Conditional Genes in Multi-Objective Evolutionary Algorithms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large-scale value-based non-linear models with many decision variables and constraints often overwhelm heuristic algorithms like Multi-objective Evolutionary Algorithms (MOEAs), leading to high computational bandwidth requirements for constraint evaluations, which can be comparable to or exceed objective evaluations.
Innovation Solution
The implementation of conditional genes in MOEAs, where the value of a conditional gene depends on other genes, reduces the number of constraint violations by selecting appropriate hidden genes based on rules, thereby decreasing the computational bandwidth needed for constraint checks and increasing the feasibility of evaluated chromosomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of decision variables and constraints is increased to model large-scale value-based non-linear models, then the model's ability to represent complex real-world problems is improved, but the computational bandwidth required for constraint evaluations increases significantly
Solution Approach 1:
The patent applies preliminary action by performing constraint evaluation checks before full objective function evaluations. The system evaluates whether potential solutions violate constraints using a simplified check that requires less computational bandwidth than complete objective evaluations, thereby filtering out infeasible solutions early in the evaluation process
Solution Approach 2:
The patent segments the evaluation process into two distinct stages: (1) constraint violation evaluation using conditional genes which requires minimal computational resources, and (2) objective function evaluation for feasible solutions. This segmentation allows the system to handle large-scale models by processing constraints separately and more efficiently
2Device complexity
If traditional MOEAs are used without conditional genes, then the algorithm structure is simpler, but the proportion of feasible chromosomes decreases due to high constraint violation rates
Solution Approach 1:
The patent introduces conditional genes as an intermediary mechanism between the decision variables and constraint evaluations. These conditional genes act as mediators that automatically ensure constraint satisfaction by selecting appropriate hidden gene values based on conditional rules, thereby increasing the proportion of feasible chromosomes without significantly complicating the overall algorithm structure
Solution Approach 2:
The patent changes the parameter representation by introducing conditional genes with multiple possible values (hidden genes) that correspond to different constraint satisfaction scenarios. The conditional gene values are dynamically selected based on the states of other genes in the chromosome, allowing the system to maintain feasibility while preserving algorithmic simplicity
3Reliability
If computational bandwidth is allocated primarily to constraint evaluations, then constraint satisfaction is improved, but the time required for objective function evaluations increases
Solution Approach 1:
The patent applies partial action by implementing a two-tiered evaluation approach where only essential constraint checks are performed initially using conditional genes, rather than conducting full constraint and objective evaluations for all chromosomes. This partial evaluation reduces the time spent on constraint checking while maintaining sufficient constraint satisfaction through the conditional gene mechanism
Data Source
AI summary
Systems and methods are provided to engage in multi-objective optimization where there may be potential solutions for evaluation (e.g., chromosomes) that each have one or more conditional genes. The value of each of the conditional genes in each of the chromosomes may be equivalent to one of a plurality of hidden genes in each of the chromosomes. The value of each of the conditional genes may be evaluated prior to determining objective values of each of the chromosomes. The objective values of each of the chromosomes may be used to evaluate the potential solutions embodied in the chromosomes and further drive to more optimized solutions. The use of the conditional genes in the chromosomes may reduce the amount of constraint violation checks that may need to be performed.