Originally posted by: Heungjo

Based on the manual, the CPLEX network optimizer solve the network flow problem very efficiently.
I would like to know if the CPLEX network optimizer can solve the generalized network flow problem or not.

If anybody knows, please let me know.

Thank you.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

CPLEX can solve multi-commodities network flow problem very fast.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Heungjo

Lessi, Thank you so much for your answer.
I know CPLEX solver can solve multi-commodities network flow problem very fast.
However, I would like to know about that more precisely.

The CPLEX manual says that it would be 100 faster to solve the network flow problem by the CPLEX network optimizer than by other general CPLEX simplex solvers.
This is the exact point that I want to know.
My formulation contains the generalized network flow problem.
I want to solve my problem by decomposing the generalized network flow problem part as subproblem.
I can solve that subproblem by using general simplex solver fast.
But, I would like to use the network optimizer, because, if possible, it would be faster than the general simplex solvers.

Does anybody have ideas about this?

Thank you.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: RWunderling

As usual with such questions, there is no clear answer. Only experimentation will tell how individual algorithms perform on your particular problems. However, the following is true:

1. The network simplex uses internal data structures for representing the simplex basis that take advantage of the network structure of your problem. This is what leads to the dramatic increase in pivoting speed compared to the general simplex algorithm.

2. The network simplex implements a primal simplex algorithm. If you require restarts from a dual feasible but not primal feasible basis, it may require more iterations than the general dual simplex algorithm.

3. If you formulate your problem using the regular general LP API CPLEX provides the network extractor to identify an embedded network. Thus I suggest you conduct your experiments first using the the general API and use the network extractor to generate the network subproblem. This way you can compare the network simplex with the general simplex performance most conveniently for your problems.

Roland
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Heungjo

Thank you Roland,
I appreciate your answer.
Your answer would be very helpful.
As you recommended, I will do some experiments.
#CPLEXOptimizers
#DecisionOptimization