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

Download:

Current browse context:

math.ST

Change to browse by:

References & Citations

Bookmark

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

Mathematics > Statistics Theory

Title: Sub-Gaussian estimators of the mean of a random vector

Abstract: We study the problem of estimating the mean of a random vector $X$ given a sample of $N$ independent, identically distributed points. We introduce a new estimator that achieves a purely sub-Gaussian performance under the only condition that the second moment of $X$ exists. The estimator is based on a novel concept of a multivariate median.
Comments: 12 pages
Subjects: Statistics Theory (math.ST); Machine Learning (stat.ML)
Cite as: arXiv:1702.00482 [math.ST]
  (or arXiv:1702.00482v1 [math.ST] for this version)

Submission history

From: Gabor Lugosi [view email]
[v1] Wed, 1 Feb 2017 22:33:36 GMT (12kb)

Link back to: arXiv, form interface, contact.