Prevent dual ray calculation when using `HiGHS`

Specific Domains Optimization (Mathematical)
1

I’m seeing the following log with an older model that I’ve just updated:

Running HiGHS 1.10.0 (git hash: fd8665394e): Copyright (c) 2025 HiGHS under MIT licence terms
LP   has 51118 rows; 38919 cols; 128913 nonzeros
Coefficient ranges:
  Matrix [2e-06, 4e+03]
  Cost   [7e-02, 7e+02]
  Bound  [0e+00, 0e+00]
  RHS    [3e-03, 5e+07]
Presolving model
3144 rows, 21250 cols, 29864 nonzeros  0s
Problem status detected on presolve: Infeasible
Model status        : Infeasible
Objective value     :  0.0000000000e+00
HiGHS run time      :          0.04
Solving LP to try to compute dual ray
LP   has 51118 rows; 38919 cols; 128913 nonzeros
Coefficient ranges:
  Matrix [2e-06, 4e+03]
  Cost   [0e+00, 0e+00]
  Bound  [0e+00, 0e+00]
  RHS    [3e-03, 5e+07]
Solving LP without presolve, or with basis, or unconstrained
Using EKK dual simplex solver - serial
  Iteration        Objective     Infeasibilities num(sum)
          0     0.0000000000e+00 Pr: 2921(2.32528e+08); Du: 0(3.18915e-09) 0s
       8921     9.6977738654e+00 Pr: 2677(1.35329e+08); Du: 0(4.50078e-05) 5s
      13852     1.3780839869e+01 Pr: 2866(2.76119e+08); Du: 0(2.02671e-05) 10s
      17865     2.3041731613e+01 Pr: 3004(2.32925e+08); Du: 0(1.5276e-05) 15s
      21885     3.9687283873e+01 Pr: 2535(7.27035e+07); Du: 0(1.71219e-05) 21s
      25139     2.5016489653e+02 23s
Model status        : Infeasible
Simplex   iterations: 25139
Objective value     :  0.0000000000e+00
Relative P-D gap    :  2.5016489664e+02
HiGHS run time      :         23.42
Solving linear system to compute dual ray

HiGHS indicates infeasibility during presolve - how can I turn off the dual ray computation that follows? I didn’t find any (/new) setting in the list of options.

2

There is no option.

HiGHS.jl always asks for the dual ray if the problem is infeasible:

Why do you want to opt-out of computing the dual ray?

3

Simply because of the the timing - comparing

Problem status detected on presolve: Infeasible
Model status        : Infeasible
Objective value     :  0.0000000000e+00
HiGHS run time      :          0.04

versus

Model status        : Infeasible
Simplex   iterations: 25139
Objective value     :  0.0000000000e+00
Relative P-D gap    :  2.5016489664e+02
HiGHS run time      :         23.42
Solving linear system to compute dual ray

This “problem” occurred during a MGA use-case, where spending >100x the time necessary to detect that a given model is infeasible “just” to also get a dual ray isn’t worth it.

4

I’ve opened an issue for a feature request in HiGHS.jl: Add option to turn off certificates · Issue #273 · jump-dev/HiGHS.jl · GitHub

1 Like
5

@sstroemer thoughts on Add HiGHS.ComputeInfeasibilityCertificates attribute by odow · Pull Request #274 · jump-dev/HiGHS.jl · GitHub?