摘要
在增量算法的基础上，利用截断（Ｃｌｉｐｐｉｎｇ）方法和蒙塔卡罗（ＭｏｎｔｅＣａｒｌｏ）算法，对以四类飞行目标平面旋转投影作为学习样本的级联神经网络互连权重进行了二值优化处理，并用非学习样本进行了容错性检验。
In this paper, based on the increment algorithm, the clipping learning method and Monto Carlo algorithm were used in optimization of a cascaded neural network. As a result, binary interconnection weights were obtained. The error tolerance of the neural network was tested by non learning sets. Computer simulation indicated that the results were satisfactory.
出处
《光学学报》
EI
CAS
CSCD
北大核心
1996年第10期1497-1500,共4页
Acta Optica Sinica
基金
国家自然科学基金和攀登计划所资助
关键词
模式识别系统
神经网络
互连权重
灰度阶
pattern recognition system, interconnection weight, gray levels, binary.
