Newest papers

Continuous-Time Policy Gradient from Discrete Data: Convergence Analysis and Model-Free Algorithms

Martin Hernandez, Xinyan Qing, Mo Zhou, Yuhua Zhu · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28580 · cited by 0 works (OpenAlex) · open-access PDF

German

Kontinuierliches Datenverstärkungslernen sucht eine optimale Politik unter unbekannter Dynamik, angetrieben von stochastischen Differentialgleichungen mit Beobachtungen, die in diskreten Zeiten gesammelt werden. Wir entwickeln ein politisches Gradientenrahmen, das strenge Konvergenzanalysen mit Lernen aus diskreten Daten kombiniert. Wir erstellen eine Leistungsdifferenz lemma und ergeben Politikgradientformeln für zufällige und parametrische Richtlinien. Wir beweisen dann die exponentielle Konvergenz eines entropie-regulierten politischen Flusses zum regulierten Optimum. Für diskrete parametrische Aktualisierungen ergeben wir eine Konvergenzbindung, die explizit die Beiträge von kritischen Fehlern, Politik-regressionen,

Machine translation (Meta M2M100); the English original follows.

Abstract

Continuous-time reinforcement learning seeks an optimal policy under unknown dynamics driven by stochastic differential equations using observations collected at discrete times. We develop a policy gradient framework that combines rigorous convergence analysis with learning from discrete data. We first establish a performance difference lemma and derive policy gradient formulas for randomized and parametrized policies. We then prove exponential convergence of an entropy-regularized policy flow to the regularized optimum. For discrete parametrized updates, we derive a convergence bound that explicitly separates the contributions of critic error, policy-regression error, and the learning step size. To enable learning under unknown dynamics, we integrate this framework with PhiBE, which constructs a continuous-time surrogate generator from discrete transition data. The resulting on-policy and off-policy model-free actor-critic algorithms preserve the continuous-time structure of the control problem. Numerical experiments on an entropy-regularized linear-quadratic regulator and a nonlinear wheeled robot demonstrate the performance and stability of the proposed methods with both known and data-driven dynamics.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Convergence (economics), Reinforcement learning, Mathematics, Lemma (botany), Stability (learning theory), Discrete time and continuous time

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange