Beyond Surprise: Improving Exploration Through Surprise Novelty
This addresses the challenge of exploration for reinforcement learning agents in sparse reward settings, offering an incremental improvement over existing surprise-driven methods.
The paper tackles the problem of inefficient exploration in reinforcement learning by proposing a new intrinsic reward based on the novelty of surprise rather than its magnitude, using a memory network to estimate this novelty. The result shows that this method significantly boosts final performance in sparse reward environments, including Noisy-TV, navigation, and Atari games.
We present a new computing model for intrinsic rewards in reinforcement learning that addresses the limitations of existing surprise-driven explorations. The reward is the novelty of the surprise rather than the surprise norm. We estimate the surprise novelty as retrieval errors of a memory network wherein the memory stores and reconstructs surprises. Our surprise memory (SM) augments the capability of surprise-based intrinsic motivators, maintaining the agent's interest in exciting exploration while reducing unwanted attraction to unpredictable or noisy observations. Our experiments demonstrate that the SM combined with various surprise predictors exhibits efficient exploring behaviors and significantly boosts the final performance in sparse reward environments, including Noisy-TV, navigation and challenging Atari games.