0

I am not sure if I should ask my question here or in maths/cross-validated webpage.

I am looking for efficient optimizer in python for quadratic programming that it is in the form:

minimize sum_{i=0}^n f_i(x) subject to linear constraints. Each f_i is a quadratic function itself (actually the Q matrix is identity) for all i.

My main issue is with efficiency (time, usually space comes with sparse matrices), for scipy it takes ages as I had to define objective function manually as the sum of n such quadratics. For cplex I couldn't manage to write my problem.

Note: The subproblems are not independent - they are constrained by common matrix.

matekrk
  • 129
  • 5
  • I am surprised by the statement "For cplex I couldn't manage to write my problem". Cplex is a very capable QP solver, so I suggest to reconsider. I am not aware of any QP problems that can not be expressed for use with Cplex. For another easy way to express QP problems consider CVXPY (with an appropriate QP solver). – Erwin Kalvelagen Mar 06 '20 at 12:07
  • Thank you for your comment. I am new to Cplex and I would appreciate which solver should I use in this case and I will dive more into it. – matekrk Mar 06 '20 at 12:18
  • Actually CVXPY can talk to Cplex. – Erwin Kalvelagen Mar 06 '20 at 12:50
  • I should add that CVXPY is not suited for all models (it cannot express high-dimensional variables). – Erwin Kalvelagen Mar 06 '20 at 13:08
  • @ErwinKalvelagen can cplex express high-dim variables? – matekrk Mar 06 '20 at 13:17
  • Absolutely. As I said before, I am not aware of any QP problems that can not be expressed for use with Cplex. – Erwin Kalvelagen Mar 06 '20 at 13:30

0 Answers0