Kernel Density Parameter Identification for Faster Optimization Tuning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing optimization algorithms, such as Simulated Annealing (SA) and Tabu Search (TS), face challenges in efficiently identifying optimal parameter values for combinatorial optimization problems, particularly in terms of solution performance and tuning time.

Innovation Solution

A computer-readable recording medium stores a parameter identification program that generates a probability density function by combining kernel functions from current and past instances to identify parameter values, using a kernel density estimation method that incorporates historical optimization results to improve parameter tuning efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional parameter tuning methods (repeated setting and evaluation) are used, then parameter values can be identified, but tuning time is excessively long

Engineering Contradiction:

Improveparameter identification accuracyVSAvoidtuning time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by collecting and storing parameter values from multiple past instances before processing the current instance. Historical parameter values are pre-processed and stored in a database, enabling the current parameter identification to leverage accumulated knowledge without repeating exhaustive search processes, thus reducing tuning time while maintaining accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system creates a probabilistic model (probability density function) that copies and generalizes patterns from historical parameter values. By modeling the distribution of optimal parameters across multiple instances, the system can infer parameters for new instances based on learned patterns rather than exhaustive searching, significantly reducing tuning time

Inventive Principle:

Principle #26Copying

2Measurement precision

If parameter values are identified through exhaustive search, then optimal parameters can be found, but solution performance deteriorates due to overfitting or suboptimal convergence

Engineering Contradiction:

Improveparameter optimization accuracyVSAvoidsolution performance

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The system implements feedback by continuously collecting evaluation results from past instances and using them to refine the probability density function. The historical parameter values and their corresponding evaluation results provide feedback that shapes the probabilistic model, enabling the system to learn from past successes and failures to improve solution performance while maintaining identification accuracy

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system changes the approach from deterministic parameter setting to probabilistic parameter selection. By modeling parameters as probability distributions rather than fixed values, the system can explore parameter spaces more effectively and avoid poor local optima, improving solution performance while maintaining accurate parameter identification through the learned distributions

Inventive Principle:

Principle #35Parameter changes

3Productivity

If historical instances are incorporated into parameter identification, then tuning efficiency improves, but system complexity increases

Engineering Contradiction:

Improveparameter tuning efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The system replaces complex mechanical-like exhaustive search processes with a probabilistic modeling approach. Instead of systematically evaluating all possible parameter combinations (mechanical enumeration), the system uses probability density functions to directly estimate optimal parameters based on historical patterns, improving tuning efficiency while the modular architecture keeps system complexity manageable

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system achieves universality by creating a general-purpose probability density function that can handle multiple different instances and parameter types. The same probabilistic modeling framework works across diverse optimization problems, allowing the system to efficiently handle various tasks without requiring instance-specific complex configurations, thus improving productivity without proportionally increasing complexity

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20250322029A1Computer-readable recording medium having stored therein parameter identification program, parameter identification method, and information processing apparatus

Publication Date: 2025.10.16 FUJITSU LTD

AI summary

A computer-readable recording medium having stored therein a parameter identification program causes a computer to execute a process includes generating a probability density function for each of a plurality of parameters used in an optimization algorithm by combining a kernel function generated from a first value of the parameter observed in a given instance, and a kernel function generated from a second value of the parameter identified in each of a plurality of instances. The process includes identifying respective values of the plurality of parameters based on the probability density function generated for the each of the plurality of parameters.