Published February 16, 2015
| Published
Journal Article
Open
Mismatched Training and Test Distributions Can Outperform Matched Ones
- Creators
- González, Carlos R.
- Abu-Mostafa, Yaser S.
Abstract
In learning theory, the training and test sets are assumed to be drawn from the same probability distribution. This assumption is also followed in practical situations, where matching the training and test distributions is considered desirable. Contrary to conventional wisdom, we show that mismatched training and test distributions in supervised learning can in fact outperform matched distributions in terms of the bottom line, the out-of-sample performance, independent of the target function in question. This surprising result has theoretical and algorithmic ramifications that we discuss.
Additional Information
© 2015 Massachusetts Institute of Technology. Published under a Creative Commons Attribution 3.0 Unported (CC BY 3.0) license Received April 17, 2014; accepted August 19, 2014. Posted Online December 16, 2014.Attached Files
Published - Gonzalez-AbuMostafa-MismatchedDists.pdf
Files
Gonzalez-AbuMostafa-MismatchedDists.pdf
Files
(585.3 kB)
Name | Size | Download all |
---|---|---|
md5:9f8eaf5a17bbd92957206d42ab0e81ef
|
585.3 kB | Preview Download |
Additional details
- Eprint ID
- 53018
- Resolver ID
- CaltechAUTHORS:20141218-110007751
- Created
-
2014-12-18Created from EPrint's datestamp field
- Updated
-
2021-11-10Created from EPrint's last_modified field