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

Download:

Current browse context:

stat.ML
< prev | next >
new | recent | 1704

Change to browse by:

cs
cs.LG
stat

References & Citations

Bookmark

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

Statistics > Machine Learning

Title: A comparative study of counterfactual estimators

Authors: Thomas Nedelec, Nicolas Le Roux, Vianney Perchet
(Submitted on 3 Apr 2017 (v1), last revised 29 Jan 2019 (this version, v3))
Abstract: We provide a comparative study of several widely used off-policy estimators (Empirical Average, Basic Importance Sampling and Normalized Importance Sampling), detailing the different regimes where they are individually suboptimal. We then exhibit properties optimal estimators should possess. In the case where examples have been gathered using multiple policies, we show that fused estimators dominate basic ones but can still be improved.
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:1704.00773 [stat.ML]
  (or arXiv:1704.00773v3 [stat.ML] for this version)

Submission history

From: Thomas Nedelec [view email]
[v1] Mon, 3 Apr 2017 19:16:06 GMT (390kb,D)
[v2] Tue, 2 May 2017 08:00:07 GMT (392kb,D)
[v3] Tue, 29 Jan 2019 13:58:22 GMT (403kb,D)
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Link back to: arXiv, form interface, contact.