Geometry-aware deep transform

Conference Paper

Many recent efforts have been devoted to designing sophisticated deep learning structures, obtaining revolutionary results on benchmark datasets. The success of these deep learning methods mostly relies on an enormous volume of labeled training samples to learn a huge number of parameters in a network, therefore, understanding the generalization ability of a learned deep network cannot be overlooked, especially when restricted to a small training set, which is the case for many applications. In this paper, we propose a novel deep learning objective formulation that unifies both the classification and metric learning criteria. We then introduce a geometry-aware deep transform to enable a non-linear discriminative and robust feature transform, which shows competitive performance on small training sets for both synthetic and real-world data. We further support the proposed framework with a formal (K)-robustness analysis.

Full Text

Duke Authors

Cited Authors

  • Huang, J; Qiu, Q; Calderbank, R; Sapiro, G

Published Date

  • February 17, 2015

Published In

Volume / Issue

  • 2015 International Conference on Computer Vision, ICCV 2015 /

Start / End Page

  • 4139 - 4147

International Standard Serial Number (ISSN)

  • 1550-5499

International Standard Book Number 13 (ISBN-13)

  • 9781467383912

Digital Object Identifier (DOI)

  • 10.1109/ICCV.2015.471

Citation Source

  • Scopus