Using this environment for greedy coloring

[sundar19]

Hello,
I am studying this project and need to know if this environment can be used for something similar to greedy_color() in networkx. My intention is to solve allocating color/Number to every node such that no two nodes next to each other have the same color. Instead of doing this with greedy_color() I want to use reinforcement learning methods to solve this problem. Any suggestions would be appreciated.

[jlaw9]

Hi @sundar19 . Sure you do that. You just need to be able to formalize the problem into an action space (e.g., select a node and change its color) with a reward (e.g., -1 * number of colors). We have a couple example problems you should be able to use as a starting point here: https://nrel.github.io/graph-env. The TSP example specifically should help.

If you need ideas for how approach the graph coloring problem with RL, looks like others have done work on this problem: https://github.com/barisbatuhan/Graph_Coloring_with_RL.

If you have more specific questions or issues, we'd be happy to help.

[sundar19]

@jlaw9 Thank you so much for the support