Solving TSPproblem with SB algorithm

[Alexqzx]

Hello arthors，your work is really amazing.Now,i am trying to use your code to solve Tsp problem, i download the dataset on Tsplib such as burma14 and so on. I implement your code and give your code J and H matrix in the ising problem which i computed by dataset.I fount its hard to get the best known answer, since add the load on constraints the best values will grow otherwise may not find the valid solution. I want to ask have you tried SB algorithm on TSP problems? I want to know whether i implemented wrongly or didn't set the correct parameters or SB algorithm does perform very well on TSP problems. I am looking forward to your precious answer!

[bqth29]

Hello @Alexqzx and thanks for your kind feeback!

We are currently working on implementing more properly the models module to allow using classical NP problems such as TSP. Unfortunately, constrained problems like TSP are not always trivial to solve using SB because the constraints must be written as quadratic penalties, and the value of the chosen penalty can influence the performance drastically.

We have not yet benchmarked TSP problems but from our first observations on other NP problems, running SB in discrete mode seems to achieve better results. For the moment, everything we can recommend is to run dSB with different parameters (possibly changing the values of the SB environment) to see which set works best.

We'll keep on working on TSP on our side too and we can use this issue to share our respective results and observations.

Do not hesitate to reach us again if you have more questions. All the best!