约束最小平方联想存储器及性能比较

CONSTRAINED LEAST SQUARES ASSOCIATIVE MEMORY AND ITS PERFORMANCE COMPARISON

鉴于Ｋｏｈｏｎｅｎ的最佳联想存储器对带噪输入会产生难以接受的联想误差，文中试图通过在Ｋｏｈｏｎｅｎ模型中引入对连接权阵的某种约束并进而优化，使修改后的Ｋｏｈｏｎｅｎ模型（ＣＬＳＡＭ）对带噪输入具有最小误差的联想．借助奇异值分解（ＳＶＤ）理论的分析和计算机模拟证实了ＣＬＳＡＭ的性能优越性．

Considering the sensitivity of Kohonen optimal associative memory to inputs with noise,a constrained least squares associative memory (CLSAM) is presented by introducing the constraints to the connection weights of Kohonen associative memory and further optimizing the memory performance of the CLSAM so that the CLSAM has minimum associative error or noise.With the help of the singular value decomposition (SVD) approach and computer simulations,the better performance of the CLSAM is confirmed compared with Kohonen associative model.