RDE-OP: A Region-Based Differential Evolution Algorithm Incorporation Opposition-Based Learning for Optimising the Learning Process of Multi-layer Neural Networks

Abstract

Learning in multi-layer neural networks (MLNNs) involves finding appropriate weights and biases and is a challenging and important task since the performance of MLNNs is directly dependent on the weights. Conventional algorithms such as back-propagation suffer from difficulties including a tendency to get stuck in local optima. Population-based metaheuristic algorithms can be used to address these issues. In this paper, we propose a novel learning approach, RDE-OP, based on differential evolution (DE) boosted by a region-based scheme and an opposition-based learning strategy. DE is a population-based metaheuristic algorithm which has shown good performance in solving optimisation problems. Our approach integrates two effective concepts with DE. First, we find, using a clustering algorithm, regions in search space and select the cluster centres to represent these. Then, an updating scheme is proposed to include the clusters in the current population. In the next step, our proposed algorithm employs a quasi-opposition-based learning strategy for improved exploration of the search space. Experimental results on different datasets and in comparison with both conventional and population-based approaches convincingly indicate excellent performance of RDE-OP.