Publication Details
Title: Rapid Uncertainty Computation with Gaussian Processes and Histogram Intersection Kernels
Author: A. Freytag, E. Rodner, P. Bodesheim, and J. Denzler
Group: Vision
Date: November 2012
PDF: http://www.icsi.berkeley.edu/pubs/vision/ICSI_rapiduncertainty12.pdf
Overview:
An important advantage of Gaussian processes is the ability to directly estimate classi cation uncertainties in a Bayesian manner. In this paper, we develop techniques that allow for estimating these uncertainties with a runtime linear or even constant with respect to the number of training examples. Our approach makes use of all training data without any sparse approximation technique while needing only a linear amount of memory. To incorporate new information over time, we further derive online learning methods leading to signi cant speed-ups and allowing for hyperparameter optimization on-the- y. We conduct several experiments on public image datasets for the tasks of one-class classi cation and active learning, where computing the uncertainty is an essential task. The experimental results highlight that we are able to compute classi cation uncertainties within microseconds even for largescale datasets with tens of thousands of training examples.
Bibliographic Information:
Proceedings of the 11th Asian Conference on Computer Vision, Daejeon, Korea. Recipient of a best paper honorable mention award.
Bibliographic Reference:
A. Freytag, E. Rodner, P. Bodesheim, and J. Denzler. Rapid Uncertainty Computation with Gaussian Processes and Histogram Intersection Kernels. Proceedings of the 11th Asian Conference on Computer Vision, Daejeon, Korea. Recipient of a best paper honorable mention award., November 2012
Author: A. Freytag, E. Rodner, P. Bodesheim, and J. Denzler
Group: Vision
Date: November 2012
PDF: http://www.icsi.berkeley.edu/pubs/vision/ICSI_rapiduncertainty12.pdf
Overview:
An important advantage of Gaussian processes is the ability to directly estimate classi cation uncertainties in a Bayesian manner. In this paper, we develop techniques that allow for estimating these uncertainties with a runtime linear or even constant with respect to the number of training examples. Our approach makes use of all training data without any sparse approximation technique while needing only a linear amount of memory. To incorporate new information over time, we further derive online learning methods leading to signi cant speed-ups and allowing for hyperparameter optimization on-the- y. We conduct several experiments on public image datasets for the tasks of one-class classi cation and active learning, where computing the uncertainty is an essential task. The experimental results highlight that we are able to compute classi cation uncertainties within microseconds even for largescale datasets with tens of thousands of training examples.
Bibliographic Information:
Proceedings of the 11th Asian Conference on Computer Vision, Daejeon, Korea. Recipient of a best paper honorable mention award.
Bibliographic Reference:
A. Freytag, E. Rodner, P. Bodesheim, and J. Denzler. Rapid Uncertainty Computation with Gaussian Processes and Histogram Intersection Kernels. Proceedings of the 11th Asian Conference on Computer Vision, Daejeon, Korea. Recipient of a best paper honorable mention award., November 2012