March 26, 2010 Workshop

March 26, 2010

Workshop: Support Vector Machines and Predictive Data Modeling Methodology

Vladimir Cherkassky
Electrical & Computer Engineering
University of Minnesota – Twin Cities
Minneapolis MN 55455 Email: cherk001@umn.edu

Friday, March 26, 2010
8:30 AM-5:00 PM

Gleacher Center
450 North Cityfront Plaza Drive
Chicago, IL 60611-4316
312-464-8787

Sponsored by the Chicago Chapter of the American Statistical Association (http://www.chicagoasa.org/).

Abstract

There is a growing interest in estimating predictive models from empirical data. The subject of data-driven modeling has been addressed in various disciplines such as statistics, pattern recognition, signal processing, genomics, artificial neural networks, machine learning, and data mining. Since late 1990’s, many researchers and practitioners alike are actively applying the so-called Support Vector Machine (SVM) methods developed under the framework of Vapnik-Chervonenkis (VC) learning theory. This workshop pursues several methodological, practical and research goals, aimed to provide in-depth understanding of SVM methodology and the underlying VC-theoretical concepts. The workshop participants will gain understanding of:

1. The difference between classical statistical and predictive learning (VC-theoretical) methodologies for estimating models from data
2. The main ideas, concepts and results of VC learning theory
3. SVM methodology for different types of learning problems, such as classification, regression and single-class learning
4. SVM model selection and other practical issues arising in practical applications;
5. New advanced learning methodologies based on recent SVM extensions and modifications. These new methodologies include Learning in the Universum environment (U-SVM), transduction (transductive SVM) and Learning with Structured Data (aka SVM+)

Course outline

Morning Session 1:

Philosophical and methodological differences between classical statistical and predictive learning approaches to data modeling

Morning Session 2:

Margin-based Methods and Support Vector Machines

• Motivation for margin-based loss
• SVM Classification
• SVM Regression
• Issues for SVM model selection

Afternoon Session 3:

Practical issues and SVM Application Studies

• Practical issues in using SVM software
• Interpretation of SVM classifiers
• Empirical comparison of different learning methods
• Application study 1: fraud detection
• Application study 2: prediction of transplant-related mortality
• Application study 3: real-time prediction of epileptic seizures

Afternoon Session 4:

Advanced SVM-based learning technologies

• Non-inductive and alternative learning settings
• Transduction
• Universum Learning
• Learning with Structured Data and Multi-Task Learning

for Chicago ASA Chapter members