光学神经网络中硬件误差对性能影响的分析

利用统计方法，分析了光学硬件误差对用Ｈｏｐｆｉｅｌｄ神经作联想记忆时出错率的影响，导了系统的误差与错率关第的近似公式，并给出了模拟结果。得到了一定限度的误差对出错率的影响并不显著的的结论，特别坚探讨测器阵列动态范围的要求，可以远远小于其探测到的最大值。这对光学神经网络系统的设计及硬件的选取，具有指导意义。

Influence of Hardware Error on Optical Neural Network

The error of optical hardware, which influences the error probability of associative memory in Hopfield neural network, is analyzed by menas of statistical method. The approximate formula and modulation result of error probability versus hardware error are proposed. It is concluded that the error within some extant does not affect the error probability sharply. Especially, the required gray level of detector array can be much less than the maxium value detected. This is very significant for designing the optical neural network.