Competitive Hopfield Neural Network with Periodic Stochastic Dynamics for Partitional Clustering

A novel competitive Hopfield network with periodic stochastic dynamics is proposed for the NP-hard partitional clustering problem in this paper. Clustering technique has been applied to a wide range of problems, such as pattern recognition and machine learning. The aim of partitional clustering is to obtain a specified number of data sets from the original data according to certain criteria. The proposed algorithm introduces periodic stochastic dynamics, which helps the neural network escape from local minima and search a possible better solution based on the solution which is obtained in the latest period. The performance is evaluated through several benchmark data sets. The simulation results show that the proposed algorithm outperforms previous approaches, such as k -means, genetic algorithm, particle swarm optimization, differential evolution, combinatorial particle swarm optimization and tabu search.