Particle Swarm Optimization-Based Extreme Learning Machine: A Single Hidden Layer Neural Network Approach

Particle Swarm Optimization-Extreme Learning Machine (PSO-ELM) is a single hidden layer neural network model that integrates Particle Swarm Optimization (PSO) algorithm with Extreme Learning Machine (ELM). This hybrid approach enhances the network's generalization capability and training efficiency by optimizing ELM's initial weights and biases through intelligent swarm-based search. In code implementation, PSO operates by maintaining a population of candidate solutions (particles) that iteratively update their positions based on personal and global best experiences.

In traditional Extreme Learning Machines, the weights and biases of hidden layer neurons are randomly initialized, which can lead to unstable training outcomes. The PSO algorithm effectively searches for optimal solutions by simulating social behavior patterns, thereby optimizing ELM parameters to improve classification accuracy and convergence speed. The optimization process typically involves defining a fitness function that evaluates model performance metrics like mean squared error or classification accuracy.

This model finds applications in data classification, fault diagnosis, and similar domains, particularly suited for handling high-dimensional data and nonlinear problems. Compared to conventional ELM, PSO-ELM demonstrates superior parameter optimization capabilities, effectively avoiding local optima while enhancing model robustness. The implementation commonly requires setting PSO parameters such as swarm size, inertia weight, and acceleration coefficients to balance exploration and exploitation during optimization.