Hybrid Structure Based PSO for ESN Optimization

Abstract

Recently, the echo state networks (ESNs) have been widely studied. In an ESN, the input weights and internal weights of reservoir are fixed after initialization, only the output weight matrix needs to be optimized. To calculate the output weights of ESN, the particle swarm optimization algorithm (PSO) with hybrid topology is proposed. The structure of the proposed PSO is mixed with regular network with strong exploration ability and scale-free network with good exploration ability. Simulation results show that the proposed ESN has good prediction performance than the traditional ESN.