A New Discriminative Kernel from Probabilistic Models
- 1 October 2002
- journal article
- Published by MIT Press in Neural Computation
- Vol. 14 (10) , 2397-2414
Abstract
Recently, Jaakkola and Haussler (1999) proposed a method for constructing kernel functions from probabilistic models. Their so-called Fisher kernel has been combined with discriminative classifiers such as support vector machines and applied successfully in, for example, DNA and protein analysis. Whereas the Fisher kernel is calculated from the marginal log-likelihood, we propose the TOP kernel derived from tangent vectors of posterior log-odds. Furthermore, we develop a theoretical framework on feature extractors from probabilistic models and use it for analyzing the TOP kernel. In experiments, our new discriminative TOP kernel compares favorably to the Fisher kernel.Keywords
This publication has 6 references indexed in Scilit:
- Soft Margins for AdaBoostMachine Learning, 2001
- A Discriminative Framework for Detecting Remote Protein HomologiesJournal of Computational Biology, 2000
- Pairwise data clustering by deterministic annealingIEEE Transactions on Pattern Analysis and Machine Intelligence, 1997
- Network information criterion-determining the number of hidden units for an artificial neural network modelIEEE Transactions on Neural Networks, 1994
- Discriminant analysis using non-metric multidimensional scalingPattern Recognition, 1993
- A tutorial on hidden Markov models and selected applications in speech recognitionProceedings of the IEEE, 1989