A Nested Neighborhood PSO Algorithm for Multi-modal Function Optimization

In PSO algorithm, a nested neighborhood model was proposed for multi-modal function optimization. First, a local structure was constructed for each particle under a predefined Euclidian distance measure(called peak-granularity). Then a nesting mechanism was presented to connect the local-best particles resulting a nested neighborhood chain. Some concepts, neighbor operator, peak neighbor and nesting counter were introduced to describe the mechanism. Based upon the mechanism, local-best information can be shared by all the particles on the neighborhood chain. This effectively improves the convergence performance of the particle swarm. And the peak-granularity can be kept small to locate more optimal basins of the function. Several benchmark functions are adopted to verify the performance.