Hi,
I have a ill-conditioned sparse matrix with 616*616 elements and I want to solve it by Eigen.
I use SparseQR solver to handle it, but to my surprise, the solution is different from matlab's result which
obtained by SVD method.
Do we have SVD Sparse solver to solve my matrix?

PS:I have noticed that there is redsvd code based on Eigen, but I don't know how robust it is .
Could anyone has experience with resvd give me some advice?

Thanks.

Note that if the matrix is not full rank, then there is not a unique solution. First thing to do is thus to check for the relative error: (A*x-b).norm()/b.norm().

Then, if you are looking for the minimal norm solution, then better use the 3.3-beta1 release which features a fast dense SVD solver: BDCSVD<>. For such a small matrix, it will be pretty fast (probably around 1s).