Information Processing with Learned Features for Simulation Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Simulating or experimenting with multi-dimensional output data requires extensive computational resources due to the need for significant calculation processing to update input parameters.
Innovation Solution
An information processing apparatus and method that acquires output data, extracts feature amounts using a learned model, generates reference feature amounts based on similarity and physical quantities, and iteratively determines input parameters until a predetermined condition is met, optimizing the process with or without user knowledge and reference images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional simulation methods are used for multi-dimensional output data, then simulation accuracy is maintained, but enormous calculation processing is required
Solution Approach 1:
The patent replaces conventional mechanical calculation-based simulation methods with a machine learning model that has learned optimal input parameters through training. Instead of performing enormous calculation processing during simulation, the system uses the trained model to directly determine optimal input parameters, substituting computational mechanics with learned patterns from data.
Solution Approach 2:
The system performs preliminary training to build a machine learning model that captures the relationship between input parameters and multi-dimensional output data. This preliminary action stores the knowledge needed for optimization, so that during actual simulation, the system can quickly query the model rather than performing extensive calculations each time.
2Productivity
If conventional simulation methods are used for multi-dimensional output data, then simulation completeness is maintained, but enormous calculation processing is required
Solution Approach 1:
The patent substitutes traditional calculation-based simulation with a machine learning approach. The system trains a model on simulation data to learn the mapping from input parameters to multi-dimensional output, then uses this model to efficiently determine optimal input parameters without repeating enormous calculations.
Solution Approach 2:
The system creates a copy of the simulation knowledge by training a machine learning model on simulation data. This model copy captures the essential relationships and can be queried repeatedly without performing the original complex simulations each time, significantly improving productivity.
3Extent of automation
If feature amounts are manually determined, then user knowledge is utilized, but the process requires significant manual input
Solution Approach 1:
The system enables automated feature amount generation where the machine learning model automatically determines relevant features and their importance based on the training data. This self-service approach reduces manual input requirements while still utilizing user knowledge through the training process, allowing the system to autonomously identify important features for optimization.
Data Source
AI summary
An information processing apparatus comprises processing circuitry that acquires output data obtained by performing an experiment or simulation based on an input parameter and a physical quantity of the output data, inputs the output data to a learned model, extracts a feature amount of the output data, generates a first reference feature amount, generates a second reference feature amount based on the physical quantity of the output data and a degree of similarity between the first reference feature amount and the feature amount of the output data, calculates degree of similarity between the second reference feature amount and the feature amount of the output data, sets an evaluation value based on the calculated degree of similarity and the physical quantity of the output data, determines an input parameter for a next experiment or simulation based on the evaluation value, and repeats the above processings until a predetermined condition is satisfied.