Adaptive Particle Swarm Optimizer With Kernel Density-Based Decoupled Exploration and Exploitation for Large-Scale Optimization

Large-scale optimization poses significant challenges for evolutionary algorithms due to the curse of dimensionality and the difficulty of balancing exploration and exploitation. This paper proposes an Adaptive Particle Swarm Optimization with Kernel Density-Based Decoupled Exploration and Exploitation (APSODEE-KD) algorithm to tackle these issues. By proposing a novel kernel density estimation, the algorithm quantifies local crowding to guide swarm distribution analysis. A tournament selection strategy, informed by this density, decouples exploration and exploitation, enabling particles to learn from both convergence and diversity exemplars. An adaptive grouping mechanism further adjusts competition intensity over time. Experiments on two large-scale benchmark suites show that APSODEE-KD outperforms state-of-the-art methods.