"Machine Learning Reductions"

04/07/06: 10:30 AM, webcast
11th Floor Large Conference Room
Host: Patrick Pantel and Jafar Adibi, schedule

Abstract: Is every learning problem unique? Or are they related? How related are they? I will describe a mathematical formalism for answering these questions. In practical applications the algorithms implied by this formalism yield effective learning algorithms capable of solving a very wide array of learning problems. These solutions are (empirically) competitive with or superior to existing techniques for the many tests we have run so far. This mathematics formalizes and extends the intuitions of a number of algorithms created by various people.

About John Langford: John Langford is a "Research Associate Professor" at TTI-Chicago. He graduated with a PhD in Computer Science from CMU in 2002 and with a physics/CS BS/BS from Caltech in 1997. He is broadly interested in machine learning with papers covering a wide array of topics. John also runs the Machine Learning (Theory) blog at http://hunch.net

Last updated: Mon Jun 19 17:44:06 2006