DETAILED ACTION The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Responsive to the communication dated 10/04/2022. Claims 1 – 20 are presented for examination. Priority ADS dated 9/22/2022 claims domestic priority to provisional application no. 63246982 dated 9/22/2021. Information Disclosure Statement IDS dated 10/04/2022, 10/04/2022 have been reviewed. See attached. Drawings The drawings dated 9/22/2022 have been reviewed. They are accepted. Specification The abstract dated 9/22/2022 has 5 lines 75 words and no legal phraseology. The abstract is accepted. Claim Rejections - 35 USC § 101 35 U.S.C. 101 reads as follows: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title. Claim 1 – 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception without significantly more. Claim 1. STEP 1: Yes. The claim recites “a design optimization method…”. A method is one of the statutory categories. STEP 2A PRONG ONE. The claim recites: “ a design optimization method comprising: preparing a symbolic tree ; updating node symbol parameters using a plurality of samples; sampling the plurality of samples with a method for solving a multi-armed bandit problem ; promoting each sample in the plurality of samples down a path of the symbolic tree ; evaluating each path with a fitness function ; and outputting a path of the symbolic tree ” which are a series of mathematical operation for solving the category of stochastic scheduling problems which the “multi-armed bandit problem” one. In the claim, a symbolic tree is a mathematical construct known as a recursive partitioning graph. The claim recites to “solve” the multi-armed bandit problem and also states to “evaluate” a path through the recursive partitioning graph using an “objective function.” An objective function is a mathematical function that is either minimized or maximized. Accordingly, the claim is directed towards a mathematical abstract idea of “optimizing” parameters (i.e., values of variables) in an equation. STEP 2A PRONG TWO: NO. While the claim recites “design” this merely characterizes the optimized parameter values as being linked to the field of “design.” This merely generally links the mathematical concept of optimizing variable values of a mathematical function to a generalized notion of “design.” Generally linking an abstract idea to a field of use is not indicative of a practical application. See MPEP 2106.05(h). While the claim recites “ and outputting a path of the symbolic tree ” this is merely a recitation to output data from the mathematical operations. Merely outputting a mathematical output is not indicative of a practical application. Additionally, even if these elements were interpreted to be something other than the abstract idea itself, merely “outputting” data is insufficient application. See MPEP 2106.05(f), MPEP 2106.05(g) STEP 2B: NO. The claim recites “and outputting a path of the symbolic tree” which is simply outputting data at a high level of generality because it simply recites “outputting.” The general “outputting” of data is a conventional computer activity. Generally reciting to output data is routine and conventional. Indeed, Sugimura_2009 states “ we use the commercial software JMP (SAS Institute) for calculating decision tree diagrams” and shows an illustration of a path through a decision tree in Fig. 3 output from the commercial JMP tool. The fact that a commercial software tool exists that outputs a path of a symbolic tree is evidence that such activities are well-understood routine and conventional. Accordingly, as outlined above, when the elements of the claim are considered individually and as a whole, the claim is found to be directed towards an abstract idea without a practical application and without significantly more. The claim is rejected under 35 USC 101. 2. The claim recites: “ further comprising: providing at least one design parameter ” however, parameters are merely variables in the mathematical equations. Naming the variables or rather linking the variables to the field of design does not provide a practical application nor indicate significantly more than the abstract idea itself. 3. The claim recites: “ wherein the at least one design parameter comprises one of: a discrete parameter; and a continuous parameter ” which merely characterizes the variables to be mathematically discrete or continuous. These elements are merely additional recitation of mathematical elements. 4. The claim recites: “ further comprising: providing a plurality of design parameters, the plurality of design parameters further comprising: discrete parameters and continuous parameters ” which merely characterizes the variables to be mathematically discrete or continuous. These elements are merely additional recitation of mathematical elements. 5. While the claim recites: “ wherein the method for solving the multi-armed bandit problem comprises Thompson sampling ” which are additional mathematical elements because Thompson sampling is a mathematical/statistical method. 6. While the claim recites: “ further comprising: sampling using batch; computing a success rate; and updating Thompson parameters ” which are merely additional mathematical elements. 7. While the claim recites “ further comprising: providing an error function, the error function defining a design objective ” which are merely additional mathematical elements as an error function is merely some sort of mathematical difference associated with the objective function and the objective function is a mathematical maximization or minimization. 8. While the claim recites “ wherein the design objective comprises an optical system design objective ” these elements merely link the objective function to the field of “optical system design” and merely linking the use of a mathematical calculation to a field of use is not a practical application nor is it significantly more. Indeed, Durand_2018 teaches to perform multi-armed bandit methods using Thompson sampling to optimize optical system design parameters. Accordingly, it is known in the art to have design objectives that include optical system design objectives. Therefore, these elements are not significantly more than the abstract idea because they are also common in the art of optical systems. 9. The limitations of claim 9 are substantially the same as those of claim 1 and are rejected due to the same reasons as outlined above for claim 1. Additionally, while the claim recites “ A computer implemented optimization method …” the mere recitation to use a generally recited computer to perform a mathematical optimization is not indicative of a practical application nor significantly more. See MPEP 2106.05(f) and MPEP 2106.05(g). 10. the claim recites “wherein the preparation phase further comprises: generating a tree node with two sets of distributions, wherein each tree node contains a Thompson Distribution” which is part of the mathematical abstract idea. 11. The claim recites: “ wherein each node contains a plurality of parameter priors for each of its respective parameters ” this is a recitation of mathematical operations as a prior (prior probability distribution). 12. The claim recites: “ wherein the parameter phase further comprises: determining a batch size and an error value for the epoch ” which are also part of the mathematical operations. 13. The claim recites “ wherein the parameter phase further comprises: setting a batch size to be a number of samples taken in each rejection phase ” which are part of the mathematical operation. 14. The claim recites: “ wherein the parameter phase further comprises: updating parameter distributions using saved samples and incrementing the epoch ” which is part of the mathematical operation. 15. the claim recites: “ wherein the rejection phase further comprises: evaluating an error function for a selected path on the symbolic tree ” which is part of the mathematical operation. 16. The claim recites: “ wherein the error function defines a design objective ” which only generally links the use of mathematical objective function to the general notion of “design” and generally linking the use of an abstract idea to a field of use is not indicative of a practical application or significantly more. 17. T he limitations of claim 17 are substantially the same as those of claim 1 and are rejected due to the same reasons as outlined above for claim 1. Additionally, while the claim recites “ An optimization system comprising: a computer system, the computer system further comprising: at least one processor; a graphical user interface; and a computer-usable medium embodying computer program code, the computer- usable medium capable of communicating with the at least one processor, the computer program code comprising instructions executable by the at least one processor and configured for ” the mere recitation to use a generally recited computer to perform a mathematical optimization is not indicative of a practical application nor significantly more. See MPEP 2106.05(f) and MPEP 2106.05(g). 18. The claim recites “ further comprising: providing at least one design parameter, the at least one design parameter comprising one of: a discrete parameter; and a continuous parameter ” are part of the mathematical abstract idea. 19. The claim recites “ wherein the method for solving the multi-armed bandit problem comprises Thompson sampling further comprising sampling using batch; computing a success rate; and updating Thompson parameters ” which are part of the mathematical abstract idea. 20. The claim recites: “ further comprising: providing an error function, the error function defining a design objective ” which are part of the abstract idea. 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. Claims 1 - 7 are rejected under 35 U.S.C. 103 as being unpatentable over Gautier_2021 (Sherlock: A M ulti- O bjective Design Space Exploration Framework, April 1, 2021) in view of Sugimura_2009 (A New Design Method based on Cooperative Data Mining from Multi-Objective Design Space , Journal of Computational Science and Technology, Vol. 3, No. 1, 2009). Claim 1. Gautier_2021 makes obvious “ A design optimization method ( abstract: “Design space exploration (DSE) provides intelligent methods to tune … optimization parameters …”; introduction: “optimizing… design…”; page 2: “… a design space exploration (DSE) framework that uses active learning to evaluate and intelligently explore the HLS design space… reach the set of Pareto optimal designs …”) comprising: p reparing a symbolic tree; Updating node symbol parameters using a plurality of samples; Sampling the plurality of samples with a method for solving a multi-armed bandit problem ; Promoting each sample in the plurality of samples down a path of the symbolic tree; Evaluating each path with a fitness functio n” ( page 2: “… we create a selection strategy based on the multi-armed bandit problem that rewards the models directly improving the actual Pareto front ; Page 5 Fig. 2: Pareto Score and choice of sampling method; page 8: “… a popular ensemble model is the Random Forest predictor based on a set of decision trees … Sherlock can use any of these types of learning algorithms . We experiment with Random Forest and Gaussian process as they can generally model complex design spaces…” : page 9 Algorithm 3 : the model selection algorithm, Page 9: “… we propose to learn the best model using a multi-armed bandit strategy that iteratively updates the importance of each model based on the Pareto set improvement…”; page 10 section 2.4 ) ; and o utputting a path of the symbolic tree. Gautier_2021 does not teach “ and outputting a path of the symbolic tree.” Sugimura_2009 makes obvious “ and outputting a path of the symbolic tree” ( Fig. 2 illustrates a decision tree analysis performed on the design database and outputting at least a list of design rules. Page 292 section 2.5 Decision Tree Analysis and Fig. 3 illustrates a path through the symbolic tree that is defined by the rules. Page 292 section 2.5: “… following this procedure, a tree diagram, as shown in Fig. 3 is obtained… a single design rule can be obtained by tracing a path to the desired result… the rule is obtained as: …[equation 8]… we use the commercial software JMP (SAS institute) for calculating decision tree diagrams…”; page 297: “… decision tree analysis was then applied to the database, and a corresponding decision tree diagram to each objective function was obtained. From these diagrams , the following rules for extremely improving corresponding objective functions were obtained… the order of design variables appearing in the condition terms represents the order of sensitivity to the corresponding objective function…” ). Gautier_2021 and Sugimura_2009 are analogous art because they are from the same field of endeavor called design optimization. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Gautier_2021 and Sugimura_2009. The rationale for doing so would have been that Gautier_2021 teaches to perform parameter optimization by doing design space exploration and Sugimura_2009 teaches “ it has become common to design products using parameter surveys and optimization that use simulations and experiments. The resultant data can be recognized as a design database… we believe that it is more important to analyze such databases to deepen understanding of design problems. Namely, we should determine a final solution after reviewing the design knowledge obtained from the design database. Therefore, it is believed that a parameter design method that practices this idea should be developed” (page 1 introduction). Accordingly, Sugimura_2009 teaches to take the results of design space exploration (DSE) and to perform a decision tree analysis and output the design rules (i.e., the trace through the decision tree). Therefore, it would have been obvious to combine design space exploration (DSE) that creates a set of designs (i.e., a database) as taught by Gautier_2021 with the decision tree analysis of Sugimura_2009 for the benefit of gaining a better understanding of the design space to better inform the selection of the most optimal design parameters to obtain the invention as specified in the claims. Claim 2. Gautier_2021 makes obvious “ further comprising: p roviding at least one design parameter ” ( page 3 section 2.1: “A design space is composed of both an input space and an output space. The input space is a set of FPGA HLS designs that met the application’s functional requirements. The difference in the designs can be described through the definition of different design parameters, also known in the DSE literature as knobs … more formally stated, the input space X is defined as X = {K1 X K2 X … X Kn } ∈ X Mxn where Ki is a knob… Figure 1 provides an example of a design space. The input space consists of five knobs (n = 5) and the output space has two objectives (o= 2)… ”). Claim 3 . Gautier_2021 makes obvious “ wherein the at least one design parameter comprises one of: a discrete parameter; and a continuous paramete r” ( page 3 section 2.1: “… Knobs take 𝑛 values, for 𝑛 ∈ [2, ∞ ]. They can be discrete, categorical, or continuous…”). Claim 4. Gautier_2021 makes obvious “ further comprising: p roviding a plurality of design parameters, the plurality of design parameters further comprising: discrete parameters and continuous parameters ” ( page 3 section 2.1: “… Knobs take 𝑛 values, for 𝑛 ∈ [2, ∞ ]. They can be discrete, categorical, or continuous…”). Claim 5. Gautier_2021 makes obvious “ wherein the method for solving the multi-armed bandit problem comprising Thompson sampling ” ( page 10: “… we update these distributions by selecting one bandit and observing the outcome. A good choice of sampling algorithm is Thompson Sampling [29] that provides a good tradeoff between exploration and exploitation…”) Claim 6. Gautier_2021 makes obvious “ further comprising: s ampling using batch; c omputing a success rate; and u pdating Thompson parameters ” ( page 9 algorithm 3; Page 10 Fig. 4; Page 10: “… In this case, we consider each model as a bandit. The outcome of observing one bandit is either an improvement in the current Pareto set, or no improvement. In other words, we are trying to learn a Bernoulli distribution for each model. Consequently, we can select the prior distribution of the bandits as a Beta distribution. We define the prior distribution with parameter 𝜃 for each model 𝑖 as 𝑃𝑖 ( 𝜃 ) = 𝐵𝑒𝑡𝑎 ( 𝛼𝑖 , 𝛽𝑖 ). We update these distributions by selecting one bandit and observing the outcome. A good choice of sampling algorithm is Thompson Sampling [29] that provides a good tradeoff between exploration and exploitation [4]. The algorithm draws a random sample from each distribution: 𝜃 ˆ 𝑖 ∼ 𝐵𝑒𝑡𝑎 ( 𝛼𝑖 , 𝛽𝑖 ) ∀𝑖 , then chooses the bandit with the largest sample value. The observation 𝑥 of the selected bandit corresponds to the improvement of hypervolume over the known designs (hypervolume( 𝐾 )), after we sample a design according to a strategy as defined in Section 2.2.3. In other words, if the model 𝑔𝑖 improved the Pareto set, 𝑥 is a positive outcome, i.e., 𝑥 = 1, otherwise 𝑥 = 0. A value of 𝑥 > 0 increases 𝛼𝑖 , while a value of 𝑥 = 0 increases the value of 𝛽𝑖 . As can be seen in Figure 4, by increasing 𝛼𝑖 and holding 𝛽𝑖 constant, the likelihood that the distribution provides are larger value (closer to 1) is increased. Likewise, increasing 𝛽𝑖 makes is more likely that smaller sample value will be selected (closer to 0). We use this updated function to compute the posterior distribution based on the outcome, and use it as prior for the next iteration. Algorithm 3 shows the details of the method and how it integrates with the Sherlock algorithm described in Algorithm 1. Note that we use an optional posterior reshaping factor 𝑟 that changes the variance of the distributions. As a result, increasing the value of 𝑟 favors exploitation over exploration (i.e., the model providing the best outcome gets selected more often), and the policy becomes more greedy . Increasing this value also has the side benefit that each positive outcome is given more consideration, and potential improvements from models later in the sampling process will re-adjust their importance faster. It provides a small chance to switch the most important model during the sampling process. A model selection algorithm is valuable…”). Claim 7. Gautier_2021 makes obvious “ further comprising: p roviding an error function, the error function defining a design objective ” ( page 11 section 3.1: “… we compute an error metric based on the ground truth design spaces, using the Average Distance to References Set (ADRS) metri c [19] ADRS measures the average normalized distance between the estimated Pareto front and the reference Pareto front. The closer it is to 0, the better the estimation is.). Claims 8 are rejected under 35 U.S.C. 103 as being unpatentable over Gautier_2021 in view of Sugimura_2009 in view of Durand_2018 (A machine learning approach for online automated optimization of super-resolution optical microscopy, Nature Communications, 2018). Claim 8. While Gautier_2021 teaches integrated circuit system design objectives and while Sugimura_2009 teaches centrifugal fan design objectives . While both Gautier_2021 and Sugimura_2009 teach the use of multi-armed bandit approach to optimizing generalized systems and accordingly it may be properly found that it would have been obvious to those of ordinary skill in the art to try such an approach on an optical system, Gautier_2021 and Sugimura_2009 do not explicitly teach to apply the multi-armed bandit solution to optical system design objectives. Nevertheless, Durand_2018 makes obvious “ wherein the design objective comprises an optical system design objective ” ( page 2: “Super-resolution techniques have revolutionized the field of optical microscopy … tuning of many parameters, such as laser excitation and depletion power, pixel size, scanning speed, detector gating, and illumination scheme … we propose here an online machine learning approach to improve the performance of optical nanoscopy by addressing an online optimization problem, where the aim is to maximize the outcome (here objectives) during the real imaging phase. We format this problem under the multi-armed bandit’s framework … we achieve multi-objective (MO) optimization…” ; page 10 diagram a: illustrates using multi-armed bandit methods with optical microscope.). Gautier_2021 and Durand_2018 are analogous art because they are from the same field of endeavor called parameter optimization. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Gautier_2021 and Durand_2018. The rationale for doing so would have been that Gautier_2021 teaches to use multi-armed bandit approach to optimize parameters in a design space and Durand_2018 teaches to use the multi-armed bandit approach to optimize parameters for an optical microscopy system. Therefore, it would have been obvious to combine the algorithm and code for design space exploration taught by Gautier_2021 with the optical system taught by Durand_2018 for the benefit of having an executable algorithm that optimizes design parameter to obtain the invention as specified in the claims. Claims 17 - 20 are rejected under 35 U.S.C. 103 as being unpatentable over Gautier_2021 in view of Sugimura_2009 in view of Balakrishnan_2020 (US 2020/0019871 A1). Claim 17 . The limitations of claim 17 are substantially the same as those of claim 1 and are therefore rejected due to the same reasons as outlined above for claim 1. While Gautier_2021 clearly teaches computer code (see algorithm 1, 2, and 3) and while this would have clearly implied to those of ordinary skill in the art “ An optimization system comprising: A computer system, the computer system further comp r ising: a t least one processor; A graphical user interface; and A computer-usable medium embodying computer program code, the computer-usable medium capable of communicating with the at least one processor, the computer program code comprising instructions executable by the at least one processor and configured for: ” as claimed, Gautier_2021 does not explicitly recite these elements. Nevertheless, Balakrishnan_2020 makes obvious “An optimization system comprising: A computer system, the computer system further comprising: at least one processor; A graphical user interface; and A computer-usable medium embodying computer program code, the computer-usable medium capable of communicating with the at least one processor, the computer program code comprising instructions executable by the at least one processor and configured for:” ( FIG. 1, FIG. 3, FIG. 4, FIG. 7, FIG. 10; par 3: “… a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a recommendation component that can recommend a decision based on one or more decision policies… the computer executable components can further comprise an explanation component that can generate an explanation of the decision…”; par 36: “… a user interface (e.g., graphical user interface (GUI), form-based interface, natural language interface, command line, documentation GUI, etc …”; par 41 : “… selection component 114 can comprise a user interface (e.g., a graphical user interface (GUI)… that can facilitate receiving input…”; par 109: “the system memory 1016 can also include volatile memory 1020 and nonvolatile memory 1022… BIOS… computer 1012 can also include removable/non-removable… computer storage media… disk storage…” ). Gautier_2021 and Balakrishnan_2020 are analogous art because they are from the same field of endeavor called decision support/recommendation system. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Gautier_2021 and Balakrishnan_2020. The rational would have been that Gautier_2021 teaches to have a design space exploration algorithm/method that uses a multi-bandit approach and Thompson sampling to optimize parameters and illustrates the algorithm using code (see FIG. 3). Optimizing parameters based on a design space exploration is a recommendation system. Balakrishnan_2020 teaches to use a computer to execute code that perform a recommendation system based on a multi-bandit operation that uses Thompson sampling (see Par 26: “… to facilitate performance of such operations described above, recommendation system… can employ one or more heuristic techniques… to address the exploration-exploitation dilemma in a multi-armed bandit setting (e.g., a constrained contextual multi-armed bandit setting) … can employ a Thompson sampling algorithm…”). Therefore, it would have been obvious to combine the algorithm of Gautier_2021 with the computer system of Balakrishnan_2020 the benefit of having a computer processor upon which the algorithm can be executed to obtain the invention as specified in the claims. Claim 18. Gautier_2021 makes obvious “ further comprising: p roviding at least one design parameter, the at least one design parameter comprising one of: a discrete parameter; and a continuous parameter ” ( page 3 section 2.1: “… Knobs take 𝑛 values, for 𝑛 ∈ [2, ∞ ]. They can be discrete, categorical, or continuous…”). Claim 19. Gautier_2021 makes obvious “ wherein the method for solving the multi-armed bandit problem comprises Thompson sampling further comprising sampling using batch; computing a success rate; and updating Thompson parameters ” ( page 9 algorithm 3; Page 10 Fig. 4; Page 10: “… In this case, we consider each model as a bandit. The outcome of observing one bandit is either an improvement in the current Pareto set, or no improvement. In other words, we are trying to learn a Bernoulli distribution for each model. Consequently, we can select the prior distribution of the bandits as a Beta distribution. We define the prior distribution with parameter 𝜃 for each model 𝑖 as 𝑃𝑖 ( 𝜃 ) = 𝐵𝑒𝑡𝑎 ( 𝛼𝑖 , 𝛽𝑖 ). We update these distributions by selecting one bandit and observing the outcome. A good choice of sampling algorithm is Thompson Sampling [29] that provides a good tradeoff between exploration and exploitation [4]. The algorithm draws a random sample from each distribution: 𝜃 ˆ 𝑖 ∼ 𝐵𝑒𝑡𝑎 ( 𝛼𝑖 , 𝛽𝑖 ) ∀𝑖 , then chooses the bandit with the largest sample value. The observation 𝑥 of the selected bandit corresponds to the improvement of hypervolume over the known designs (hypervolume( 𝐾 )), after we sample a design according to a strategy as defined in Section 2.2.3. In other words, if the model 𝑔𝑖 improved the Pareto set, 𝑥 is a positive outcome, i.e., 𝑥 = 1, otherwise 𝑥 = 0. A value of 𝑥 > 0 increases 𝛼𝑖 , while a value of 𝑥 = 0 increases the value of 𝛽𝑖 . As can be seen in Figure 4, by increasing 𝛼𝑖 and holding 𝛽𝑖 constant, the likelihood that the distribution provides are larger value (closer to 1) is increased. Likewise, increasing 𝛽𝑖 makes is more likely that smaller sample value will be selected (closer to 0). We use this updated function to compute the posterior distribution based on the outcome, and use it as prior for the next iteration. Algorithm 3 shows the details of the method and how it integrates with the Sherlock algorithm described in Algorithm 1. Note that we use an optional posterior reshaping factor 𝑟 that changes the variance of the distributions. As a result, increasing the value of 𝑟 favors exploitation over exploration (i.e., the model providing the best outcome gets selected more often), and the policy becomes more greedy . Increasing this value also has the side benefit that each positive outcome is given more consideration, and potential improvements from models later in the sampling process will re-adjust their importance faster. It provides a small chance to switch the most important model during the sampling process. A model selection algorithm is valuable…”). Claim 20. Gautier_2021 makes obvious “ further comprising: providing an error function, the error function defining design objective ” (Page 4: “Since the goal of DSE is to find the Pareto front P… understand the design space around the Pareto font… DSE outputs an estimated Pareto front P. To understand the quality of the estimated Pareto front, a metric is needed to compare the estimated Pareto designs with the actual Pareto front. Average Distance to Reference Set (ADRS) [19] measures the average normalized distance between the estimated Pareto fron P and the actual Pareto front… 0 indicates that every estimated Pareto point is on the actual Pareto front…”; page 11 section 3.1: “… we compute an error metric based on the ground truth design spaces, using the Average Distance to References Set (ADRS) metric [19] ADRS measures the average normalized distance between the estimated Pareto front and the reference Pareto front. The closer it is to 0, the better the estimation is.). Claims 9 – 16 are rejected under 35 U.S.C. 103 as being unpatentable over Gautier_2021 in view of Balakrishnan_2020. Claim 9. Gautier_2021 makes obvious “a computer implemented optimization method comprising: initializing a symbolic tree in a preparation phase; Updating a parameter held by each node in the symbolic tree using samples collected during an epoch in a parameter phase; Evaluating at least one sample down the symbolic tree with Thompson sampling in order to select at least one sample in a Thompson phase; and Updating parameter distributions using the selected at least one sample and incrementing the epoch in a rejection phase” ( page 2: “… we create a selection strategy based on the multi-armed bandit problem that rewards the models directly improving the actual Pareto front ; Page 5 Fig. 2: Pareto Score and choice of sampling method; page 8: “… a popular ensemble model is the Random Forest predictor based on a set of decision trees … Sherlock can use any of these types of learning algorithms . We experiment with Random Forest and Gaussian process as they can generally model complex design spaces…” : page 9 Algorithm 3 : the model selection algorithm , Page 9: “… we propose to learn the best model using a multi-armed bandit strategy that iteratively updates the importance of each model based on the Pareto set improvement…” ; page 9-10 section 2.4; page 10 FIG 4 ); Balakrishnan_2020 makes obvious “computer implemented” ( FIG. 1, FIG. 3, FIG. 4, FIG. 7, FIG. 10; par 3: “… a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a recommendation component that can recommend a decision based on one or more decision policies… the computer executable components can further comprise an explanation component that can generate an explanation of the decision…”; par 36: “… a user interface (e.g., graphical user interface (GUI), form-based interface, natural language interface, command line, documentation GUI, etc …”; par 41 : “… selection component 114 can comprise a user interface (e.g., a graphical user interface (GUI)… that can facilitate receiving input…”; par 109: “the system memory 1016 can also include volatile memory 1020 and nonvolatile memory 1022… BIOS… computer 1012 can also include removable/non-removable… computer storage media… disk storage…”). Gautier_2021 and Balakrishnan_2020 are analogous art because they are from the same field of endeavor called decision support/recommendation system. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Gautier_2021 and Balakrishnan_2020. The rational would have been that Gautier_2021 teaches to have a design space exploration algorithm/method that uses a multi-bandit approach and Thompson sampling to optimize parameters and illustrates the algorithm using code (see FIG. 3). Optimizing parameters based on a design space exploration is a recommendation system. Balakrishnan_2020 teaches to use a computer to execute code that perform a recommendation system based on a multi-bandit operation that uses Thompson sampling (see Par 26: “… to facilitate performance of such operations described above, recommendation system… can employ one or more heuristic techniques… to address the exploration-exploitation dilemma in a multi-armed bandit setting (e.g., a constrained contextual multi-armed bandit setting) … can employ a Thompson sampling algorithm…”). Therefore, it would have been obvious to combine the algorithm of Gautier_2021 with the computer system of Balakrishnan_2020 the benefit of having a computer processor upon which the algorithm can be executed to obtain the invention as specified in the claims. Claim 10. Gautier_2021 makes obvious “wherein the preparation phase further comprises: generating a tree node with two sets of distributions, wherein each tree node contains a Thompson Distribution” (page 9-10 section 2.4; page 10 FIG 4) Balakrishnan_2020 makes obvious “wherein the preparation phase further comprises: generating a tree node with two sets of distributions, wherein each tree node contains a Thompson Distribution” ( FIG. 2 “Thompson Sampling”; page 26, 27, 30). Claim 11. Gautier_2021 makes obvious “ wherein each node contains a plurality of parameter priors for each of its respective parameters ” page 9-10 section 2.4: “… the prior distributions of the bandits as a Beta distribution. We define the prior distribution with parameter θ … and use it as prior for the next iteration. Algorithm 3 shows the details of the method…”). Balakrishnan_2020 makes obvious “ wherein each node contains a plurality of parameter priors for each of its respective parameters ” (FIG. 2 “Thompson Sampling”; page 26, 27, 30). Claim 12. Gautier_2021 makes obvious “ Wherein the parameter phase further comprises: determining a batch size and an error value for the epoch ” (page 2: “… design space exploration (DSE) framework that uses active learning to evaluate and intelligently explore the HLS design space. Sherlock can quickly reach the set of Pareto optimal designs by minimizing the initialization size, and performing sample selection … using a strategy that balances exploration and exploitation…”; FIG. 2: pareto score/increase/decrease exploit/explore “ sample”… ”; page 6: “… 2.2.3 Sample Selection… decide at every iteration the index I of the next candidate design to sample…” page 11: “… the number of samples …”). Balakrishnan_2020 makes obvious “Wherein the parameter phase further comprises: determining a batch size and an error value for the epoch” (FIG. 2 “Thompson Sampling”; page 26, 27, 30). Claim 13 . Gautier_2021 makes obvious “ wherein the parameter phase further comprises: setting a batch size to be a number of samples taken in each rejection phase ” (page 2: “… design space exploration (DSE) framework that uses active learning to evaluate and intelligently explore the HLS design space. Sherlock can quickly reach the set of Pareto optimal designs by minimizing the initialization size, and performing sample selection … using a strategy that balances exploration and exploitation…”; FIG. 2: pareto score/increase/decrease exploit/explore “ sample”… ”; page 6: “… 2.2.3 Sample Selection… decide at every iteration the index I of the next candidate design to sample…” page 11: “… the number of samples …”). Claim 14 . Gautier_2021 makes obvious “ wherein the parameter phase further comprises: updating parameter distributions using saved samples and incrementing the epoch ” (Algorithm 3, Figure 4). Claim 15. Gautier_2021 makes obvious “ wherein the rejection phase further comprises: evaluating an error function for a selected path on the symbolic tree ” (Figure 2, Algorithm 3). Claim 16. Gautier_2021 makes obvious “ wherein the error function defines a design objective ” ( Page 4: “Since the goal of DSE is to find the Pareto front P… understand the design space around the Pareto font… DSE outputs an estimated Pareto front P. To understand the quality of the estimated Pareto front, a metric is needed to compare the estimated Pareto designs with the actual Pareto front. Average Distance to Reference Set (ADRS) [19] measures the average normalized distance between the estimated Pareto fron P and the actual Pareto front… 0 indicates that every estimated Pareto point is on the actual Pareto front…”; page 11 section 3.1: “… we compute an error metric based on the ground truth design spaces, using the Average Distance to References Set (ADRS) metric [19] ADRS measures the average normalized distance between the estimated Pareto front and the reference Pareto front. The closer it is to 0, the better the estimation is.). Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to FILLIN "Examiner name" \* MERGEFORMAT BRIAN S COOK whose telephone number is FILLIN "Phone number" \* MERGEFORMAT (571)272-4276 . The examiner can normally be reached FILLIN "Work Schedule?" \* MERGEFORMAT 8:00 AM - 5:00 PM . Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, FILLIN "SPE Name?" \* MERGEFORMAT Emerson Puente can be reached at FILLIN "SPE Phone?" \* MERGEFORMAT 571-272-3652 . 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. /BRIAN S COOK/ Primary Examiner, Art Unit 2187