Temporally-Extended ε-Greedy Exploration
This work addresses the exploration challenge in RL for practitioners by offering a simple yet effective method, though it is incremental as it builds on the well-known ε-greedy approach.
The paper tackles the problem of exploration in reinforcement learning by proposing a temporally extended ε-greedy algorithm that reduces dithering by repeating actions for random durations, which improves performance on a broad set of domains, with specific distributions inspired by animal foraging yielding strong results.
Recent work on exploration in reinforcement learning (RL) has led to a series of increasingly complex solutions to the problem. This increase in complexity often comes at the expense of generality. Recent empirical studies suggest that, when applied to a broader set of domains, some sophisticated exploration methods are outperformed by simpler counterparts, such as ε-greedy. In this paper we propose an exploration algorithm that retains the simplicity of ε-greedy while reducing dithering. We build on a simple hypothesis: the main limitation of ε-greedy exploration is its lack of temporal persistence, which limits its ability to escape local optima. We propose a temporally extended form of ε-greedy that simply repeats the sampled action for a random duration. It turns out that, for many duration distributions, this suffices to improve exploration on a large set of domains. Interestingly, a class of distributions inspired by ecological models of animal foraging behaviour yields particularly strong performance.