如果输出为“0”(即结果错误)，则把网络连接权值朝着减小综合输入加权值的方向调整，其目的在于使网络下次再遇到“A”模式输入时，减小犯同样错误的可能性。如此操作调整，当给网络轮番输入若干个手写字母“A”、“B”后，经过网络按以上学习方法进行若干次学习后，网络判断的正确率将大大提高。这说明网络对这两个模式的学习已经获得了成功，它已将这两个模式分布地记忆在网络的各个连接权值上。当网络再次遇到其中任何一个模式时，能够作出迅速、准确的判断和识别。一般说来，网络中所含的神经元个数越多，则它能记忆、识别的模式也就越多。(If output is "0", that is the result error, then adjust the weight of network connection to reduce the direction of the comprehensive input weighted value. The purpose is to reduce the possibility of making the same mistake when the network enters the "A" mode again. In this way, when we input several handwritten letters "A" and "B" into the network, the accuracy of network judgement will be greatly improved after several times of learning by network. This shows that the network has been successful in learning the two patterns, and it has distributed the two patterns in the network's various connection weights. When the network meets any of them again, it can make quick and accurate judgment and recognition. In general, the more the number of neurons in the network is, the more patterns it can remember and recognize.)