Feature Extraction: Foundations and Applications
This book is organized around the results of a benchmark that took place in 2003 (the website of the challenge is still active: http://www.nipsfsc.ecs.soton.ac.uk/) and whose results were discussed at the NIPS 2003 workshop on feature extraction: http://clopinet.com/isabelle/Projects/NIPS2003/. Dozens of research groups competed on five large feature selection problems from various application domains: medical diagnosis, text processing, drug discovery, and handwriting recognition. The results of this effort pave the way to a new generation of methods capable of analyzing data tables with million of lines and/or columns. This book is a step towards validating, unifying, and formalizing approaches. The data and sample code with useful baseline methods implemented in Matlab are available as a single big archive (http://clopinet.com/isabelle/Projects/ETH/clopinet.com/isabelle/Projects/NIPS2003/DataNcode.zip: 93 MB). Course material is also available: http://clopinet.com/isabelle/Projects/ETH/. The introduction chapter is downloadable (courtesy of Springer): http://clopinet.com/fextract-book/IntroFS.pdf.