论文部分内容阅读
Selective neural network ensemble is an effective way to enhance the generalization accuracy of networks, which is much better than the ensemble of all available networks. However, there are still some problems in the selective ensemble, which are lack of unified definition of diversity among component neural networks (NNs) and hard to improve the accuracy by selecting if the diversities of available NNs are small. Aiming at the above problems, the output errors of NNs are firstly vectorized, the diversity of NNs is defined based on the error vectors and the size of ensemble is analyzed. And then an error vectorization based selective neural network ensemble (EVSNE) is proposed in the paper, in which the error vector of each network can offset that of the other networks by training the component NNs orderly. Thus, the component networks have large diversity. Experiments and comparisons over standard datasets and actual industrial dataset of high-density polyethylene (HDPE) demonstrate that EVSNE performs better in generalization ability with higher accuracy.