search

EMS-TU-Ilmenau / fastmat

A library to build up lazily evaluated expressions of linear transforms for efficient scientific computing.
https://fastmat.readthedocs.io
Apache License 2.0
24 stars 8 forks source link

Inverse and pseudo-inverse Matrices #62

Closed SebastianSemper closed 4 years ago

SebastianSemper commented 5 years ago

I opened this pull request not in order to merge it, but rather to get the discussion going.

  1. We now need a solid way to test these new factories.
  2. We need to find out, which scipy solvers perform best and if we should change the depending on the class type.
  3. We need to handle A.inverse.inverse.
  4. We need to find out, what A.pinverse.pinverse means.
  5. We need to provide means of preconditioning, maybe on demand, maybe automatically.
  6. We have to work through the classes and provide specialized inverses, when they exist.
SebastianSemper commented 5 years ago

I added some small test cases and around half of them fail. I think we need a generator for matrices, that can be inverted in a stable fashion. They do not have to be used everywhere but for the cases in question we really need those. A simple generator drawing the entries from a normal distribution should be sufficient.

SebastianSemper commented 4 years ago

Has been merged in v0.2. Closing.