Open domokane opened 3 years ago
I'd be happy to have a go at this
I am not sure this is still needed. Look in utils at solver_cg.py and solver_nm.py to see the functions that were added and where they are used to see if everything is as fast as it can be. I had lots of issues with numba but I think they now work. But I have sort of forgotten so I am not 100% sure.
I saw solver_cg but it didn't look like it was being used anywhere... and it looks like equity_vol_surface (and others) are using the scipy solver. Though it looks like it tries solver_nm and only uses scipy's CG when that fails to converge (or when the user specifies to use it).
Yes. My main focus was on the FX surface solver. If you see a need to work on it let me know. I am jammed on other things right now but if there is stuff to be done on this it would be very helpful.
The conjugate gradient descent algorithm in scipy is slow and prevents the calling function from being jitted by numba. Idea is to find a python version of the CG minimiser that can be jitted to give faster minimisations.