DinerDash Gym: A Benchmark for Policy Learning in High-Dimensional Action Space
This provides a domain-specific benchmark for researchers in reinforcement learning to evaluate algorithms in complex, high-dimensional action spaces, though it is incremental as it builds on existing policy learning methods.
The authors tackled the lack of a benchmark for policy learning in hierarchical tasks with high-dimensional action spaces by proposing DinerDash Gym, a new benchmark with a 57-dimensional action space, and introduced the Decomposed Policy Graph Modelling (DPGM) algorithm, which achieved significant improvements over baselines.
It has been arduous to assess the progress of a policy learning algorithm in the domain of hierarchical task with high dimensional action space due to the lack of a commonly accepted benchmark. In this work, we propose a new light-weight benchmark task called Diner Dash for evaluating the performance in a complicated task with high dimensional action space. In contrast to the traditional Atari games that only have a flat structure of goals and very few actions, the proposed benchmark task has a hierarchical task structure and size of 57 for the action space and hence can facilitate the development of policy learning in complicated tasks. On top of that, we introduce Decomposed Policy Graph Modelling (DPGM), an algorithm that combines both graph modelling and deep learning to allow explicit domain knowledge embedding and achieves significant improvement comparing to the baseline. In the experiments, we have shown the effectiveness of the domain knowledge injection via a specially designed imitation algorithm as well as results of other popular algorithms.