I've a matrix of dimension 5000x12. I have reduced the dimension of the matrix using SVD. Now I want to calculate the compression ratio. But the problem is that the orthogonal matrices, U and V have the dimension of 5000x12 and 12x12 respectively and S has the dimension of 12x12 i.e. only 12 diagonal elements. As the size of U is same as that of the data matrix, there won't be any compression. Hence, can anyone suggest me how to get a good CR. Is there possible to calculate U or approximate of U matrix using only S and V matrices.

Question

anonymous · Answer

hello sibashankar - i am from iit kanpur what is going on in ur maths 
lecures in iit guwahati ????