An evolutionary algorithm for variable-sized ANN ensemble construction
Lin‐Yu Tseng, Wen‐Ching Chen · 2010
For some difficult problems, artificial neural network (ANN) ensemble classifiers, instead of a single ANN classifier, are considered. The ensemble usually has better generalization performance than any individual network for classification problems. But, it is not easy to construct the ANN ensemble. In the previous study, the authors presented the systematic trajectory search algorithm (STSA) to train the ANN and the experimental results revealed the performance of the STSA is good. Based on the STSA, an evolutionary algorithm for variable-sized ANN ensemble construction, called the VSEC, is proposed in this paper. A penalty term is added to the error function in order to guarantee the diversity of the ensemble. Besides, a variable-sized ANN ensemble construction method, based on three operations, is proposed to update the ensemble members. The performance of the proposed algorithm is evaluated by applying it to train a class of feedforward neural networks to solve the large n-bit parity problems. By comparing with the previous studies, the experimental results revealed that the neural network ensemble classifiers trained by the VSEC have very good classification ability.