0

Working with python, i have a quadratic program (minimize : x' M x - 2 x' b) with constraints (E x = f, G x >= h) where M,E,G are matrices, and b,f,h are vectors.

This is the kind of classical quadratic program that software as OSQP handle well.

What now if the loss function is non linear anymore ? I now that for constraints specifications $l <= x <= u$, SLSQP in scipy does the job, but for matrix/vector constraint i did not found the right optimizer.

Could you provide one ?

lrnv
  • 1,038
  • 8
  • 19
  • `scipy.minimize` takes arbitrary constraints as far as I can tell. Why isn't that suited for your specific problem? Maybe provide a minimal code example of the problem where you know the solution? – user2653663 Jan 27 '20 at 16:02
  • Yeah i did not saw that. I did not think it was possible because `scipy.opitmize.slsqp`, which seems to be the direct call to `slsqp`, does NOT take arbitrary constraints. I thought that `scipy.minimize` was just a wrapper around other algorithms. Thanks – lrnv Jan 27 '20 at 16:30

0 Answers0