Warning of max constraint violation exceeding tolerance in MIQCP
回答済みI am using Gurobi version 10.0.1 to solve problems formulated as mixed-integer programs with linear objective function and all linear constraints except for a set of constraints which are quadratic convex.
The quadratic constraints are perspective constraints of the form:
v[i]^2 <= u[i]y[i] for i={1,...,100}
where v,u are non-negative continuous variables, whereas y is a binary variable. They can also be expressed as rotated second-order cones.
I am trying to solve instances of this problem, but on every instance there is a warning in the log file that says (I am copying and pasting one of these warnings) "Warning: max constraint violation (4.8070e-01) exceeds tolerance". I tried setting several parameters to avoid this, without success. I don't know what to do because I can't rely on the optimal solution found if it says it is violating some constraints.
The parameters I tested are: MIPGap = 0.0, FeasibilityTol = 1e-9, IntFeasTol = 1e-9, OptimalityTol = 1e-9, Presolve = 0, MIPGapAbs = 0, PSDTol = 0, BarQCPConvTol = 0, BarConvTol = 0. I also tested PreMIQCPForm = 2 together with BarHomogeneous = 1.
I am solving the problem without any callback, simply using the function model.optimize(). I am not setting other parameters than the ones I listed above. The problem admits solution (e.g., with y=1), so it cannot be infeasible, nor unbounded. I can send some instances/logs if it can help understanding what is going on (in case, please let me know how). I'd be grateful if you could provide some guidance.
-
正式なコメント
This post is more than two years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support. -
You can try different PreMIQCPForm = 0, or 1 or PreQLinearize = 0, 1, 2.
Also if you have any warnings in the log you may want to try the NumericFocus = 1, 2, or 3 or Quad = 1.
You may also want to reduce the bounds for the variables if possible.
Cheers,
David0
投稿コメントは受け付けていません。
コメント
2件のコメント