Timezone: »

 
Poster
Implicit Task-Driven Probability Discrepancy Measure for Unsupervised Domain Adaptation
Mao Li · Kaiqi Jiang · Xinhua Zhang

Tue Dec 07 08:30 AM -- 10:00 AM (PST) @
in Poster Session 1 »

Probability discrepancy measure is a fundamental construct for numerous machine learning models such as weakly supervised learning and generative modeling. However, most measures overlook the fact that the distributions are not the end-product of learning, but are the basis of downstream predictor. Therefore it is important to warp the probability discrepancy measure towards the end tasks, and we hence propose a new bi-level optimization based approach so that the two distributions are compared not uniformly against the entire hypothesis space, but only with respect to the optimal predictor for the downstream end task. When applied to margin disparity discrepancy and contrastive domain discrepancy, our method significantly improves the performance in unsupervised domain adaptation, and enjoys a much more principled training process.

Keywords: Optimization · Machine Learning · Generative Model · Domain Adaptation

Author Information

Mao Li (University of Illinois at Chicago)
Kaiqi Jiang (University of Illinois at Chicago)

I am currently a Ph.D. student concentrating on machine learning working with Professor Xinhua Zhang. My current research is domain adaptation and fairness.

Xinhua Zhang (University of Illinois at Chicago (UIC))

More from the Same Authors