Abstract
This paper presents a dynamic mutation particle swarm optimization (DMPSO). The particle swarm optimization (PSO) is a popular swarm algorithm, which has exhibited good performance on many optimization problems. However, similar to other swarm intelligence algorithms, PSO also suffers from premature convergence. Sub swarm and mutation are widely used strategies in the PSO algorithm to overcome the premature convergence. In the DMPSO, firstly, the population is divided into two groups, and some special operations - dynamic crossover and dynamic mutation are employed in the evaluation. The DMPSO algorithm was used in the training process of neural network, and compared to other three different types of NN training algorithms, and the experiment results show that the new algorithm performances better on two representative types of application of neural network.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have