Adaptive cooperation of multi-swarm particle swarm optimizer-based hidden Markov model

The classical PSO algorithm can be affected with premature convergence when it comes to more complex optimization problems; the resolution easily can be trapped into local optima. The primary concern is to accelerate the convergence speed and to prevent the local optima solutions. To defeat these weaknesses and to enhance the overall performances, a new technique is offered building a dynamic multi-swarm design with cooperative rules based on a machine-learning design, namely the hidden Markov classification model. In this approach, a new design with multiple processes implemented inside the PSO that are the control of parameters adaptively with the improvement in the topological structure by setting a multi-swarm layer. Another process of information exchange between swarms is also considered. According to an HMM classification, the entire swarm will be then divided into dynamic cooperating sub-swarms. The size of each sub-swarm is going to be also adjusted at each iteration to suit the search stage. All sub-swarms share information between each other in order to ensure the best exploration of the search space and most effective exploitation. Adaptiveness of both acceleration coefficient and inertia weight strategies is customized with the account of the multi-swarm dynamic evolution and the history of achievements. The approach is simulated and compared by experimental tests to the best-known state of the art.