Min-Max Statistical Alignment for Transfer Learning

Samitha Herath, Mehrtash Harandi, Basura Fernando, Richard Nock; The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 9288-9297

Abstract


A profound idea in learning invariant features for transfer learning is to align statistical properties of the domains. In practice, this is achieved by minimizing the disparity between the domains, usually measured in terms of their statistical properties. We question the capability of this school of thought and propose to minimize the maximum disparity between domains. Furthermore, we develop an end-to-end learning scheme that enables us to benefit from the proposed min-max strategy in training deep models. We show that the min-max solution can outperform the existing statistical alignment solutions, and can compete with state-of-the-art solutions on two challenging learning tasks, namely, Unsupervised Domain Adaptation (UDA) and Zero-Shot Learning (ZSL).

Related Material


[pdf] [supp]
[bibtex]
@InProceedings{Herath_2019_CVPR,
author = {Herath, Samitha and Harandi, Mehrtash and Fernando, Basura and Nock, Richard},
title = {Min-Max Statistical Alignment for Transfer Learning},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2019}
}