Singular Value Decomposition?

Hi!

I'm really stuck with a nxn matrix (principially a linear equation system represented by a 400x400 matrix). To solve the problem i have to invert the matrix, but it gives "Matrix is singular to working precision." error.

That is there are linear dependencies, rank(Mat) gives me 393 - so I have 7 linear depenencies.

Question is, how can I PIN OUT these linear dependencies. Is it possible with svd or svds command, I get some arrays out of results of these functions but how can I find row/column indices of my arrays linearly dependent rows/columns??

I have made basic iterative script for comparing all rows and columins by pair with coeficients from -10:10 (all possible combiations) but the problem seems to be more sophisticated

Any help would be appreciated!