Showing papers on "Multiple kernel learning published in 2006"

Large Scale Multiple Kernel Learning

Abstract: While classical kernel-based learning algorithms are based on a single kernel, in practice it is often desirable to use multiple kernels. Lanckriet et al. (2004) considered conic combinations of kernel matrices for classification, leading to a convex quadratically constrained quadratic program. We show that it can be rewritten as a semi-infinite linear program that can be efficiently solved by recycling the standard SVM implementations. Moreover, we generalize the formulation and our method to a larger class of problems, including regression and one-class classification. Experimental results show that the proposed algorithm works for hundred thousands of examples or hundreds of kernels to be combined, and helps for automatic model selection, improving the interpretability of the learning result. In a second part we discuss general speed up mechanism for SVMs, especially when used with sparse feature maps as appear for string kernels, allowing us to train a string kernel SVM on a 10 million real-world splice data set from computational biology. We integrated multiple kernel learning in our machine learning toolbox SHOGUN for which the source code is publicly available at http://www.fml.tuebingen.mpg.de/raetsch/projects/shogun .

Learning interpretable SVMs for biological sequence classification.

Abstract: Support Vector Machines (SVMs) – using a variety of string kernels – have been successfully applied to biological sequence classification problems. While SVMs achieve high classification accuracy they lack interpretability. In many applications, it does not suffice that an algorithm just detects a biological signal in the sequence, but it should also provide means to interpret its solution in order to gain biological insight. We propose novel and efficient algorithms for solving the so-called Support Vector Multiple Kernel Learning problem. The developed techniques can be used to understand the obtained support vector decision function in order to extract biologically relevant knowledge about the sequence analysis problem at hand. We apply the proposed methods to the task of acceptor splice site prediction and to the problem of recognizing alternatively spliced exons. Our algorithms compute sparse weightings of substring locations, highlighting which parts of the sequence are important for discrimination. The proposed method is able to deal with thousands of examples while combining hundreds of kernels within reasonable time, and reliably identifies a few statistically significant positions.

Object categorization using kernels combining graphs and histograms of gradients

Abstract: This paper presents a method for object categorization. This problem is difficult and can be solved by combining different information sources such as shape or appearance. In this paper, we aim at performing object recognition by mixing kernels obtained from different cues. Our method is based on two complementary descriptions of an object. First, we describe its shape thanks to labeled graphs. This graph is obtained from morphological skeleton, extracted from the binary mask of the object image. The second description uses histograms of oriented gradients which aim at capturing objects appearance. The histogram descriptor is obtained by computing local histograms over the complete image of the object. These two descriptions are combined using a kernel product. Our approach has been validated on the ETH80 database which is composed of 3280 images gathered in 8 classes. The results we achieved show that this method can be very efficient.

Large Scale Multiple Kernel Learning

Abstract: While classical kernel-based learning algorithms are based on a single kernel, in practice it is often desirable to use multiple kernels. Lanckriet et al. (2004) considered conic combinations of ke...

Detecting Generic Visual Eventswith Temporal Cues

Abstract: We present novel algorithms for detecting generic visual events from video. Target event models will produce binary decisions on each shot about classes of events involving object actions and their interactions with the scene, such as airplane taking off, exiting car, riot. While event detection has been studied in scenarios with strong scene and imaging assumptions, the detection of generic visual events from an unconstrained domain such as broadcast news has not been explored. This work extends our recent work [3] on event detection by (1) using a novel bag-of-features representation along with the earth movers' distance to account for the temporal variations within a shot, (2) learn the importance among input modalities with a double-convex combination along both different kernels and different support vectors, which is in turn solved via multiple kernel learning. Experiments show that the bag-of-features representation significantly outperforms the static baseline; multiple kernel learning yields promising performance improvement while providing intuitive explanations for the importance of the input kernels.