IFlowNets: Extending Generative Samplers to Learn Strategies in Incomplete Information Games

While many algorithms blend reinforcement learning (RL) with counterfactual regret (CFR) methods to leverage tradeoffs in computational speed and performance, there are fewer investiga- tions into generative sampling frameworks in game theoeretic applications in incomplete informa- tion games. We extend a generative flow network framework, Adversarial Flow Networks (AFNs), to incomplete information games, called Information Flow Networks (IFNs). We prove that pre- viously established constraints for generative flow networks in complete information games are inadmissible for obtaining valid densities (corresponding to player strategies) and a valid training objective. We show that our proposed generalization, IFlowNets, alleviates this issue and strictly generalizes AFlowNets. In preliminary results for three standard game environments, IFlowNets perform comparably to or better than OS-MCCFR and standard RL-based methods in performance and speed.