Yixuan Qiu

@bodhi1991 I think you can first test the CG solver in one iteration, and see if is faster than an LDLT solver. And then you may compare the number of...

@GYengera It's hard to say "optimum", but generally `SymEigsShiftSolver` is a good starting point. You can always benchmark different software packages to pick the best for you.

@y-yao I may change that to `Eigen::Index`, which is expanded to `std::ptrdiff_t` by default. Will let you know when ready. Thanks.

I don't think there is a standard algorithm to do this, at least from the user guide of ARPACK. But actually there is a simple method to achieve this goal...

It would be very helpful if you can provide a concrete example matrix, otherwise there is little I can start with.

I tried both Spectra and ARPACK on this matrix and both failed to converge, so I guess this eigenvalue problem is hard in itself. I'll do more research later.

Thank you! If you mean the eigen decomposition of a complex matrix, then I would say that the code needs to be completely rewritten, since for now it only applies...

Sorry for being late, but it seems the link to the matrix file is now invalid. Could you share the file again?

I haven't tested the example in details, but theoretically the shift cannot be any of the true eigenvalues. If you know all the eigenvalues are nonnegative and you want to...

I see. I will improve the documentation in the future. Thanks for the feedback!