Timezone: »

A Smoother Way to Train Structured Prediction Models
Krishna Pillutla · Vincent Roulet · Sham Kakade · Zaid Harchaoui

Thu Dec 06 02:00 PM -- 04:00 PM (PST) @ Room 517 AB #102

We present a framework to train a structured prediction model by performing smoothing on the inference algorithm it builds upon. Smoothing overcomes the non-smoothness inherent to the maximum margin structured prediction objective, and paves the way for the use of fast primal gradient-based optimization algorithms. We illustrate the proposed framework by developing a novel primal incremental optimization algorithm for the structural support vector machine. The proposed algorithm blends an extrapolation scheme for acceleration and an adaptive smoothing scheme and builds upon the stochastic variance-reduced gradient algorithm. We establish its worst-case global complexity bound and study several practical variants. We present experimental results on two real-world problems, namely named entity recognition and visual object localization. The experimental results show that the proposed framework allows us to build upon efficient inference algorithms to develop large-scale optimization algorithms for structured prediction which can achieve competitive performance on the two real-world problems.

Author Information

Krishna Pillutla (University of Washington)
Vincent Roulet (UW)
Sham Kakade (University of Washington)
Zaid Harchaoui (University of Washington)

More from the Same Authors