Path Relinking Initialization for Diverse Combinatorial Search

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional path relinking methods for combinatorial optimization problems often confine the initial solution to the vicinity of the best solution, limiting the search space and hindering the discovery of diverse and potentially better solutions.

Innovation Solution

A computer program that determines higher-ranked solutions based on evaluation function values, adjusts selection probabilities to favor solutions closer to the best solution, and generates an initial solution using path relinking with a second, diverse solution, thereby expanding the search space.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the best solution is used to generate the initial solution in path relinking, then the initial solution is similar to the best solution, but the search space is limited and diversity is reduced

Engineering Contradiction:
Improvesolution qualityVSAvoidsearch space coverage
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent changes the selection parameter from deterministic (always selecting the best solution) to probabilistic (selecting from multiple high-ranking solutions based on probability). This allows the system to balance between exploiting known good solutions and exploring diverse regions of the search space, resolving the contradiction between solution quality and search space coverage.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces dynamic selection where the choice of initial solution varies probabilistically based on ranking probabilities. Instead of a static rule always selecting the best solution, the system dynamically selects from multiple candidates, enabling adaptability between exploitation and exploration phases of the search.

Inventive Principle:
Principle #15Dynamics

2Adaptability or versatility

If two solutions are randomly selected from the solution pool, then the search space is expanded, but good solutions are searched to the same extent as other solutions, reducing efficiency

Engineering Contradiction:
Improvesearch space coverageVSAvoidsearch efficiency
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent transforms the uniform random selection parameter into a non-uniform probability distribution where selection probability is inversely related to the rank of the solution. This ensures that higher-quality solutions have higher probabilities of being selected, maintaining search efficiency while still allowing diverse solutions to be chosen occasionally for exploration.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If the best solution is always selected, then solution quality is maintained, but diversity in initial solutions is reduced

Engineering Contradiction:
Improvesolution qualityVSAvoidsolution diversity
Core Design Contradiction:
ReliabilityVSStability of the object's composition

Solution Approach 1:

The patent introduces dynamic probabilistic selection that balances reliability and diversity. The selection process is dynamic rather than static, allowing the system to maintain reliability by favoring high-ranking solutions while simultaneously preserving diversity through probabilistic variation in selection outcomes.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4715634A1Computer program, data processing apparatus, and data processing method
Publication Date: 2026.03.25 FUJITSU LTD
  • EP4715634A1 patent drawingFigure 1
  • EP4715634A1 patent drawingFigure 2
  • EP4715634A1 patent drawingFigure 3

AI summary

A processing unit determines, from a plurality of solutions to a combinatorial optimization problem stored in a storage unit, a plurality of higher-ranked solutions based on evaluation function values of the plurality of solutions, determines, for each of the plurality of higher-ranked solutions, a selection probability such that higher-ranked solutions having evaluation function values closer to that of the highest-ranked solution among the plurality of higher-ranked solutions are more likely to be selected, selects a first solution from the plurality of higher-ranked solutions according to the selection probabilities, selects a second solution different from the first solution, from the plurality of solutions, generates a third solution with a path relinking method using the selected first solution and second solution, and performs a solution search on the combinatorial optimization problem using the third solution as an initial solution.