An enhance multimodal multiobjective optimization genetic algorithm with special crowding distance for pulmonary...

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Evolutionary Computation

Background:

• Multiobjective optimization typically assumes a one-to-one mapping between decision and objective spaces.
• Multimodal multiobjective problems (MMOPs) arise when multiple decision variables map to the same objective value, leading to multiple Pareto Sets (PS) mapping to a single Pareto Front (PF).
• Existing evolutionary computing methods face challenges in efficiently handling the complexities of MMOPs.

Purpose of the Study:

• To propose an enhanced multimodal multiobjective genetic algorithm (ESNSGA-II) designed to effectively solve MMOPs.
• To improve the search capabilities for multiple Pareto Sets (PS) in complex optimization landscapes.
• To enhance both decision and objective space diversity for better convergence and solution accuracy.

Main Methods:

• Development of ESNSGA-II, incorporating a specialized crowding distance calculation considering both decision and objective space diversity.
• Implementation of a novel crossover mechanism combining Simulated Binary Crossover (SBX) with Pareto solution properties to generate offspring.
• Evaluation of ESNSGA-II using the CEC2020 MMF1-MMF8 benchmarks and comparison with state-of-the-art multimodal multiobjective evolutionary algorithms.

Main Results:

• ESNSGA-II demonstrated superior performance in efficiently searching for numerous PSs of MMOPs compared to existing methods.
• The algorithm successfully balanced convergence and diversity in both decision and objective spaces, enhancing PS distribution and PF accuracy.
• Application to a real-world MMOP for pulmonary hypertension detection showed ESNSGA-II outperforming other algorithms.

Conclusions:

• The proposed ESNSGA-II is an effective algorithm for tackling multimodal multiobjective optimization problems.
• ESNSGA-II shows significant potential as a valuable tool for complex real-world applications, including medical diagnostics like pulmonary hypertension detection.