Originally posted by: UserCplex

Hello,

I am running a series of slightly different shortest path problems on the same underlying network of nodes. Each time, the cost of the arcs of the networks will change.

I intend to use the CPLEX Network Simplex interface to accomplish this. It is attractive that if the problem is changed, the optimal solution from a previous iteration provides a starting basis.

I was wondering if CPLEX Network Simplex developers are aware of some benchmark comparisons between their algorithm and other libraries -- in particular, the Boost Graph Library -- BGL -- that help solve the same purposes. https://www.boost.org/doc/libs/1_66_0/libs/graph/doc/index.html

I don't know how BGL is internally interpreted, perhaps it implements its own network simplex algorithm. But I doubt that BGL would be able to exploit the final optimal solution of one problem to solve another slightly different problem.

If there are some benchmark results comparing CPLEX Network Simplex with BGL that CPLEX folks have undertaken, please do let me know.

Thanks.