摘要
多维联想记忆神经网络可以回忆图像。在算法相同且库向量维数也相同的情况下,选取图像矩阵中两列跨度为图像矩阵列秩一半的向量作为库向量,回忆得到的图像最清晰、回忆时间最短,此库向量是最佳的。理论分析和计算机仿真都表明该结论是正确的。
Multidimensional associative memory neural networks can be used for image recalling.If library vector is made up of two column vectors whose span is half of image matrix column rank taken out from image matrix,the restored image is clearest and the recalling process takes least time in the same algorithm and the same dimensions of library vector.So the library vector is best.The theory and computer simulation both justify the conclusion.
出处
《滁州学院学报》
2010年第2期29-31,共3页
Journal of Chuzhou University
关键词
多维联想记忆神经网络
库向量
投影
multidimensional associative memory neural networks
library vector
projection
