A Bivariate Compactly Support Non-tensor Product Adaptive Classified Pre-wavelet Neural Network

摘要

In this paper, we proposed a bivariate non-tensor product adaptive classified pre-wavelet neural network (ACPWNN), which can virtually make use of the learning ability of the network to select wavelets to self-adaptively search for the optimal classification feature and performance. And we put forward effective initialization and training algorithms for ACPWNN.