论文部分内容阅读
提出一种多目标旋转不变分类识别的新方法。其基本思想是多对一异联想存储模型，为便于光学实现、充分利用系统的时间、空间带宽积，采用了全单极（输入模式和权重均为单极）形成，并对权重进行了二值化截取。一个分布阈值用来克服单极联想存储模型０，１状态分布几率不等时存储容量急剧下降的缺点，同时通过巧妙地选取目标所对应的异联想码将异联想模型和ＷＴＡ模型有机地结合起来，提高了整个系统的存储容量和容错性。本文给出了计算机仿真模拟及一个光电混合系统的实验结果。
A new method of multi-objective rotation invariant classification and classification is proposed. The basic idea is a multi-to-one allxynchrony storage model, in order to facilitate the optical realization, make full use of the system time and space bandwidth product, the use of the full monopole (input mode and weight are unipolar) formation, and the weight of two Value cut. A distribution threshold is used to overcome the shortcomings of the storage capacity dropping sharply when the unipolar associative memory model has different probability distributions in the 0,1 state. At the same time, the heterogeneous association model and the WTA model are organically integrated through the ingenious selection of the target association codes , Improving the storage capacity and fault tolerance of the entire system. This paper presents the computer simulation and experimental results of a hybrid optical system.