Mathematics > Dynamical Systems
[Submitted on 31 Mar 2015 (v1), revised 18 Jan 2016 (this version, v8), latest version 25 Feb 2017 (v14)]
Title:Two Timescale Stochastic Approximation with Controlled Markov noise and Off-policy temporal difference learning
View PDFAbstract:We present for the first time an asymptotic convergence analysis of two-timescale stochastic approximation driven by controlled Markov noise. In particular, both the faster and slower recursions have non-additive 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 invariant probability measures associated with the controlled Markov processes. Finally, we show how to solve the off-policy convergence problem for temporal difference learning with linear function approximation using our results and proving stability of the iterates in this case. Moreover, in general, we emphasize the fact that all the reinforcement learning scenarios where function approximation of value function is deployed needs to consider Markov noise in the convergence proofs.
Submission history
From: Prasenjit Karmakar [view email][v1] Tue, 31 Mar 2015 16:10:55 UTC (32 KB)
[v2] Thu, 2 Apr 2015 17:18:37 UTC (32 KB)
[v3] Thu, 30 Apr 2015 04:11:39 UTC (32 KB)
[v4] Tue, 4 Aug 2015 12:49:32 UTC (33 KB)
[v5] Wed, 5 Aug 2015 14:02:19 UTC (33 KB)
[v6] Thu, 6 Aug 2015 12:53:51 UTC (33 KB)
[v7] Fri, 1 Jan 2016 12:10:22 UTC (33 KB)
[v8] Mon, 18 Jan 2016 15:29:21 UTC (33 KB)
[v9] Mon, 21 Mar 2016 19:25:28 UTC (30 KB)
[v10] Sat, 26 Mar 2016 04:53:48 UTC (30 KB)
[v11] Sun, 17 Apr 2016 13:11:17 UTC (33 KB)
[v12] Thu, 16 Feb 2017 09:37:38 UTC (33 KB)
[v13] Wed, 22 Feb 2017 17:06:39 UTC (33 KB)
[v14] Sat, 25 Feb 2017 18:46:13 UTC (35 KB)
Current browse context:
math.DS
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
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.