Most classification algorithms were originally designed for fixed-size vectors. However, important machine learning problems in computational biology, or text and speech processing, require the analysis of
variable-length sequences or, more generally, distributions over variable-length sequences.
This tutorial introduces Rational Kernels, a unifying learning framework for the analysis of text, speech, and biological sequences.
It describes general and efficient algorithms for computing rational kernels, presents several effective kernels, and illustrates their
successful use in some difficult prediction problems.
It also presents several theoretical results guiding the design of the rational kernels and shows that most similarity measures commonly used
in text classification or computational biology are special instances of rational kernels.
[Joint work with Corinna Cortes and Patrick Haffner]
Presentation (PDF File)
Video of Talk (RealPlayer File)
Copyright. All Rights Reserved.