[Theory Lunch 10/16] Haoming Li: Classification with Strategically Withheld Data
5 views
Skip to first unread message
Haoming Li
Oct 12, 2020, 11:49:30 PM10/12/20
to usc-theo...@googlegroups.com, usc-t...@googlegroups.com
USC CS Theory Lunch:
Classification with Strategically Withheld Data
Speaker: Haoming Li (https://haoming.li/)
Time: 10/16/20 12:00pm
Location: https://usc.zoom.us/j/94386654763
Abstract:
Machine learning techniques can be useful in applications such as credit approval and college admission. However, to be classified more favorably in such contexts, an agent may decide to strategically withhold some of her features, such as bad test scores. This is a missing data problem with a twist: which data is missing depends on the chosen classifier, because the specific classifier is what may create the incentive to withhold certain feature values. We address the problem of training classifiers that are robust to this behavior.
Classification with Strategically Withheld Data
Speaker: Haoming Li (https://haoming.li/)
Time: 10/16/20 12:00pm
Location: https://usc.zoom.us/j/94386654763
Abstract:
Machine learning techniques can be useful in applications such as credit approval and college admission. However, to be classified more favorably in such contexts, an agent may decide to strategically withhold some of her features, such as bad test scores. This is a missing data problem with a twist: which data is missing depends on the chosen classifier, because the specific classifier is what may create the incentive to withhold certain feature values. We address the problem of training classifiers that are robust to this behavior.
Based on two papers:
Reply all
Reply to author
Forward
0 new messages