Originally posted by: SystemAdmin

[sjtudai said:]

Here, I have a question about the Linear Programming Solver in Cplex.

For solving a Linear Programming iteratively, its constraints are the same at each iteration, but the objective function at iteration k is different from that at iteration k+1, so I want to know:
Is there an option for LP solver for "solve the current LP based on (starting from) the solution of the LP solved in last time"?
How to set it respectively by C (ILOG Callable Library) or C++ (ILOG Concert Technology) ?
I think Cplex may have such an option for its LP solver. Because In this way, the computation time for solving LP can be greatly reduced.

Thank you very much in advance.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[prubin said:]

ILOG calls this an "advanced start" (the colloquial term is "hot start").  The parameter you are looking for is the advanced start switch (AdvInd in the programming APIs), and it is turned on by default, so you probably need not do anything special in your code.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[notdarkyet said:]

Just a remark: the advanced start indicator is on by default in the current versions of cplex, it used to default to off in the older versions.
So it can't do any harm to set it to on explicitly.
notdarkyet
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

[achterberg said:]

If you change the objective function you would usually use the primal simplex in order to resolve a modified problem. This is because changing the objective function does not destroy primal feasibility and hence, the primal simplex can immediately start in phase 2 of the algorithm.

In contrast, if you add constraints or tighten bounds it is usually much better to use the dual simplex because these operations do not destroy dual feasibility.

Warm starts with the barrier are not possible.

#CPLEXOptimizers
#DecisionOptimization