In many Mixed-Integer Programming (MIP) applications, not all instances solve equally well.
While some cases reach optimality quickly, harder instances may hit time or resource limits long before the solver has finished its search by reducing the gap below accepted tolerances.
With a standard AIMMS solve statement, restarting the solve means starting over: the model is regenerated and the solver search is reset.
Continued Solve in a nutshell
The GMP library separates model generation from solver execution.
Once a GMP instance has been created, it remains available in memory and can be solved multiple times.
For solvers that support continued search (such as CPLEX and GUROBI), this makes it possible to:
-
Preserve the solver’s internal search state
-
Reuse progress made in earlier solve attempts
-
Continue a search instead of restarting it
This capability is one example of the power of the GMP library, others include retrieving progress information, changing coefficients, adding rows, columns, transposing, and so on.
For a quick presentation on Youtube: https://www.youtube.com/watch?v=n03ccm7PT5U
For a detailed presentation see how-to: https://how-to.aimms.com/Articles/685/685-continued-solve.html