2

Does there exist in R an interface to a quadratic programming solver that can handle 2000+ parameters and 20000+ constraints ?

I know that my constraints are not contradictory (a solution exists), but there are too many of them and rounding errors makes limSolve::lsei(type=1) AND quadprog::solve.QP() think the authorised space is empty ("inequality contradictory").

If i get the dimension of the problem down a little, say 200+ parameters for 2000+ constraints, both solve.QP and lsei works perfectly, but not if the problem is too big.

So my question is : is there a quadratic programming solver accessible from R that will handle high-dimensional problems a little better ? Or, more generaly, is there another QP solver accessible from R ?

lrnv
  • 1,038
  • 8
  • 19
  • I would probably look at commercial (free for academics) solvers like Cplex, Gurobi and Mosek. They have R interfaces. Having said this, for large models I usually do the modeling outside R and just export the data from R and solution back into R. – Erwin Kalvelagen Sep 07 '19 at 00:19

0 Answers0