论文部分内容阅读
证明了区间小波神经网络具有一致及Ｌ２逼近性质，且为相容的函数估计子，其学习收敛速度在ｄ维情形不随ｄ增大而减慢，本质上克服了神经网络高维学习的“维数灾难”问题，模拟实例验证了理论的正确性．
It is proved that the interval wavelet neural network has consistent and L2 approximation properties and is a consistent function estimator. Its learning convergence speed does not slow down with increasing d in the d-dimensional case, essentially overcoming the “dimensionality Several catastrophe ”problem, the simulation example verified the correctness of the theory.