Units: | - |
Default Value: | 0.0001 |
Validation Rule: | ≥0 |
Description: | Declare the integer solution optimal when this gap is reached between the current integer solution and best-bound linear relaxation (this is not a measure of optimality). |
Performance MIP Relative Gap is a
solver setting for the mixed integer optimizer. The parameter may take
any value in the range 0.0 to 1.0.
When the value:
|bestbound-bestinteger|/(1e-10+|bestinteger|)
falls below the value of this parameter, the mixed integer
optimization is stopped.
Here:
- 'bestinteger' refers to the objective function value associated
with the best integer feasible solution found so far; and
- 'bestbound' refers to the objective function value of any linear
relaxation to the integer problem that has yet to be 'ruled out' as
potentially leading to a better integer solution
For example, to instruct the solver to stop as soon as it has found a
feasible integer solution
proved to be within half a percent
of optimal, set the relative MIP gap tolerance to 0.5%. Note that this does not guarantee that the current integer
solution is half a percent worse than optimal, it only says that the
true optimal solution is no more than half a percent better than the
current solution. Put another way, the solution at a gap of 0.5% might
in fact be optimal, but the solver has yet to prove it so.
For unit commitment/economic dispatch problems a relative gap of
between 0.1-1% can provide a good balance of quality and performance.
See also: