一个基于机器学习的神经网络初始化方法

A METHOD OF NEURAL NETWORK INITIALIZATION BASED ON MACHINE LEARNING

本文针对ＢＰ（ｂａｃｋｐｒｏｐａｇａｔｉｏｎ）前馈神经网络存在的训练时间长、容易陷入局部极小等问题，研究了一个基于机器学习的神经网络初始化方法．实验结果表明，用这种方法初始化神经网络，提高了神经网络的学习效率和泛化能力，并且可以有效地抑制陷入局部极小的可能性．

In order to solve the problem of long training time and local minima in BP neural network, a method of neural network initialization based on machine learning is studied in this paper. The experiment results show that the learning efficiency and generalization ability of the initialized neural network are increased and the probability of falling to local minima is decreased.