Comment by whatever1

Sorry if I am harsh, but a 1200 node tsp problem is a toy problem. We can find proven optimal solutions to these in a fraction of the time you spent.

RL is probably best suited for uncertainty infected instances.