Natural Language-conditioned reinforcement learning (RL) enables the agents to follow human instructions. Previous approaches generally implemented language-conditioned RL by providing human instructions in natural language (NL) and training a following policy. In this outside-in approach, the policy needs to comprehend the NL and manage the task simultaneously. However, the unbounded NL examples often bring much extra complexity for solving concrete RL tasks, which can distract policy learning from completing the task. To ease the learning burden of the policy, we investigate an inside-out scheme for natural language-conditioned RL by developing a task language (TL) that is task-related and unique. The TL is used in RL to achieve highly efficient and effective policy training. Besides, a translator is trained to translate NL into TL. We implement this scheme as TALAR (TAsk Language with predicAte Representation) that learns multiple predicates to model object relationships as the TL. Experiments indicate that TALAR not only better comprehends NL instructions but also leads to a better instruction-following policy that improves 13.4% success rate and adapts to unseen expressions of NL instruction. The TL can also be an effective task abstraction, naturally compatible with hierarchical RL.

Fictitious play with reinforcement learning is a general and effective framework for zerosum games. However, using the current deep neural network models, the implementation of fictitious play faces crucial challenges. Neural network model training employs gradient descent approaches to update all connection weights, and thus is easy to forget the old opponents after training to beat the new opponents. Existing approaches often maintain a pool of historical policy models to avoid the forgetting. However, learning to beat a pool in stochastic games, i.e., a wide distribution over policy models, is either sample-consuming or insufficient to exploit all models with limited amount of samples. In this paper, we propose a learning process with neural fictitious play to alleviate the above issues. We train a single model as our policy model, which consists of sub-models and a selector. Everytime facing a new opponent, the model is expanded by adding a new sub-model, where only the new sub-model is updated instead of the whole model. At the same time, the selector is also updated to mix up the new sub-model with the previous ones at the state-level, so that the model is maintained as a behavior strategy instead of a wide distribution over policy models. Experiments on Kuhn poker, a grid-world Treasure Hunting game, and Mini-RTS environments show that the proposed approach alleviates the forgetting problem, and consequently improves the learning efficiency and the robustness of neural fictitious play.

In many decision-making tasks, some specific actions are limited in their frequency or total amounts, such as "fire" in the gunfight game and "buy/sell" in the stock trading. We name such actions as "sparse action". Sparse action often plays a crucial role in achieving good performance. However, their Q-values, estimated by classical Bellman update, usually suffer from a large estimation error due to the sparsity of their samples. The greedy policy could be greatly misled by the biased Q-function and takes sparse action aggressively, which leads to a huge sub-optimality. This paper constructs a reference distribution that assigns a low probability to sparse action and proposes a regularized objective with an explicit constraint to the reference distribution. Furthermore, we derive a regularized Bellman operator and a regularized optimal policy that can slow down the propagation of error and guide the agent to take sparse action more carefully. The experiment results demonstrate that our method achieves state-of-the-art performance on typical sparse action tasks.

No abstract