摘要
本文提出了一种异联想记忆模型的优化学习算法.首先,我们将反映神经元网络性能的标准转化为一个易于控制的代价函数,从而将权值的确定过程自然地转化为一个全局最优化过程.优化过程采用了梯度下降技术.这种学习算法可以保证每个训练模式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.在理论上,我们讨论了异联想记忆模型的存储能力,训练模式的渐近稳定性和吸引域的范围.计算机实验结果充分说明了算法的有效性.
This paper presents an optimized learning strategy for a bidirectional associative memory. According to a cost function that measures the goodness of the neural network, the problem of determining the connection weights is converted into a global optimiza-tion which is solved by a gradient descent method. The optimized learning rule guarantees to store each traning pattern with basin of attraction as large as possible. The storage capacity, the asymptotic stability, and domain of attraction are examined mathematically. Experimental results convincingly verify the efficiency of the proposed algorithm.
出处
《计算机学报》
EI
CSCD
北大核心
1993年第2期97-105,共9页
Chinese Journal of Computers
关键词
异联想记忆
模型
优化学习算法
Bidirectional associative memories, optimized learning algorithm, cost function, global optimization, asymptotic stability.
