Distributionally Robust Learning for Unsupervised Domain Adaptation

10/08/2020
by   Haoxuan Wang, et al.
2

We propose a distributionally robust learning (DRL) method for unsupervised domain adaptation (UDA) that scales to modern computer vision benchmarks. DRL can be naturally formulated as a competitive two-player game between a predictor and an adversary that is allowed to corrupt the labels, subject to certain constraints, and reduces to incorporating a density ratio between the source and target domains (under the standard log loss). This formulation motivates the use of two neural networks that are jointly trained - a discriminative network between the source and target domains for density-ratio estimation, in addition to the standard classification network. The use of a density ratio in DRL prevents the model from being overconfident on target inputs far away from the source domain. Thus, DRL provides conservative confidence estimation in the target domain, even when the target labels are not available. This conservatism motivates the use of DRL in self-training for sample selection, and we term the approach distributionally robust self-training (DRST). In our experiments, DRST generates more calibrated probabilities and achieves state-of-the-art self-training accuracy on benchmark datasets. We demonstrate that DRST captures shape features more effectively, and reduces the extent of distributional shift during self-training.

READ FULL TEXT

page 3

page 7

page 15

page 16

research
11/18/2020

FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation

Unsupervised domain adaptation (UDA) methods for learning domain invaria...
research
09/10/2018

Improving Adversarial Discriminative Domain Adaptation

Adversarial discriminative domain adaptation (ADDA) is an efficient fram...
research
09/29/2017

Unsupervised Domain Adaptation with Copula Models

We study the task of unsupervised domain adaptation, where no labeled da...
research
09/03/2022

Source-Free Unsupervised Domain Adaptation with Norm and Shape Constraints for Medical Image Segmentation

Unsupervised domain adaptation (UDA) is one of the key technologies to s...
research
07/12/2022

Domain Gap Estimation for Source Free Unsupervised Domain Adaptation with Many Classifiers

In theory, the success of unsupervised domain adaptation (UDA) largely r...
research
01/14/2020

A Sample Selection Approach for Universal Domain Adaptation

We study the problem of unsupervised domain adaption in the universal sc...
research
12/16/2019

Incorporating Unlabeled Data into Distributionally Robust Learning

We study a robust alternative to empirical risk minimization called dist...

Please sign up or login with your details

Forgot password? Click here to reset