An improved differential evolution algorithm for artificial neural networks
Wei Li, Yu Lei · 2018
DE, which is a well-known intelligence optimization algorithm, has been widely used in solving science and engineering problems. However, the search performance of DE algorithm depends on control parameter settings. It is possible that poor settings for the parameters may cause the population to move towards undesirable values. In order to improve the performance of DE, this paper proposed an improved optimization algorithm based on JADE (MJADE). In MJADE, the number of best individuals selected from the current population is dynamically adjusted to keep a balance between exploration and exploitation. Moreover, a modified mutation operation is designed. To verify the effectiveness of MJADE, numerical experiments are carried on ten benchmark problems from CEC2014. In addition, an application problem of artificial neural network is examined. The experimental results show that MJADE is competitive with respect to other compared algorithms in this simulation example.