Ethan Knight

Joshua Achiam, Ethan Knight, Pieter Abbeel

Deep Q-Learning (DQL), a family of temporal difference algorithms for control, employs three techniques collectively known as the `deadly triad' in reinforcement learning: bootstrapping, off-policy learning, and function approximation. Prior work has demonstrated that together these can lead to divergence in Q-learning algorithms, but the conditions under which divergence occurs are not well-understood. In this note, we give a simple analysis based on a linear approximation to the Q-value updates, which we believe provides insight into divergence under the deadly triad. The central point in our analysis is to consider when the leading order approximation to the deep-Q update is or is not a contraction in the sup norm. Based on this analysis, we develop an algorithm which permits stable deep Q-learning for continuous control without any of the tricks conventionally used (such as target networks, adaptive gradient optimizers, or using multiple Q functions). We demonstrate that our algorithm performs above or near state-of-the-art on standard MuJoCo benchmarks from the OpenAI Gym.

Ethan Knight, Osher Lerner

We present a novel algorithm to train a deep Q-learning agent using natural-gradient techniques. We compare the original deep Q-network (DQN) algorithm to its natural-gradient counterpart, which we refer to as NGDQN, on a collection of classic control domains. Without employing target networks, NGDQN significantly outperforms DQN without target networks, and performs no worse than DQN with target networks, suggesting that NGDQN stabilizes training and can help reduce the need for additional hyperparameter tuning. We also find that NGDQN is less sensitive to hyperparameter optimization relative to DQN. Together these results suggest that natural-gradient techniques can improve value-function optimization in deep reinforcement learning.