Nonparametric Kernel Estimators for Image Classification (2012)

Barnabas Poczos, Liang Xiong, Dougal J. Sutherland, Jeff Schneider


We introduce a new discriminative learning method for image classification. We assume that the images are represented by unordered, multi-dimensional, finite sets of feature vectors, and that these sets might have different cardinality. By means of consistent nonparametric divergence estimators we define new kernels over these sets, and then apply them in kernel classifiers. Our numerical results demonstrate that in many cases this approach can outperform state-of-the-art competitors on both simulated and challenging real-world datasets.

Full text

Download (application/pdf, 666.5 kB)


Code to reproduce the experiments: feature extraction, divergence estimation, and SVM evaluation. Includes the OT dataset and static versions of the C++ estimator for Linux x64.

Related Software

Nonparametric divergence estimators (C++ static versions included in the .tar.bz2 above; this link includes a pure-Matlab version as well as the C++ source).
C++ support distribution machine wrappers (not used above; a pure-Matlab version is used instead).

Approximate BibTeX Entry

    Howpublished = {CVPR 2012},
    Year = {2012},
    Journal = {CVPR 2012},
    Booktitle = {CVPR 2012},
    Author = { Barnabas Poczos, Liang Xiong, Dougal J. Sutherland, Jeff Schneider },
    Title = {Nonparametric Kernel Estimators for Image Classification}

Copyright 2010, Carnegie Mellon University, Auton Lab. All Rights Reserved.