Bayesian Optimization Parameter Regeneration for Simulation Series

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Bayesian optimization methods can be time-consuming when searching for input parameters in series of simulations, particularly when sub-simulations have different processing periods, leading to prolonged calculation times and reduced efficiency.

Innovation Solution

An information processing system with a parameter provision apparatus that includes a calculator and a generator, which calculates evaluation values and regenerates input parameters based on selected output parameters, adjusting the number of input parameters for each sub-processing to optimize the total processing time by executing sub-processing with shorter periods more frequently.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If Bayesian optimization is performed on series of simulations with multiple sub-simulations, then search accuracy is improved, but total calculation time increases enormously

Engineering Contradiction:

Improvesearch accuracyVSAvoidtotal calculation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the series of simulations into multiple sub-simulations with different processing periods. By identifying and separating sub-simulations with shorter processing periods from those with longer periods, the system can execute the shorter sub-simulations multiple times within one iteration cycle, thereby improving search accuracy without proportionally increasing total calculation time.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent dynamically adjusts the number of executions for each sub-simulation based on its processing period. Sub-simulations with shorter periods are executed more frequently (multiple times per iteration), while sub-simulations with longer periods are executed less frequently. This dynamic execution strategy optimizes the balance between search accuracy and total calculation time.

Inventive Principle:

Principle #15Dynamics

2Productivity

If the number of simulation iterations is reduced to shorten processing time, then calculation efficiency is improved, but search accuracy degrades

Engineering Contradiction:

Improvecalculation efficiencyVSAvoidsearch accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

By segmenting the simulation series into sub-simulations with different processing periods, the system can execute shorter sub-simulations multiple times within a single iteration. This segmentation allows the system to maintain effective sample size and search accuracy while reducing the total number of iteration cycles needed, thereby improving calculation efficiency.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by executing only certain sub-simulations (those with shorter processing periods) multiple times within one iteration, rather than executing all sub-simulations the same number of times. This selective repeated execution of specific sub-simulations provides sufficient data for accurate parameter search without requiring a proportional increase in total iterations.

Inventive Principle:

Principle #16Partial or excessive action

3Device complexity

If sub-simulations with different processing periods are executed sequentially, then system complexity is reduced, but total processing period increases

Engineering Contradiction:

Improvesystem complexityVSAvoidtotal processing period

Core Design Contradiction:

Device complexityVSDuration of action of moving object

Solution Approach 1:

The patent introduces dynamic execution scheduling where sub-simulations with shorter processing periods are executed multiple times within one iteration cycle, while sub-simulations with longer periods are executed once. This dynamic approach reduces the total processing period without requiring complex parallel processing infrastructure, thus maintaining relatively simple system architecture.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements periodic action by executing shorter sub-simulations repeatedly within each iteration cycle. This periodic re-execution of specific sub-simulations allows the system to gather more evaluation data without proportionally increasing the number of iteration cycles, thereby reducing total processing time while keeping system complexity manageable.

Inventive Principle:

Principle #19Periodic action

Data Source

PatentUS12197942B2Information processing apparatus, information processing method, non-transitory storage medium, and information processing system

Publication Date: 2025.01.14 KK TOSHIBA

AI summary

One embodiment of the invention provides an apparatus preventing degradation of accuracy of a search result while reducing a total required period of a series of processing in a case where input parameters for a series of processing including a plurality of processing with different required periods are searched for. The apparatus includes a calculator and a generator. The calculator calculates evaluation values for output parameters of a series of processing including first processing and second processing. The first processing uses a first input parameter. The second processing use a second input parameter. The generator regenerates first and second input parameters corresponding to one time of a series of processing based on first and second input parameters corresponding to selected output parameters. The number of input parameters for shorter one of the first and the second processing is larger than the number of input parameters for the other.