Welcome to the new version of CaltechAUTHORS. Login is currently restricted to library staff. If you notice any issues, please email coda@library.caltech.edu
Published February 16, 2015 | Published
Journal Article Open

Mismatched Training and Test Distributions Can Outperform Matched Ones


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 (585.3 kB)
Name Size Download all
585.3 kB Preview Download

Additional details

August 20, 2023
October 18, 2023