基于遗传算法的神经认知机选择度调整方法

A genetic algorithm based method for adjusting selectivity of the neocognitron

基于视觉模型而建立的神经认知机能正确识别具有变形、位移和缩放的输入模式．研究表明，选择度参数直接影响着神经认知机的识别能力．设计了一个目标函数，通过对该函数的优化能够得到最佳的选择度．这是一种简单而有效的方法．经该方法调整后，可使各特征选择平面对不同训练样本的响应达到均匀一致，从而提高整个系统的识别能力．对于0-9十个手写阿拉伯数字的仿真结果表明，该方法可有效改善神经认知机的性能．

The neocognitron, which is proposed based on the model of biological vision, has been acclaimed as a shift and distortion tolerant character recognition system. Unfortunately, studies show that the performance of the neocognitron is affected greatly by the value of its selectivity. The neocognitron has a poor recognition rate if the value of selective is not reasonable. A genetic algorithm based method for adjusting necognitron＇ s selectivity is proposed in this paper. By using the proposed method, the responses of S - plane are uniform. The proposed method is tested on 10 digits, and the simulation results show that it is capable of improving the recognition rate of the neocognitron.