Post #60

Reinforcement Learning for Combinatorial Optimization: A Survey Our new submission to IJCAI survey track. We surveyed all of the literature we found on applying RL methods for combinatorial optimization problems (e.g. TSP, Knapsack, MaxCut). There are three types of the RL approaches we categorized the papers: Value-based, Policy-based, and Monte-Carlo Tree Search based. This is one of the domains that appeared very recently, a few years ago, and has an increasing number of successful applications to traditional problems each year. I would say it's a good topic for a fresh Ph.D. student to start working on.