具有广义行或列对称性结构的矩阵(即广义延拓矩阵)的奇异值和奇异向量与原矩阵(母矩阵)的奇异值和奇异向量存在定量关系,从而对于广义延拓矩阵的奇异值分解,可利用母矩阵奇异值分解来实现。这不但可以节省计算量和运算量,而且不影响任何数据的精度。
The quantitative correspondence of the singular values and singular vectors between the generalized row or column symmetric matrix ( namely, generalized extended matrix) and its original (a. k. a mother matrix) were derived. The singular value decomposition of generalized extended matrix could save dramatically the memory by using the mother matrix and alleviate considerably the computation burden without loss of any numerical precision.