Breaking the Deadly Triad with a Target Network
This addresses a foundational stability problem in reinforcement learning for researchers and practitioners, offering a theoretical and practical solution to enable more reliable algorithms.
The paper tackles the instability in reinforcement learning caused by the deadly triad (off-policy learning, function approximation, and bootstrapping) by proposing a novel target network update rule with projections, resulting in the first convergent linear Q-learning algorithms under nonrestrictive and changing behavior policies without bi-level optimization.
The deadly triad refers to the instability of a reinforcement learning algorithm when it employs off-policy learning, function approximation, and bootstrapping simultaneously. In this paper, we investigate the target network as a tool for breaking the deadly triad, providing theoretical support for the conventional wisdom that a target network stabilizes training. We first propose and analyze a novel target network update rule which augments the commonly used Polyak-averaging style update with two projections. We then apply the target network and ridge regularization in several divergent algorithms and show their convergence to regularized TD fixed points. Those algorithms are off-policy with linear function approximation and bootstrapping, spanning both policy evaluation and control, as well as both discounted and average-reward settings. In particular, we provide the first convergent linear $Q$-learning algorithms under nonrestrictive and changing behavior policies without bi-level optimization.