I have a program to estimate a vector X from noisy measurements.I use the singular value decomposition SVD to solve the linear equation AX=0. where the solution will be the last vector in the matrix V (assuming [USV] = SVD (A)). the problem is that when I test the program using true values of the measured quantities (quantities without noise) the estimated vector is correct, however when I add the noise the estimated vector will have some flipped signs which make the difference large between true and estimated vectors!
I will be grateful if you could help me.
