aiqbench analysis
This paper introduces a $\lambda$-target update mechanism that geometrically averages periodic hard target updates to enhance stability in linear Q-learning. By bridging the gap between discrete one-period updates and continuous projected Q-value iteration, the study offers a theoretically grounded approach for stabilizing deep Q-learning with function approximation. The analysis, conducted via switching-system models, provides a deterministic framework that extends to stochastic reinforcement learning settings.
Geometrically Averaged Hard Target Updates for Linear Q-Learning
arXiv:2606.10835v1 Announce Type: new Abstract: Periodic hard target updates are among the most common stabilization devices in modern deep Q-learning. Recent studies suggest that target updates can improve stability in Q-learning with function approximation, including linear function approximation. We introduce and analyze the so-called $\lambda$-target update, obtained by averaging the $m$-periodic target update maps with $\lambda$-geometric weights $(1-\lambda)\lambda^{m-1}$, $\lambda \in [0,1]$. The endpoint $\lambda=0$ recovers the one-period target update, while the continuous endpoint $\lambda\uparrow1$ recovers projected Q-value iteration. We study this mechanism for Q-learning with linear function approximation, namely linear Q-learning, using a switching-system model and related tools. For clarity, the paper treats a deterministic version; the formulation extends to stochastic reinforcement-learning settings.
Abstract:Periodic hard target updates are among the most common stabilization devices in modern deep Q-learning. Recent studies suggest that target updates can improve stability in Q-learning with function approximation, including linear function approximation. We introduce and analyze the so-called $\lambda$-target update, obtained by averaging the $m$-periodic target update maps with $\lambda$-geometric weights $(1-\lambda)\lambda^{m-1}$, $\lambda \in [0,1]$. The endpoint $\lambda=0$ recovers the one-period target update, while the continuous endpoint $\lambda\uparrow1$ recovers projected Q-value iteration. We study this mechanism for Q-learning with linear function approximation, namely linear Q-learning, using a switching-system model and related tools. For clarity, the paper treats a deterministic version; the formulation extends to stochastic reinforcement-learning settings.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2606.10835 [cs.LG]
(or arXiv:2606.10835v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.10835 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 205 words