Peak Identification in Evolutionary Multimodal Optimization: Model, Algorithms, and Metrics.

Area of Science:

• Computational Science
• Optimization Algorithms
• Machine Learning

Background:

• Multimodal optimization aims to find multiple solutions in complex search spaces.
• Existing algorithms often suffer from redundant solutions, reducing efficiency.
• Accurate identification of distinct optima is crucial for many applications.

Purpose of the Study:

• To develop a two-phase multimodal optimization model for efficient and accurate identification of multiple optima.
• To introduce a novel peak identification (PI) procedure to filter non-optimal and redundant solutions.
• To propose and evaluate two specific PI algorithms: HVPI and HVPIC.

Main Methods:

• A population-based search algorithm is used in the first phase to locate potential optima.
• A novel peak identification (PI) procedure, including HVPI and HVPI with bisecting K-means clustering (HVPIC), is employed in the second phase.
• Performance is evaluated using the F-measure, assessing both accuracy and redundancy.

Main Results:

• The proposed PI algorithms effectively filter out non-optimal and redundant solutions.
• HVPI and HVPIC demonstrated high precision and recall in identifying distinct optima.
• Extensive experiments on benchmark functions and engineering problems showed significant outperformance compared to traditional methods.

Conclusions:

• The presented two-phase multimodal optimization model successfully addresses the challenge of redundant solutions.
• The novel PI algorithms offer an effective approach for accurate multimodal optimization.
• The findings suggest improved performance and efficiency for complex optimization problems.