Features Selection

in
PhD Seminar Course on

Features Selection

Cagliari, Giugno, 7-10, 2011
Instructor: Dr. Gavin Brown
Gavin Brown is a lecturer in Machine Learning at the School of Computer Science, University of Manchester, UK. His research career began in the Multiple Classifier Systems field, studying the phenomenon of `diversity'. Since this time he has diverged to study the feature selection problem in an information theoretic framework. Applications of his work currently include bio-health informatics and adaptive compilers.
Duration: 8 hours
Schedule:
  • Motivations and Basic Methods (June 7, 9-11 A.M.)  
    • Assessment methods and Wrappers. Slides
    • Uni-variate vs Multivariate Filter Methods. Slides
  • Advanced Filters and Embedded Methods (June 8, 9-11 A.M.)  
    • Information Theoretic Filters. Slides
  • The Variance of Feature Selection (June 9, 9-11 A.M.)  
    • Why does Feature Selection Work? Slides
  • Probabilistic Perspectives (June 10, 9-11 A.M.) 
    • Markov Blanket Algorithms. Slides
    • A Unifying View via Conditional Likelihood 
Venue:
  • June 7,8,10 - room X
  • June 9 - room "Mocci" (DIEE - A Building)
Topics:
This course will introduce the problem of "feature selection" in pattern recognition and machine learning. We will first cover the motivations, assumptions, and basic techniques used in current practice - covering the three main families, "wrappers", "filters" and "embedded" methods. Once we have this foundation in place, we will discuss more advanced topics such as how these methods relate to the multiple classifier systems literature, and how cause/effect can be taken into account in the feature selection process. Finally, we will conclude with current research issues for the field, including a unifying viewpoint on the literature from an information theoretic perspective.
Organizer: Prof. Giorgio Fumera
Dep. of Electrical and Electronic Engineering
University of Cagliari, Italy
Email: fumera(at)diee(dot)unica(dot)it