained first proxy model a first predicted target dataset property, wherein the first synthetic dataset is sampled based on the first predicted target dataset property; and (see Geffner section [B.4] “DECI may also be used to evaluate densities under intervened distributions. For a given graph, ”
determining using the trained first evaluation model a first predicted synthetic dataset property, wherein the first matching score is computed between the first predicted target dataset property and the first synthetic dataset property. (see Geffner section [3.3] “Finally, we use these samples to obtain a Monte Carlo estimate of the expectations required for ATE computation”, and see Watson paragraph [0007] “The synthetic dataset(s) can be evaluated by comparing first results from the training of the first model to second results from the training of the second model.”)
As to claim 8, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 7, wherein the first predicted target dataset property comprises a first predicted target dataset causal property, wherein the first predicted synthetic dataset property comprises a first predicted synthetic dataset causal property. (See Geffner section [3.3] “DECI returns q‑(G), an approximation of the posterior over graphs given observational data. Then, interventional distributions and treatment effects can be obtained by marginalizing over graphs as”)
As to claim 9, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 8, wherein the first predicted target dataset causal property and the first predicted synthetic dataset causal property are embodied in respective causal graphs sampled from the trained first proxy model and the trained first evaluation model respectively. (See Geffiner section [3.3] “We sample a graph G q‑(G) and a set of exogenous noise variables z pz. We then input this noise into the learnt DECI structural equation model to simulate x”)
As to claim 10, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, wherein the first proxy model and the second proxy model are causal models, the method comprising: (see Geffner section [3.1] “DECI takes a Bayesian approach to causal discovery [15]. We model the causal graph G jointly with the observations x1…… xN as”)
sampling a causal graph from the trained second proxy model or another trained proxy model derived from the trained second proxy model via additional sampling and training operations. (see Geffner section [3.3] “we can use the model learnt by DECI to simulate new samples x from p_(xjG)… the “mutilated” graph obtained … We can then use eq. (11) to estimate causal quantities by marginalizing over these distributions.”)
As to claim 11, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 10, comprising:
determining a treatment action based on the causal graph sampled from the trained second proxy model or the other proxy model; and (see Geffner section [2] “The ATE and CATE quantities allow us to estimate the impact of our actions (treatments)”)
performing the treatment action on a physical or logical system. (See Geffner section [1] “Causal-aware decision making is pivotal in many fields such as economics [3, 70] and healthcare [4, 20, 63]. For example, in healthcare, caregivers may wish to understand the effectiveness of different treatments given only historical data. They aspire to estimate treatment effects from observational data, with incomplete or no knowledge of the causal relationships between variables.”)
As to claim 12, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, comprising:
wherein the first evaluation model is one of multiple first evaluation models trained based on respective first candidate hyperparameters, resulting in multiple trained first evaluation models, wherein training the second proxy model based on the target dataset, the first candidate hyperparameter and the first matching score comprises: (see Moharrer paragraph [0083] “Maximum number of iterations T, N number of random samples, and n number of trials at each round Output: Best Hyperparameter configuration … and pick top n performing instances”)
selecting the first candidate hyperparameter from the respective first candidate hyperparameters based on respective first matching scores computed based on the multiple trained first evaluation models and the trained first proxy model, and (see Moharrer paragraph [0047] “Propose a new hyperparameter configuration H.sup.r+1 Return H* = argmax.sub.H∈{H.sub.1.sub.,...,H.sub.T.sub.} S(A(H), D) 2.0 Hyperstars”, and see Moharrer paragraph [0083] “Maximum number of iterations T, N number of random samples, and n number of trials at each round Output: Best Hyperparameter configuration … and pick top n performing instances”)
fitting the second proxy model to the target dataset using the first candidate hyperparameter as selected based on based on respective first matching scores, resulting in the trained second proxy model. (see Geffner section [B.4] “We then marginalise the graphs using Monte Carlo: ”, and see Moharrer paragraph [0047] “Propose a new hyperparameter configuration H.sup.r+1 Return H* = argmax.sub.H∈{H.sub.1.sub.,...,H.sub.T.sub.} S(A(H), D) 2.0 Hyperstars”, and see Moharrer paragraph [0083] “Maximum number of iterations T, N number of random samples, and n number of trials at each round Output: Best Hyperparameter configuration … and pick top n performing instances”)
As to claim 13, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 12,
determining a first likelihood of the trained first proxy model with respect to the target dataset; (see Geffner section [B.4] “DECI may also be used to evaluate densities under intervened distributions.”, and see Geffner section [A] “the maximum likelihood estimate (MLE) recovers the ground truth”)
determining a second likelihood of the trained second proxy model with respect to the target dataset; (see Geffner section [A.2] "we will show that under a correctly specified model and with maximum likelihood training with infinite data, DECI can recover the unique ground truth graph G∗= G0 and the true data generating distribution pθ∗(x; G∗) = p(x; G0), where (θ∗,G∗) are MLE solutions...")
determining an additional second likelihood of the trained additional second proxy model with respect to the target dataset; and (see Geffner section [A.2] "we will show that under a correctly specified model and with maximum likelihood training with infinite data, DECI can recover the unique ground truth graph G∗= G0 and the true data generating distribution pθ∗(x; G∗) = p(x; G0), where (θ∗,G∗) are MLE solutions...")
selecting the trained second proxy model from a set comprising the trained second proxy model and the trained additional second proxy model based on the first likelihood, the second likelihood and the additional second likelihood. (See Moharrer paragraph [0024] “The general metamodel predicts the performance score of the ML model for each of the randomly generated set of hyperparameter configurations. … A few top predicted hyperparameters configurations may be selected”)
Geffner does not explicitly teach “wherein selecting the first candidate hyperparameter comprises selecting a subset of the respective first candidate hyperparameters based on the respective first matching scores, the subset comprising the first candidate hyperparameter and an additional first candidate hyperparameter;”, and “fitting an additional second proxy model to the target dataset using the additional first candidate hyperparameter, resulting in a trained additional second proxy model;”
However, Moharrer teaches
wherein selecting the first candidate hyperparameter comprises selecting a subset of the respective first candidate hyperparameters based on the respective first matching scores, the subset comprising the first candidate hyperparameter and an additional first candidate hyperparameter; (see Moharrer paragraph [0024] “the general metamodel for hyperparameters optimization of the ML model given a new dataset. Metafeatures about the new dataset are used with the trained metamodel to warm start a hyperparameters optimization algorithm. The general metamodel predicts the performance score of the ML model for each of the randomly generated set of hyperparameter configurations.”, and see Moharrer paragraph [0083] “pick top n performing instances”)
Additionally, Tee teaches
fitting an additional second proxy model to the target dataset using the additional first candidate hyperparameter, resulting in a trained additional second proxy model; (see Tee Paragraph [0007] “a second tuning region that excludes the failure region; identifies additional tuned hyperparameter values for each of the two or more hyperparameters based on results of the second tuning;”)
As to claim 14, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 12, wherein the respective first candidate hyperparameters are generated randomly. (see Moharrer paragraph [0083] “Generate n random hyperparameter configurations H.sup.1, H.sup.2…”)
As to claim 15, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 12, wherein the respective first candidate hyperparameters are generated via a Bayesian search based on the respective first matching scores. (see Tee paragraph [0051] “The ensemble of optimization models 140 … Tree-structured Parzen Estimators (TPE) model and variants thereof”, and see Tee paragraph [0059] “a Bayesian method (e.g., naïve Bayes, averaged one-dependence estimators, Bayesian belief network, etc.)”)
As to claim 16, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, wherein determining the initial hyperparameter comprises:
sampling an initial synthetic dataset based on the initial proxy model, (see Geffner section [3.3] “After training, we can use the model learnt by DECI to simulate new samples x”
Geffner does not explicitly teach “generating an initial proxy model,”, “training multiple initial evaluation models based on the initial synthetic dataset and respective initial hyperparameters,”, “training multiple initial evaluation models based on the initial synthetic dataset and respective initial hyperparameters”, “computing an initial matching score based on each trained initial evaluation model and the initial proxy model, and”, and “selecting the initial candidate hyperparameter from the respective initial hyperparameters based on the initial matching score computed for each trained initial evaluation model.”
However, Moharrer teaches
generating an initial proxy model, (see Moharrer [0025] “for each training dataset, a computer derives, from the training dataset, values for dataset metafeatures. The computer performs, for each hyperparameters configuration, of a machine learning (ML) model, including landmark hyperparameters configurations: configuring the ML model based on the hyperparameters configuration, training the ML model based on the training dataset, and obtaining an empirical quality score”
training multiple initial evaluation models based on the initial synthetic dataset and respective initial hyperparameters, (see Moharrer paragraph [0025] “for each hyperparameters configuration … configure the ML model based on the hyperparameters configuration; train the ML model based on the training dataset;”)
selecting the initial candidate hyperparameter from the respective initial hyperparameters based on the initial matching score computed for each trained initial evaluation model. (See Moharrer paragraph [0083] “pick top n performing instances”)
computing an initial matching score based on each trained initial evaluation model and the initial proxy model, and (See Watson claim [3] “comparing first results from the training of the first model to second results from the training of the second model.”)
As to claim 17, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, wherein the respective initial hyperparameters are: randomly generated, or (see Moharrer paragraph [0083] “Generate n random hyperparameter configurations H.sup.1, H.sup.2”)
determined via tuning of a likelihood of the target dataset with respect to an initial proxy model. ( see Geffner section [A] “The main idea is to first show that the maximum likelihood estimate (MLE) recovers the ground truth due to the correctly specified model. Then, we prove that optimal solutions from maximizing the ELBO are closely related to the MLE under mild assumptions.”)
As to claim 18, Geffner-Watson-Moharrer as modified by Tee teaches the method of claim 1, comprising, based on the trained second proxy model:
tuning, adapting, modifying or replacing an industrial machine, (see Geffner section [1] “Causal-aware decision making is pivotal in many fields such as economics [3, 70] and healthcare [4, 20, 63]. For example, in healthcare, caregivers may wish to understand the effectiveness of different treatments given only historical data. They aspire to estimate treatment effects from observational data, with incomplete or no knowledge of the causal relationships between variables.”)
examiner note: Analogous to medical diagnosis, the action can be machine repair.
performing a maintenance or repair action performed on a machine, (optional limitation not addressed in this office action)
generating image data, audio data, text or other content, or (optional limitation not addressed in this office action)
performing a security mitigation action. (optional limitation not addressed in this office action)
As to claim 19, this is directed to a system or a computing device claim that corresponds to method claim 1. See the rejection for claim 1 above, which also applies to claim 19, additionally
at least one memory configured to store computer-readable instructions; and (see Watson paragraph [0062] “a computer-accessible medium 1115 (e.g., as described herein above, a storage device such as a hard disk, floppy disk, memory stick,”)
at least one hardware processor coupled to the at least one memory, wherein the computer-readable instructions are configured to cause the at least one hardware processor to perform operations comprising: (see Watson paragraph [0062] “a computer-accessible medium 1115 (e.g., as described
herein above, a storage device such as a hard disk, floppy disk, memory stick, CD-ROM, RAM, ROM, etc., or a collection thereof) can be provided (e.g., in communication with the processing arrangement 1105). The computer-accessible medium 1115 can contain executable instructions 1120 thereon.”)
As to claim 20, this is directed to a computer program claim that corresponds to method claim 1. See the rejection for claim 1 above, which also applies to claim 20.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ABDULLAH K ABOUD whose telephone number is (571)272-0025. The examiner can normally be reached Mon-Fri 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, Li B Zhen, can be reached at (571) 272-3768. 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.
/ABDULLAH KHALED ABOUD/Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights