Will all contingencies in Trial Event 1 Dataset be feasible?

2 posts / 0 new
Last post
Dr. Olga Anna Kuchar
Will all contingencies in Trial Event 1 Dataset be feasible?

From the webinars, it was noted that GO Competition organizers preserve the right to include infeasible contingencies in data sets. Is this true?

Total votes: 631
Dr. Olga Anna Kuchar
From Jesse Holzer

Generally, no, not all contingencies may be feasible.

The problem has hard constraints and soft constraints. Soft constraints are those which can be violated at a cost that appears in the objective function. A soft constraint violation does not render a solution infeasible. Hard constraints cannot be violated at all. A hard constraint violation renders a solution infeasible.

A solution is called hard-feasible if it does not violate any hard constraints. A solution is called soft-feasible if it does not violate any constraints at all. A hard-feasible solution will be regarded as feasible by our evaluation code. A soft-feasible solution will be feasible and will have relatively small constraint violation penalties in its objective.

We do guarantee that every scenario will have a hard-feasible solution. We describe a method of producing such a solution in the formulation document.

We do not guarantee that every scenario will have a soft-feasible solution. Nor do we guarantee that there is any scenario not having a soft-feasible solution.

For each scenario, our validation procedure attempts to find a hard-feasible solution with very small soft constraint violations, and we have used our judgment in discarding or revising scenarios where we are unable to find such a solution. We do not make any particular quantitative statement about the magnitude or number of soft constraint violations that we might accept in validating a scenario.