Evolutionary Computing Optimization Using E-Dominance Archive
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing evolutionary computing based optimization techniques (ECBOTs) face challenges with multi-objective problems, particularly experiencing dominance resistance and deterioration in convergence rate as the number of objectives increases, leading to inefficiencies in finding solutions along the entire tradeoff surface, especially for problems with four or more objectives.
Innovation Solution
The implementation of an ε-dominance archive, auto-adaptive operators, detection of search stagnation through ε-progress, and adaptive population sizing, along with randomized restarts and multiple recombination operators, enhances the ability to maintain convergence and diversity in multi-objective optimization problems, particularly for those with four or more objectives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing ECBOTs are applied to multi-objective problems with four or more objectives, then the proportion of locally non-dominated solutions becomes large, but the convergence rate degrades to be no better than random search due to dominance resistance
Solution Approach 1:
The patent transforms the traditional Pareto dominance relation into an ε-dominance relation by introducing a parameter ε. This parameter change modifies the dominance criterion: a solution u ε-dominates v if ui ≤ vij + ε for all i and ui < vij for at least one j. This transformation reduces the proportion of non-dominated solutions when the number of objectives increases, thereby maintaining convergence rate without degrading to random search performance.
2Adaptability or versatility
If selection mechanism promotes diverse solutions along the entire extent of the tradeoff surface, then diversity is improved, but deterioration occurs when solution set at time i contains solutions dominated by solutions at time j
Solution Approach 1:
The patent implements a dynamic archive management mechanism where the archive of non-dominated solutions is continuously updated and maintained. When a new solution is generated, it is compared against the archive using ε-dominance. If the new solution ε-dominates existing archive members, those members are removed. This dynamic update process ensures that the archive always contains high-quality diverse solutions without including dominated solutions, thus preventing deterioration while maintaining diversity along the tradeoff surface.
3Reliability
If ε-dominance is used to eliminate dominance resistance, then convergence is improved, but solutions along the entire extent of the tradeoff surface are not produced
Solution Approach 1:
The patent employs a feedback mechanism through the iterative update of the archive and the adaptive adjustment of ε. The algorithm monitors the distribution of solutions in the archive and adjusts ε dynamically to ensure both convergence and diversity. When the archive contains solutions clustered in certain regions, the feedback mechanism increases ε to allow broader exploration, thereby producing solutions along the entire extent of the tradeoff surface while maintaining convergence through the ε-dominance criterion.
Data Source
AI summary
Some embodiments for performing evolutionary based optimization feature: a dominance archive, auto-adaptive operators, detection of search stagnation, exploitation of randomized restarts to escape local optima, and/or selection of recombination operators based on their success in generating high quality solutions.