LGAIDec 31, 2021

A Theoretical Understanding of Gradient Bias in Meta-Reinforcement Learning

arXiv:2112.15400v415 citations
Originality Incremental advance
AI Analysis

This work addresses gradient bias issues in meta-reinforcement learning, which is crucial for improving adaptation efficiency in AI agents, though it is incremental as it builds on existing frameworks.

The paper identifies that existing stochastic meta-gradient estimators in gradient-based meta-reinforcement learning are biased due to compositional and multi-step Hessian estimation sources, with theoretical bounds provided, and demonstrates how methods like off-policy learning can mitigate this bias in experiments on tabular MDPs, Iterated Prisoner's Dilemma, and Atari games.

Gradient-based Meta-RL (GMRL) refers to methods that maintain two-level optimisation procedures wherein the outer-loop meta-learner guides the inner-loop gradient-based reinforcement learner to achieve fast adaptations. In this paper, we develop a unified framework that describes variations of GMRL algorithms and points out that existing stochastic meta-gradient estimators adopted by GMRL are actually \textbf{biased}. Such meta-gradient bias comes from two sources: 1) the compositional bias incurred by the two-level problem structure, which has an upper bound of $\mathcal{O}\big(Kα^{K}\hatσ_{\text{In}}|τ|^{-0.5}\big)$ \emph{w.r.t.} inner-loop update step $K$, learning rate $α$, estimate variance $\hatσ^{2}_{\text{In}}$ and sample size $|τ|$, and 2) the multi-step Hessian estimation bias $\hatΔ_{H}$ due to the use of autodiff, which has a polynomial impact $\mathcal{O}\big((K-1)(\hatΔ_{H})^{K-1}\big)$ on the meta-gradient bias. We study tabular MDPs empirically and offer quantitative evidence that testifies our theoretical findings on existing stochastic meta-gradient estimators. Furthermore, we conduct experiments on Iterated Prisoner's Dilemma and Atari games to show how other methods such as off-policy learning and low-bias estimator can help fix the gradient bias for GMRL algorithms in general.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes