Two Timescale Stochastic Approximation with Controlled Markov noise and Off-policy temporal difference learning
This addresses theoretical convergence issues in reinforcement learning algorithms, particularly for off-policy learning scenarios.
The paper tackles the asymptotic convergence analysis of two timescale stochastic approximation with controlled Markov noise, providing the first such analysis for this framework. It applies these results to solve the off-policy convergence problem for temporal difference learning with linear function approximation.
We present for the first time an asymptotic convergence analysis of two time-scale stochastic approximation driven by `controlled' Markov noise. In particular, both the faster and slower recursions have non-additive controlled Markov noise components in addition to martingale difference noise. We analyze the asymptotic behavior of our framework by relating it to limiting differential inclusions in both time-scales that are defined in terms of the ergodic occupation measures associated with the controlled Markov processes. Finally, we present a solution to the off-policy convergence problem for temporal difference learning with linear function approximation, using our results.