The impact of VMAX activation function in particle swarm optimization neural network

Abstract

Back propagation (BP) Network is the most common technique in Artificial Neural Network (ANN) learning. However, major disadvantages of BP are its convergence rate is relatively slow and always being trapped at the local minima. Therefore, latest optimization technique, Particle Swarm Optimization (PSO) is chosen and applied in feed forward neural network to enhance the network learning. In conventional PSO, maximum velocity, Vmax serves as a constraint that controls the maximum global exploration ability PSO can have. By setting a too small maximum velocity, maximum global exploration ability is limited and PSO will always favor a local search no matter what the inertia weight is. By setting a large maximum velocity, PSO can have a large range of exploration ability. Therefore, in this study, different activation functions will apply in the PSO Vmax function in order to control global exploration of particles and increase the convergence rate as well as correct classification. The preliminary results show that Vmax hyperbolic tangent function give promising results in term of convergence rate and classification compared to Vmax sigmoid function and standard Vmax function.