Robust Constraint Satisfaction in Bayesian Optimization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing optimization methods, such as Bayesian optimization, may deviate from optimal parameter settings due to various factors, leading to suboptimal performance in manufacturing and other systems, requiring efficient methods to minimize influence on objective and constraint function values while ensuring robust constraint satisfaction without multiple simulations.

Innovation Solution

An information processing device estimates a robust satisfaction probability by calculating the probability that parameter settings within a neighborhood range satisfy a robust constraint, allowing for efficient determination of optimal parameter values using Bayesian optimization and robust acquisition functions, reducing the need for multiple simulations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple simulations are executed to determine whether set values satisfy constraints, then constraint satisfaction accuracy is improved, but processing time and computational cost increase

Engineering Contradiction:

Improveconstraint satisfaction accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies preliminary action by estimating constraint function values and their uncertainties before actual simulation execution. The system predicts which parameter settings are likely to satisfy constraints based on preliminary analysis, allowing it to focus computational resources only on critical cases that require full simulation verification, thus reducing overall processing time while maintaining accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating surrogate models or approximations of the constraint functions based on limited simulation data. These copied representations allow rapid evaluation of constraint satisfaction for many parameter settings without executing full simulations, reserving actual simulations only for final verification of promising candidates

Inventive Principle:

Principle #26Copying

2Reliability

If optimization methods account for deviations from optimal parameter settings, then robustness of constraint satisfaction is improved, but calculation complexity increases

Engineering Contradiction:

Improverobustness of constraint satisfactionVSAvoidcalculation complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies parameter changes by transforming the original constraint satisfaction problem into a probabilistic framework where constraint functions are characterized by their expected values and uncertainties. This transformation allows the use of statistical methods to evaluate robustness, converting a complex deterministic optimization problem into a more tractable probabilistic formulation that naturally accounts for parameter deviations

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces an intermediary probabilistic model that acts as a mediator between the uncertain parameter settings and the constraint requirements. This intermediary layer estimates the probability of constraint satisfaction by combining preliminary constraint evaluations with uncertainty quantification, simplifying the direct assessment of robustness under parameter variations

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If fewer simulation trials are used to determine optimal parameters, then productivity is improved, but determination accuracy of optimal values deteriorates

Engineering Contradiction:

Improveefficiency of parameter optimizationVSAvoidaccuracy of optimal parameter determination

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent employs copying by constructing surrogate models that replicate the behavior of the full simulation system based on limited data. These surrogate models enable rapid evaluation of many parameter settings, allowing the system to identify promising candidates efficiently while using actual simulations only for final verification, thus maintaining accuracy with fewer trials

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent implements feedback by using the results of limited simulations to update and refine the surrogate models iteratively. Each simulation trial provides feedback that improves the accuracy of the probabilistic predictions, allowing the system to progressively converge on optimal parameter settings with high confidence using fewer total trials than traditional methods would require

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20240045923A1Information processing device, information processing method, and computer program product

Publication Date: 2024.02.08 KK TOSHIBA

AI summary

According to an embodiment, an information processing device includes one or more hardware processors configured to: based on a constraint function value that is an output of a constraint function when one or more first set values are input for one or more parameters, estimate a first estimation value and a first estimation error of the constraint function value; and based on the first estimation value and the first estimation error, calculate a robust satisfaction probability representing a probability that one or more second set values for the one or more parameters satisfy a robust constraint that the constraint function value when inputting a plurality of third set values should satisfy, the plurality of third set values being obtained by changing the one or more second set values in a neighborhood range determined in advance based on the one or more second set values.