Competitive Island Cooperative Neuro-evolution of Feedforward Networks for Time Series Prediction

摘要

Problem decomposition, is vital in employing cooperative coevolution for neuro-evolution. Different problem decomposition methods have features that can be exploited through competition and collaboration. Competitive island cooperative coevolution (CICC) implements decomposition methods as islands that compete and collaborate at different phases of evolution. They have been used for training recurrent neural networks for time series problems. In this paper, we apply CICC for training feedforward networks for time series problems and compare their performance. The results show that the proposed approach has improved the results when compared to standalone cooperative coevolution and shows competitive results when compared to related methods from the literature.