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

Download:

Current browse context:

cs
< prev | next >
new | recent | 1712

Change to browse by:

cs.LG
math
math.PR
math.ST
stat

References & Citations

DBLP - CS Bibliography

listing | bibtex

Bookmark

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

Mathematics > Statistics Theory

Title: Exact upper and lower bounds on the misclassification probability

Authors: Iosif Pinelis
(Submitted on 3 Dec 2017 (v1), last revised 9 Feb 2018 (this version, v4))
Abstract: Exact lower and upper bounds on the best possible misclassification probability for a finite number of classes are obtained in terms of the total variation norms of the differences between the sub-distributions over the classes. These bounds are compared with the exact bounds in terms of the conditional entropy obtained by Feder and Merhav.
Comments: Version 3: exact upper bounds are added; results are compared with ones by Feder and Merhav. Version 4: additional discussion presented
Subjects: Statistics Theory (math.ST); Machine Learning (cs.LG); Probability (math.PR)
MSC classes: 60E15, 62H30, 91B06 (Primary) 26D15, 26D20, 62C10, 68T10 (Secondary)
Cite as: arXiv:1712.00812 [math.ST]
  (or arXiv:1712.00812v4 [math.ST] for this version)

Submission history

From: Iosif Pinelis [view email]
[v1] Sun, 3 Dec 2017 18:42:40 GMT (6kb)
[v2] Wed, 6 Dec 2017 16:47:32 GMT (7kb)
[v3] Mon, 5 Feb 2018 02:00:29 GMT (128kb,D)
[v4] Fri, 9 Feb 2018 02:12:50 GMT (130kb,D)
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Link back to: arXiv, form interface, contact.