多层感知器三种学习算法的比较

The Compare Between Three Training Algorighms of Multilayer Perceptorns

多层感知器是一种多层前馈神经网络,常用的快速训练算法有共轭梯度法、拟牛顿法。通过模式分类实验对这两种算法和BP算法进行比较,并由试验数据得出这几种算法的复杂性、可靠性,以及由算法产生的多层感知器的泛化能力。

Multilayer Perceptrons is a kind of multilayer feed-forward network, it has two fast training algorighm-conjugate grads method and q-newton method. This paper compare the training speed,time and space cost of those algorighms through classifer experiment.