University of Cambridge > Talks.cam > Machine Learning Reading Group @ CUED > Large-scale sequential experimentation

Large-scale sequential experimentation

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Elre Oldewage.

Standard hypothesis tests require you to commit to a prespecified sample size. It is not allowed to peek at the data during the collection process and stop prematurely. Several famous works have unsuspectingly fallen victim to this subtle catch, impelling the authors to eventually retract the paper. In this reading group, we will look at uniform confidence sequences (UCSs), which can be used to construct testing procedures that do allow you to peek at the data and stop prematurely, and see how these UCSs connect to betting games. In preparation for the reading group, we recommend watching [1] (resp. reading [2] on which the talk is based), and optionally also watching [3] (resp. reading [4]).

[1] https://www.youtube.com/watch?v=po0EsW-ixig

[2] https://arxiv.org/abs/1810.08240

[3] https://www.youtube.com/watch?v=N3N26HQBWow

[4] https://arxiv.org/abs/2010.09686

This talk is part of the Machine Learning Reading Group @ CUED series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.

 

© 2006-2024 Talks.cam, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity