0

I want to solve a linear system of equations Ax = b subject to x_i >= 0.

This is of course a case of linear programming. The problem is that A is too large to be created. I can, however, calculate Ax efficiently.

So I have access to f(x) = Ax, x, and b.

I prefer using pytorch (to utilize the GPU). What is the best way to tackle this?

khelwood
  • 55,782
  • 14
  • 81
  • 108
Yakir
  • 1
  • 1
  • Does this answer your question? [Python: multivariate non-linear solver with constraints](https://stackoverflow.com/questions/43995862/python-multivariate-non-linear-solver-with-constraints) – Pranav Hosangadi Sep 26 '22 at 13:30
  • 1
    All serious LP solvers assume A is sparse. – Erwin Kalvelagen Sep 26 '22 at 15:50
  • If it's sparse and only dimensions are big, follow Erwins comment. If not, there is the are of *matrix-free* or (sometimes just) *first-order* (LP) methods. But available software is limited. – sascha Sep 27 '22 at 18:52

0 Answers0