A New Wavelet Neural Network with Boundary Value Constraints
Lu Ran, Shurong Li, Yulei Ge · 2018
A new wavelet neural network (WNN) with boundary value constraints (BVC) is proposed in this paper. The network has a novel topology, which is able to satisfy a series of BVC automatically. The difference of the proposed network with conventional WNN is that the proposed BVC- WNN considers not only the prior knowledge, but also the observation data. Here the prior knowledge means the BVC. The orthogonal least squares algorithm (OLS) is applied to identify the BVC- WNN. And the ability of satisfying BVC is discussed in this paper. By comparing the simulation result for a numerical example using BVC-WNN and conventional WNN, the experimental result proves that the BVC-WNN has better accuracy.