本文用计算机仿真研究了一种适于光学实现的非线性神经网络模型的存储客量α_c和寻址能力,提出了一个改进其触突互联矩阵的蒙特卡洛学习算法.数值研究表明,经过学习修正后的神经网络模型的寻址能力及存储容量都有较大的改进. In this paper, we study the storage capacity α_c and addressing ability of a nonlinear neural network model suitable for optical implementation by computer simulation, and propose a Monte Carlo learning algorithm to improve its contact matrix. Numerical results show that after learning The modified neural network model has great improvement in addressing ability and storage capacity.