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 § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding claim 1, Under Step 1 of the Subject Matter Eligibility Test of Products and
Processes, the claim is directed towards a method which is considered a process, which is one of the four statutory categories.
Next, under a Step 2A Prong 1 Analysis, the following limitations from the claim are interpreted to be, under the broadest reasonable interpretation, a grouping of abstract idea:
tuning hyperparameters of a machine learning model (mental process)
generating, for each hyperparameter, a performance attribution statistic corresponding to an evaluation metric of the machine learning model based on historical experiment statistics for the evaluation metric and the machine learning model (mental process)
assigning a numerical weight to each hyperparameter based on the performance attribution statistic of the hyperparameter (mental process)
iteratively updating, in a series of experiments, the hyperparameters based on the numerical weight assigned to each hyperparameter, wherein the numerical weight assigned to a hyperparameter indicates a relative likelihood of selection for updating the hyperparameter in any particular iteration compared to other hyperparameters (mental process)
updating the machine learning model with a non-zero set of the hyperparameters for the machine learning model from one of the experiments, wherein the non-zero set of the hyperparameters results in a recorded value of the evaluation metric that satisfies a tuning condition. (mental process)
Therefore, we have to examine the claim under Step 2A prong 2, which considers the additional elements within the claim. Since there are no additional elements within the claim, the claim is directed towards an abstract idea.
Under a Step 2B analysis, the claim's additional elements do not amount to significantly
more than the judicial exception as explained above in Step 2A prong 2. Therefore, the claim is ineligible.
Regarding claim 8, Under Step 1 of the Subject Matter Eligibility Test of Products and
Processes, the claim is directed towards a machine, which is one of the four statutory categories.
Next, under a Step 2A Prong 1 Analysis, the following limitations from the claim are interpreted to be, under the broadest reasonable interpretation, a grouping of abstract idea:
tuning hyperparameters of a machine learning model (mental process)
generate, for each hyperparameter, a performance attribution statistic corresponding to an evaluation metric of the machine learning model based on historical experiment statistics for the evaluation metric and the machine learning model (mental process)
assign a numerical weight to each hyperparameter based on the performance attribution statistic of the hyperparameter (mental process)
iteratively execute a series of experiments… selection of the at least one hyperparameter in each experiment being based on the numerical weight assigned to each hyperparameter, wherein the numerical weight assigned to a hyperparameter indicates a relative likelihood of selection for updating the hyperparameter in any particular iteration compared to other hyperparameters; (mental process)
update the machine learning model with a non-zero set of the hyperparameters for the machine learning model from one of the experiments, wherein the non- zero set of the hyperparameters results in a recorded value of the evaluation metric that satisfies a tuning condition. (mental process)
Therefore, we have to examine the claim under Step 2A prong 2, which considers the additional elements within the claim. The additional elements include:
One or more hardware processors
A performance attributor
A hyperparameter weight assessor
A hyperparameter updater
the machine learning model using different values of at least one hyperparameter of the machine learning model,
and a hyperparameter selector
The limitations, as drafted, merely acts as an indication of the technological environment or field of use, and “generally links” the judicial exception, to processors, an attributor, an updater, an assessor, a machine learning model with different hyperparameter values, and a selector that can perform it. (See MPEP 2106.05(h)) Therefore, the claim is directed towards an abstract idea.
Under a Step 2B analysis, the claim's additional elements do not amount to significantly
more than the judicial exception as explained above in Step 2A prong 2. Therefore, the claim is ineligible.
Regarding claim 15, Under Step 1 of the Subject Matter Eligibility Test of Products and
Processes, the claim is directed towards a manufacture, which is one of the four statutory categories.
Next, under a Step 2A Prong 1 Analysis, the following limitations from the claim are interpreted to be, under the broadest reasonable interpretation, a grouping of abstract idea:
A process of tuning hyperparameters of a machine learning model (mental process)
generating, for each hyperparameter, a performance attribution statistic corresponding to an evaluation metric of the machine learning model based on historical experiment statistics for the evaluation metric and the machine learning model (mental process)
assigning a numerical weight to each hyperparameter based on the performance attribution statistic of the hyperparameter; (mental process)
executing a series of experiments… selection of the at least one hyperparameter in each experiment being based on the numerical weight assigned to each hyperparameter, wherein the numerical weight assigned to a hyperparameter indicates a relative likelihood of selection for updating the hyperparameter in any particular iteration compared to other hyperparameters (mental process)
and updating the machine learning model with a non-zero set of the hyperparameters for the machine learning model from one of the experiments, wherein the non- zero set of the hyperparameters results in a recorded value of the evaluation metric that satisfies a tuning condition. (mental process)
Therefore, we have to examine the claim under Step 2A prong 2, which considers the additional elements within the claim. The additional elements include:
One or more tangible processor-readable storage media embodied with instructions for executing on one or more processors and circuits of a computing device
the machine learning model using different values of at least one hyperparameter of the machine learning model
The limitations, as drafted, merely indicates the technological environment or field of use, and “generally links” the judicial exception, to media, processors, circuitry, and the machine learning model with different hyperparameter values that can perform it. (See MPEP 2106.05(h)) Therefore, the claim is directed towards an abstract idea.
Under a Step 2B analysis, the claim's additional elements do not amount to significantly
more than the judicial exception as explained above in Step 2A prong 2. Therefore, the claim is ineligible.
Regarding claims 2, 9, and 16, the claims recite “the evaluation metric corresponds to a performance objective of the machine learning model to which the hyperparameters are being tuned.” The limitation, as drafted , is considered to be a “mental process”, which is a grouping of abstract idea. Therefore, the claims are rejected for the same reasons as set forth in the rejections of claims 1, 8, and 15.
Regarding claims 3, 10, and 17, the claims recite “the historical experiment statistics track historical values of the evaluation metric against different values of each hyperparameter.” The limitation, as drafted , is considered to be a “mental process”, which is a grouping of abstract idea. Therefore, the claims are rejected for the same reasons as set forth in the rejections of claims 1, 8, and 15.
Regarding claims 4, 11, and 18, the claims recite “wherein the performance attribution statistic corresponding to the hyperparameter indicates a sensitivity of the evaluation metric to changes in the hyperparameter” The limitation, as drafted , is considered to be a “mental process”, which is a grouping of abstract idea. Therefore, the claims are rejected for the same reasons as set forth in the rejections of claims 1, 8, and 15.
Regarding claim 5, 12, and 19, the claims recite “for multiple iterations limited by a compute budget, the hyperparameter updater is further configured to: randomly select a hyperparameter based on the numerical weight assigned to the hyperparameter,”, “update the hyperparameter to a new value”, “execute an experiment on the machine learning model based on the new value of the hyperparameter,”, and “record a value of the evaluation metric resulting from the experiment.” The limitations, as drafted, all are considered to be “mental processes”, which is a grouping of abstract idea. Therefore, the claims are rejected for the same reasons as set forth in the rejections of claims 1, 8, and 15.
Regarding claims 6, 13, and 20, the claims recite “the hyperparameter updater is further configured to update the hyperparameter to a new value based on maximizing a growth rate of the evaluation metric based on changes in the hyperparameter and minimizing a covariance of the evaluation metric based on changes in the hyperparameter.” The limitation, as drafted, under the broadest reasonable interpretation, is interpreted to be a “mathematical concept”, which is a grouping of abstract idea. Therefore, the claims are rejected for the same reasons as set forth in the rejections of claims 1, 8, and 15.
Regarding claims 7 and 14, the claims recite “the hyperparameter updater is further configured to update the hyperparameter to a new value using a Bayesian model.” The limitation, as drafted, under the broadest reasonable interpretation, is interpreted to be a “mathematical concept”, which is a grouping of abstract idea. Therefore, the claims are rejected for the same reasons as set forth in the rejections of claims 1 and 8.
Claim Rejections - 35 USC § 102
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 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.
Claim(s) 1-3, 5, 7-10, 12, 14-17 and 19 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Guoxin Sui et al. (Herein referred to as Sui) (Bayesian Contextual Bandits for Hyper Parameter Optimization).
Regarding claim 1, Sui teaches a method of tuning hyperparameters of a machine learning model, the method comprising: generating, for each hyperparameter, a performance attribution statistic corresponding to an evaluation metric of the machine learning model based on historical experiment statistics for the evaluation metric and the machine learning model; (“yi,k represents the performance of the model trained using xi with k units of resources (e.g., iterations)… The objective is to find the maximum yi,k with a limited budget B… If we pull the ith drawn arm k times, we observe a value yi,k. The best value observed in the history y∗ = max[yi,k] is stored in the state S. Rx (r) = P[r|x] is a probability distribution over rewards.”, pg. 3, right column, first and second paragraphs; See also Algorithm 2 on pg. 3) (yi,k correspond to performance attribution statistic. Algorithm 2 shows the evaluation of hyperparameters, with f(x*) in step 4 and 6 corresponding to an evaluation metric, and H corresponding to historical experiment statistics) assigning a numerical weight to each hyperparameter based on the performance attribution statistic of the hyperparameter; (“Given the reward distribution F(ri), vanilla Thompson sampling cannot be directly applied since we have infinite arms. A strategy is needed to decide how many new arms to be added to the ‘‘basket.’’ Here we propose a dynamic Thompson sampling for HPO… Each round, we follow the process in Algorithm 3.”, pg. 3, left column, bottom paragraph; See also Algorithm 3 on pg. 5) (Algorithm 3 shows the evaluation of hyperparameters with Thompson Sampling, with step 4 corresponding to a performance attribution statistic. yi,k is uses to determine F(ri,k), which corresponds to the numerical weight. (See Equations 17-21 on pg. 5 for more details.)) iteratively updating, in a series of experiments, the hyperparameters based on the numerical weight assigned to each hyperparameter, wherein the numerical weight assigned to a hyperparameter indicates a relative likelihood of selection for updating the hyperparameter in any particular iteration compared to other hyperparameters; (See Algorithm 3 on pg. 5. F(ri,k), the numerical weight, is used to update xi, which corresponds to a hyperparameter. The distribution F(ri,k) gives a probability of a hyperparameter being selected, as the hyperparameter with the largest reward, ri, is selected, fully teaching the limitation.) and updating the machine learning model with a non-zero set of the hyperparameters for the machine learning model from one of the experiments, wherein the non- zero set of the hyperparameters results in a recorded value of the evaluation metric that satisfies a tuning condition. (“N is the maximum number of iterations that a hyper parameter can be trained. The objective is to find the maximum yi,k with a limited budget B.”, pg. 3, left column, first paragraph) (Training a hyperparameter for a number of iterations corresponds to a tuning condition.)
Regarding claim 8, Sui teaches A computing system for tuning hyperparameters of a machine learning model, the computing system comprising: one or more hardware processors; (“The experiment for predictive performance analysis in this part was run in a machine with an Intel Core i7-8700 CPU and 16GB RAM”, pg. 7, left column, second paragraph) a performance attributor executable by the one or more hardware processors and configured to generate, for each hyperparameter, a performance attribution statistic corresponding to an evaluation metric of the machine learning model based on historical experiment statistics for the evaluation metric and the machine learning model; (“yi,k represents the performance of the model trained using xi with k units of resources (e.g., iterations)… The objective is to find the maximum yi,k with a limited budget B… If we pull the ith drawn arm k times, we observe a value yi,k. The best value observed in the history y∗ = max[yi,k] is stored in the state S. Rx (r) = P[r|x] is a probability distribution over rewards.”, pg. 3, right column, first and second paragraphs; See also Algorithm 2 on pg. 3) (yi,k correspond to performance attribution statistic. Algorithm 2 shows the evaluation of hyperparameters, with f(x*) in step 4 and 6 corresponding to an evaluation metric, and H corresponding to historical experiment statistics) a hyperparameter weight assessor executable by the one or more hardware processors and configured to assign a numerical weight to each hyperparameter based on the performance attribution statistic of the hyperparameter; (“Given the reward distribution F(ri), vanilla Thompson sampling cannot be directly applied since we have infinite arms. A strategy is needed to decide how many new arms to be added to the ‘‘basket.’’ Here we propose a dynamic Thompson sampling for HPO… Each round, we follow the process in Algorithm 3.”, pg. 3, left column, bottom paragraph; See also Algorithm 3 on pg. 5) (Algorithm 3 shows the evaluation of hyperparameters with Thompson Sampling, with step 4 corresponding to a performance attribution statistic. yi,k is uses to determine F(ri,k), which corresponds to the numerical weight. (See Equations 17-21 on pg. 5 for more details.)) a hyperparameter updater executable by the one or more hardware processors and configured to iteratively execute a series of experiments, on the machine learning model using different values of at least one hyperparameter of the machine learning model, selection of the at least one hyperparameter in each experiment being based on the numerical weight assigned to each hyperparameter, wherein the numerical weight assigned to a hyperparameter indicates a relative likelihood of selection for updating the hyperparameter in any particular iteration compared to other hyperparameters; (See Algorithm 3 on pg. 5. F(ri,k), the numerical weight, is used to update xi, which corresponds to a hyperparameter. The distribution F(ri,k) gives a probability of a hyperparameter being selected, as the hyperparameter with the largest reward, ri, is selected, fully teaching the limitation.) and a hyperparameter selector executable by the one or more hardware processors and configured to update the machine learning model with a non-zero set of the hyperparameters for the machine learning model from one of the experiments, wherein the non- zero set of the hyperparameters results in a recorded value of the evaluation metric that satisfies a tuning condition. (“N is the maximum number of iterations that a hyper parameter can be trained. The objective is to find the maximum yi,k with a limited budget B.”, pg. 3, left column, first paragraph) (Training a hyperparameter for a number of iterations corresponds to a tuning condition.)
Regarding claim 15, Sui teaches One or more tangible processor-readable storage media embodied with instructions for executing on one or more processors and circuits of a computing device a process of tuning hyperparameters of a machine learning model, (While a tangible processor-readable storage media isn’t explicitly taught, one would implicitly need those components to distribute and run the method of Sui.) the process comprising: generating, for each hyperparameter, a performance attribution statistic corresponding to an evaluation metric of the machine learning model based on historical experiment statistics for the evaluation metric and the machine learning model; (“yi,k represents the performance of the model trained using xi with k units of resources (e.g., iterations)… The objective is to find the maximum yi,k with a limited budget B… If we pull the ith drawn arm k times, we observe a value yi,k. The best value observed in the history y∗ = max[yi,k] is stored in the state S. Rx (r) = P[r|x] is a probability distribution over rewards.”, pg. 3, right column, first and second paragraphs; See also Algorithm 2 on pg. 3) (yi,k correspond to performance attribution statistic. Algorithm 2 shows the evaluation of hyperparameters, with f(x*) in step 4 and 6 corresponding to an evaluation metric, and H corresponding to historical experiment statistics) assigning a numerical weight to each hyperparameter based on the performance attribution statistic of the hyperparameter; (“Given the reward distribution F(ri), vanilla Thompson sampling cannot be directly applied since we have infinite arms. A strategy is needed to decide how many new arms to be added to the ‘‘basket.’’ Here we propose a dynamic Thompson sampling for HPO… Each round, we follow the process in Algorithm 3.”, pg. 3, left column, bottom paragraph; See also Algorithm 3 on pg. 5) (Algorithm 3 shows the evaluation of hyperparameters with Thompson Sampling, with step 4 corresponding to a performance attribution statistic. yi,k is uses to determine F(ri,k), which corresponds to the numerical weight. (See Equations 17-21 on pg. 5 for more details.)) executing a series of experiments, on the machine learning model using different values of at least one hyperparameter of the machine learning model, selection of the at least one hyperparameter in each experiment being based on the numerical weight assigned to each hyperparameter, wherein the numerical weight assigned to a hyperparameter indicates a relative likelihood of selection for updating the hyperparameter in any particular iteration compared to other hyperparameters (See Algorithm 3 on pg. 5. F(ri,k), the numerical weight, is used to update xi, which corresponds to a hyperparameter. The distribution F(ri,k) gives a probability of a hyperparameter being selected, as the hyperparameter with the largest reward, ri, is selected, fully teaching the limitation.) and updating the machine learning model with a non-zero set of the hyperparameters for the machine learning model from one of the experiments, wherein the non- zero set of the hyperparameters results in a recorded value of the evaluation metric that satisfies a tuning condition. (“N is the maximum number of iterations that a hyper parameter can be trained. The objective is to find the maximum yi,k with a limited budget B.”, pg. 3, left column, first paragraph) (Training a hyperparameter for a number of iterations corresponds to a tuning condition.)
Regarding claims 2, 9, and 16, Sui teaches the method, system, and medium of claims 1, 8, and 15 respectively, as well as the evaluation metric corresponds to a performance objective of the machine learning model to which the hyperparameters are being tuned. (“For a hyper parameter configuration xi ∈ X, yi,k represents the performance of the model trained using xi”, pg. 3, left column, bottom paragraph)
Regarding claims 3, 10 and 17, Sui teaches the method, system, and medium of claims 1, 8, and 15 respectively, as well as the historical experiment statistics track historical values of the evaluation metric against different values of each hyperparameter. (“Algorithm 1 gives the general structure of the BO frame work. Mt is the surrogate function used to predict the performance of configurations and the historical observations are stored in H, which will be used to update Mt. S is the acquisition function, which determines the best configuration x∗ to evaluate in the next iteration.”, pg. 2, right column, above “Algorithm 1”; See Algorithm 1)
Regarding claims 5, 12 and 19, Sui teaches the method, system, and medium of claims 1, 8, and 15 respectively, as well as for multiple iterations limited by a compute budget, selecting a hyperparameter based on the numerical weight allocated assigned to the hyperparameter, (“with k units of resources (e.g., iterations), s.t. 1 ≤ k ≤ N, N is the maximum number of iterations that a hyper parameter can be trained. The objective is to find the maximum yi,k with a limited budget B… Probability matching strategy selects an action x according to the probability that x is the optimal action. Thompson sampling[20], [21] implements probability matching by sampling a reward distribution from posterior. Define r(x) as the reward function, the probability that xi is the optimal action is thus:”, pg. 3, right column, top paragraph; pg. 4, left column, under “2) PROBABILITY MATCHING WITH THOMPSON SAMPLING”) updating the hyperparameter to a new value, (“the training data contains the performances of 35 different hyper parameters, where 5 of them have been trained to the end (27 training epochs) and 30 of them are trained to the halfway (sampled uniformly from [6,24]).“, pg. 7, left column, second paragraph) (The hyper parameters are updated during training.) executing an experiment on the machine learning model based on the new value of the hyperparameter, and recording a value of the evaluation metric resulting from the experiment. (See algorithms 3 and 4 on pg. 5) (Hyper parameters are input into Algorithm 3 or 4 to help optimize the hyper parameters using Thompson Sampling or Entropy Search.)
Regarding claims 7 and 14, Sui teaches the method of claim 1, and the system of claim 8, as well as updating the hyperparameter to a new value using a Bayesian model. (“The fundamental challenge in bandits problems is the need for balancing exploration and exploitation. Here we introduce three strategies in Bayesian bandits.”, pg. 4, left column, under “B. STRATEGIES IN BAYESIAN BANDITS”)
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 4, 11, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Sui in view of Honghe Jin. (Herein referred to as Jin) (Hyperparameter Importance for Machine Learning Algorithms)
Regarding claims 4, 11, and 18, Sui teaches the method, system, and medium of claims 1, 8, and 15 respectively, but does not explicitly teach the performance attribution statistic corresponding to the hyperparameter indicates a sensitivity of the evaluation metric to changes in the hyperparameter.
Jin teaches the performance attribution statistic corresponding to the hyperparameter indicates a sensitivity of the evaluation metric to changes in the hyperparameter. (“The intuition behind Definition 1 is that Varθj (R∗ (θ)) measures the variability of the risk when θj is changed. If the risks change much when θj varies, then θj should be tuned carefully since it can influence the results.”, pg. 2, second to last paragraph) (Jin uses variance of risk as a performance attribution statistic, which intrinsically indicates a sensitivity to changes in the hyperparameter.)
Therefore, it would have been considered obvious to one of ordinary skill in the art, prior to the application’s filing date, to combine the hyper parameter optimization of Sui with the evaluation metrics of Jin. One would be motivated to combine the teachings, prior to the application’s filing date, as this allows for different quantifications of the importance of the hyperparameter, as disclosed in Jin. (“More specifically, we use the variance of risks with varying hyperparameter values to quantify the importance of the hyperparameter, which is proved to be consistent when subsampling a subset from the entire data set”, pg. 2, second paragraph)
Claims 6, 13, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Sui in view of Bobak Shahriari et al. (Herein referred to as Shahriari) (Taking the Human Out of the Loop: A Review of Bayesian Optimization)
Regarding claims 6, 13, and 20, Sui teaches the method, system, and media of claims 1, 8, and 15 respectively, but does not explicitly teach updating the hyperparameter to a new value based on maximizing a growth rate of the evaluation metric based on changes in the hyperparameter and minimizing a covariance of the evaluation metric based on changes in the hyperparameter.
Shahriari teaches updating the hyperparameter to a new value based on maximizing a growth rate of the evaluation metric based on changes in the hyperparameter (“These acquisition functions trade off exploration and exploitation; their optima are located where the uncertainty in the surrogate model is large (exploration) and/or where the model prediction is high (exploitation). Bayesian optimization algorithms then select the next query point by maximizing such acquisition functions”, pg. 150; See also Fig. 1) (Fig. 1 shows graphic changes based on exploitation and exploration, indicating some sort of rate) and minimizing a covariance of the evaluation metric based on changes in the hyperparameter. (“One early approach to modeling large n with GPs considered using m G n inducing pseudoinputs to reduce the rank of the covariance matrix to m, resulting in a significant reduction in computational cost.” pg. 158, left column, top of the page)
Therefore, it would have been considered obvious to one of ordinary skill in the art,
prior to the current application’s filing date, to combine the hyperparameter optimization and selection of Sui with the exploration, exploitation, and covariance reduction of Shahriari. One would be motivated to combine the two teachings, prior to the filing date of the current application, as these functionality allow for cheaper computational resources, as disclosed in Shahriari. (“Naturally, these acquisition functions are often even more multimodal and difficult to optimize, in terms of querying efficiency, than the original black-box function f. Therefore, it is critical that the acquisition functions be cheap to evaluate or approximate: cheap in relation to the expense of evaluating the black box f… One early approach to modeling large n with GPs considered using m G n inducing pseudoinputs to reduce the rank of the covariance matrix to m, resulting in a significant reduction in computational cost.” pg. 150, right column; pg. 158, left column, top of the page)
Response to Arguments
Applicant's arguments filed on February 6th, 2026 have been fully considered but they are not persuasive. The applicant argues in substance:
The examiner respectfully disagrees. According to MPEP 2106.04(a)(2), a claim that requires a computer may still recite a mental process. The examiner interprets the steps, under the broadest reasonable interpretation, as a concept that is performed in the human mind, and merely performed on a generic computer and/or a computing environment. (See MPEP 2106.04(a)(2) under III. Mental Processes) Therefore, the steps fall inside the category of mental processes.
Argument 1: The amended claims present the same type of computer-implemented, resource- constrained optimization that the USPTO has already characterized as eligible in Examples 37 and 42 of the subject-matter eligibility guidance.
Specifically, the applicant asserts:
The claims are not directed to mere organizing information or mental steps.
Specifically, the Applicant states that the claimed steps interpreted to be mental processes in the rejection are machine-executed operations, and therefore cannot be performed in the human mind. The Applicant further states that the feedback-and-control scheme requires physical use of the processor and memory, and cannot be carried out with the aid of pencil and paper. Furthermore, the applicant states that a human cannot practically execute model training experiments, update hyperparameters, or generate evaluation metrics. Moreover, the claims are not directed toward mere mathematics.
The Office Action did not characterize any of the limitations as organizing information nor mere mathematics. However, as pointed out in the rejection, the claim does recite mental steps. According to MPEP 2106.04(a)(2), a claim that requires a computer may still recite a mental process. The examiner interprets the steps, under the broadest reasonable interpretation, as mental processes performed within a computing environment, and merely using a computer as tool to perform the mental processes. As such the claims are directly toward mental steps, which is a grouping of abstract idea.
Like Example 42, Claim 1 specifies a machine-learning-centric control loop
Specifically, the Applicant asserts that the claimed tuning of hyperparameters and Example 42's training (adjusting network weights) are analogous, although technically different, activities. Although hyperparameters and model weights differ, both govern machine execution behavior, and both require iterative machine computation not performable mentally.
As pointed out in point 1, claims that require a computer may still recite a mental process. Under Step 2A Prong 1, the adjusting of weights, and iterative computation are interpreted under the BRI to be mental processes which are performed in a computing environment, and use a computer as a tool to perform the mental processes. Under Step 2A Prong 2, the additional elements, comprising machine execution components are interpreted to be mere instructions to apply a judicial exception, as the components are used as tools to perform the abstract idea. Under Step 2B, the components do not amount to significantly more than the judicial exception. When properly analyzed under the subject matter eligibility test, the claim does not recite eligible subject matter under 35 U.S.C. 101.
Claim 1 follows a similar pattern to Example 37 because both improve a technical field
Specifically, the Applicant asserts that Claim 1 improves the technical field of ML-model deployment and compute-budget utilization. As in Example 37, the output (the selected hyperparameter set) is immediately usable by the machine-learning system and is not an abstract data display.
The examiner respectfully disagrees. In Example 37, the claim recited automatically moving the most used icons to a position on the GUI closest to the start icon of the computer system, which amounted to a recitation of the technical improvement in the claim. In the instance claim 1, the selection of hyperparameters isn’t reflected in the claims, rather the claim recites updating the model. Updating the model amounts to a mental process rather than an improvement to a technology. As such, the claim does not recite an improvement of a technical field, but rather the improvement of an abstract idea.
Under Step 2A-Prong 2, the claims are integrated into a practical application, specifically an improvement in computer operation analogous to the efficiency improvement credited in Example 42. Because each step is tied to concrete data structures held in memory and executed by the claimed processors, the claims do not pre-empt all forms of hyperparameter tuning; it is limited to a specific, machine-implemented technique.
The examiner respectfully disagrees. In Example 42 the claim recited automatically generating a message... by the content server whenever updated information has been stored; and transmitting the message to all of the users over the computer network in real time. As pointed out in point 3, the steps of claim 1 are directed towards abstract idea, and not additional elements that can be integrated into a practical application. Even if the claims recite the steps as additional elements, the claims would not integrate into a practical application by improving computer operations, but rather recite the improvement of an abstract idea.
Even if the mathematical calculations were viewed as an abstract idea, Step 2B is satisfied: the claim adds significantly more than a generic computer. It recites specialized components (performance attributor, weight assessor, updater, selector) and a compute-budgeted control policy that applies numeric weights to influence real-time experimental selection, elements not found in the routine or conventional art cited by the Office.
The examiner respectfully disagrees. The performance attributor, weight assessor, updater, selector, and policy are tools, found in a computing environment, used to perform the abstract idea. Given the BRI, each of these elements amounts to a hardware or software component of a computer, use as tools to perform the abstract ideas. These amount to generic computer components and not specialized components, as asserted by the Applicant. As such, the elements do not add significantly more to the abstract ideas. For these reasons, the 101 rejections are maintained.
Argument 3: Jin does not teach the limitation, “historical experiment statistics”, nor does Jin teach the newly amended limitations of claims 1, 8, and 15.
Applicant’s arguments with respect to claim(s) 1, 8, and 15 have been considered but are not persuasive. The limitation, “historical experiment statistics” is taught in the prior action by previous subsample results which corresponds to the limitation. In this rejection the argument is are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Therefore, the 102 and 103 rejections are maintained.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Tyler E Iles whose telephone number is (571)272-5442. The examiner can normally be reached 9:00am - 5:00pm.
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, Kakali Chaki can be reached at (571) 272-3719. 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.
/T.E.I./Patent Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
