Eigendecomposition for asymmetric matrix

i want to compute the eigenvalues and eigenvectors of a hidden layer using mxnet, 
i have seen mxnet.ndarray.linalg.syevd ,but it only support the symmetric matrix, if the matrix is asymmetric how to solve?

i resolved this problem by using “.asnumpy()”, and then using " .eig(T)" support by numpy ,but i found it run really slow.Does anyone has any good idea to solve it?
thank u very much

@asmushetzel any ideas?

@Eva_Hua are there any other properties of your matrix beyond asymmetry?

@mseeger What are your thoughts?

Hello, I wrote syevd.
Are you asking for the singular value decomposition (SVD), which can be computed for any matrix?
If so, I should say that I have derived the maths for the SVD, but did not get around to implement it. I myself do not have a use case for it. I could be convinced spending some work, if somebody told me about a good use case.

Or are you asking about the eigendecomposition of a non-symmetric (but square) matrix? Note that in this case, both eignvalues and eigenvectors are complex-valued, a datatype which MXNet does not support. So no, it does not make much sense to implement such an operator. The SVD does.