Tamir Hazan, Subhransu Maji, Joseph Keshet, Tommi Jaakkola

University of Haifa; University of Massachusetts, Amherst; Bar-Ilan University; Massachusetts Institute of Technology

Poster: Learning Efficient Random Maximum A-Posteriori Predictors with Non-Decomposable Loss Functions

7:00 – 11:59pm Friday, December 06, 2013

Harrah's Special Events Center, 2nd Floor

This is part of the Poster Session which begins at 19:00 on Friday December 6, 2013


Get the paper

In this work we develop efficient methods for learning random MAP predictors for structured label problems. In particular, we construct posterior distributions over perturbations that can be adjusted via stochastic gradient methods. We show that every smooth posterior distribution would suffice to define a smooth PAC-Bayesian risk bound suitable for gradient methods. In addition, we relate the posterior distributions to computational properties of the MAP predictors. We suggest multiplicative posteriors to learn super-modular potential functions that accompany specialized MAP predictors such as graph-cuts. We also describe label-augmented posterior models that can use efficient MAP approximations, such as those arising from linear program relaxations.