针对传统ART2型神经网络的缺点,提出了一种增强了网络执行速度的改进的ART2型神经网络。改进后的算法避免了传统ART2因输入次序不同而导致的输出结果不同的缺陷。应用了一种新的方法计算输入模式与所有模式的相似度。为了解决传统ART2型神经网络的模式漂移问题引入了激活深度的概念。改善了ATR2型神经网络的适用性。
This paper indicates the shortage of standard ART2 neural network.A simplified ART2 network structure is presented to enhance speed of network performance.The simplified network avoids the different results of standard ART2 neural network because of inputting different sequences.And the paper uses a new method to compute similarity.In order to solve the pattern excursion problem of ART2 neural network,it indicates enabled depth to avoid the problem.It improves the applicative effect of ART2 neural network.