论文部分内容阅读
本文提出了一种对称互连神经元网络的学习策略,利用全局约束优化方法确定连接权。优化过程采用了梯度下降技术。这种学习算法可以保证训练样本成为系统的稳定吸引子,并且具有优化意义上的最大吸引域。本文讨论了网络的存储容量,训练样本的渐近稳定性和吸引域大小。计算机实验结果说明了学习算法的优越性。
In this paper, a learning strategy of symmetric interconnected neuronal networks is proposed, which uses the global constraint optimization method to determine the connection rights. The optimization process uses a gradient descent technique. This learning algorithm can ensure that the training sample becomes a stable attractor of the system and has the largest attraction domain in the sense of optimization. This paper discusses the storage capacity of the network, the asymptotic stability of the training samples and the size of the attracting domain. Computer experimental results illustrate the superiority of learning algorithms.