Friday, December 4, 2015 - 12:00pm to 1:00pm
Location:McWilliams Classroom 4303 Gates & Hillman Centers
Speaker:SHIVA KAUL, Ph.D. Student http://www.cs.cmu.edu/~skkaul
Linear classifiers are convenient to train, but cannot cope with noisy or complex data. Nonlinear kernel classifiers are powerful, but are prone to overfitting and are cumbersome to train. A new kind of classifier addresses the limitations of both of its forebears, as evidenced by its surprising statistical and algebraic properties. Our classifier underlies a new learning algorithm of notable practical and theoretical interest.
Joint work with Geoff Gordon.
Presented in Partial Fulfillment of the CSD Speaking Skills Requirement.
Speaker Skills Poster