Hare escape optimization algorithm with applications in engineering and deep learning.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Nature-Inspired Computing

Background:

• Metaheuristic algorithms are crucial for solving complex optimization problems.
• Existing algorithms often struggle with balancing exploration and exploitation, leading to local optima.
• Novel approaches are needed to improve convergence speed and solution quality.

Purpose of the Study:

• Introduce the Hare Escape Optimization (HEO) algorithm, a novel metaheuristic inspired by hare evasion tactics.
• Evaluate HEO's performance against state-of-the-art algorithms on benchmark functions and real-world engineering problems.
• Demonstrate HEO's effectiveness in optimizing hyperparameters for Convolutional Neural Networks (CNNs).

Main Methods:

• HEO integrates Levy flight dynamics and adaptive directional shifts for enhanced exploration-exploitation balance.
• Tested HEO on 43 benchmark functions from CEC 2015 and CEC 2020 testbeds.
• Applied HEO to four constrained engineering design problems and CNN hyperparameter optimization.

Main Results:

• HEO demonstrated superior performance on unimodal and multimodal benchmark functions compared to 29 other metaheuristics.
• Achieved better solution feasibility and computational efficiency in engineering design optimization.
• Significantly improved accuracy and convergence speed for CNNs in image classification.

Conclusions:

• HEO is a robust and adaptable metaheuristic optimization tool.
• Its unique search mechanism offers a new perspective for intelligent optimization.
• HEO shows significant promise for applications in engineering and deep learning.