We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: O$^2$TD: (Near)-Optimal Off-Policy TD Learning

Abstract: Temporal difference learning and Residual Gradient methods are the most widely used temporal difference based learning algorithms; however, it has been shown that none of their objective functions is optimal w.r.t approximating the true value function $V$. Two novel algorithms are proposed to approximate the true value function $V$. This paper makes the following contributions: (1) A batch algorithm that can help find the approximate optimal off-policy prediction of the true value function $V$. (2) A linear computational cost (per step) near-optimal algorithm that can learn from a collection of off-policy samples. (3) A new perspective of the emphatic temporal difference learning which bridges the gap between off-policy optimality and off-policy stability.
Comments: 10 pages, 7 figures
Subjects: Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:1704.05147 [cs.LG]
  (or arXiv:1704.05147v2 [cs.LG] for this version)

Submission history

From: Daoming Lyu [view email]
[v1] Mon, 17 Apr 2017 23:18:48 GMT (1747kb,D)
[v2] Wed, 19 Apr 2017 22:22:52 GMT (1762kb,D)

Link back to: arXiv, form interface, contact.