A Differential Evolution based Memetic Algorithm for Neural Networks Designing

L. Zhang (PRC)


Optimization, Differential Evolution, Memetic algorithm, Neural networks designing


Memetic algorithms have stimulated research interests in many academic fields, due to its inclusion of the choice of local search methods so as to significantly affect the efficiency of problem searches. Based on the ideas that a local search with multiple faster training algorithms can enrich the local searching behavior and to avoid premature convergence, this paper proposes an effective Differential Evolution algorithm (DE) based memetic algorithm (MA) for designing artificial neural network. In the proposed DE-based MA (DEMA), the evolutionary searching mechanism of DE is used to perform global exploration. Adaptive high-performance faster training algorithms are employed to enhance the local exploitation search. In order to well balance the exploration and exploitation abilities of DEMA, an effective adaptive Meta-Lamarckian learning strategy is employed to decide which local search method to be used so as to concentrate computing effort on promising neighbour solutions. Simulation results and comparisons demonstrate the effectiveness and efficiency of the pro-posed method.

Important Links:

Go Back