[SCIP] Help With Infeasibility and Constraint Issues in SCIP Implementation

[Marc Pfetsch]

Dear Mahdi,

SCIP does not currently have an infeasibility analyzer, but will likely 
have one in SCIP 10. Issue 2 below seems to rather be a logical error 
than a numerical one. I suggest to output the problem and check whether 
the problem looks as it should be.

Moreover, have you made sure that the constraints you add are 
modifiable? Only in this case, you can perform correct pricing on them. 
Also note that you have to take the dual variables of the new 
constraints into account in the pricing problem.

Best

Marc

On 12/12/2024 16:25, Mahdi Mohammadi wrote:
> Dear SCIP Development Team,
> 
> First, I would like to express my gratitude for providing such a 
> powerful and freely accessible framework to the scientific community. 
> Your work has significantly advanced research in optimization and 
> related fields.
> 
> I am reaching out regarding an issue I have been struggling with while 
> implementing my model in SCIP. My implementation is based on the 
> bin-packing problem example you provided in C language. However, I am 
> encountering difficulties with one of my pricing constraints.
> Specifically, when I use the following linear constraint:
> 
> SCIPcreateConsBasicLinear(subscip, &cons, "sum_cap", nvars, vars, vals, 
> capacity, capacity);
> 
> The problem becomes infeasible. I am unsure how to debug and identify 
> the source of the infeasibility, as I have not found a way to generate 
> detailed diagnostic reports similar to those in GAMS, CLPEX and etc.
> Meanwhile, when I replace the linear constraint with a knapsack 
> inequality as shown below, the instances become feasible but yield 
> incorrect solutions for some cases:
> 
> SCIPcreateConsBasicKnapsack(subscip, &cons, "sum_cap", nvars, vars, 
> vals, capacity);
> 
> Interestingly, when I set the left-hand side of the linear constraint to 
> zero, the problem becomes feasible again but produces another incorrect 
> feasible solution. Since the problem's objective is maximizing this 
> should not happen.
> 
> To summarize, I would greatly appreciate your guidance on the following:
> 
> 1.How can I debug infeasibility in SCIP similar to GAMS to identify the 
> root cause? I want to see the result of summation in the constraint that 
> is infeasible.
> 2.Could this issue be related to numerical precision? For example, I 
> have an instance that is solved correctly with the knapsack constraint 
> but becomes infeasible when replaced with the linear constraint while 
> everything is OK.
> For reference, I have attached the relevant model definitions.
> 
> Thank you for your time and assistance. I look forward to hearing your 
> advice on how to resolve this issue.
> 
> Best regards,
> Mahdi Mohammadi
> ------------------------------------------------
> PS:
> I sent this email some minutes ago, but back then my email hasn't been 
> aproved by moderater and I've just recieved the email. I'm sorry if it 
> is duplicated.
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip