Readspeaker Menü

Approximations of Gaussian Process Regression

Paul Bodesheim, Alexander Freytag, Erik Rodner, and Joachim Denzler

 

[Bodesheim13:AOG]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler. Approximations of Gaussian Process Uncertainties for Visual Recognition Problems. Scandinavian Conference on Image Analysis (SCIA). 2013. pp. 182--194. [pdf] [bib]

[Bodesheim13:AEA]

Paul Bodesheim and Alexander Freytag and Erik Rodner and Joachim Denzler: An Efficient Approximation for Gaussian Process Regression. Technical Report TR-FSU-INF-CV-2013-01, Computer Vision Group, Friedrich Schiller University Jena, Germany. 2013. [pdf] [bib]


Code available!

The Matlab source code is available [here].

lgpl-logo


Summary

Gaussian processes offer the advantage of calculating the classification uncertainty in terms of predictive variance associated with the classification result. This is especially useful to select informative samples in active learning and to spot samples of previously unseen classes known as novelty detection. However, the Gaussian process framework suffers from high computational complexity leading to computation times too large for practical applications. Hence, we propose an approximation of the Gaussian process predictive variance leading to rigorous speedups. The complexity of both learning and testing the classification model regarding computational time and memory demand decreases by one order with respect to the number of training samples involved. The benefits of our approximations are verified in experimental evaluations for novelty detection and active learning of visual object categories on the datasets C-Pascal of Pascal VOC 2008, Caltech-256, and ImageNet.


Downloads

  • The SCIA'13 [paper]
  • A [technical report] containing additional information
  • Matlab source [code]