FoCM 2014 conference
Workshop C3 - Learning Theory
December 18, 17:30 ~ 18:00 - Room B23
Trivial Pursuit: A Shallow Learning Retrospective
Benjamin Recht
University of California, Berkeley, USA - brecht@eecs.berkeley.edu
In this talk, I review the history of random features in machine learning over the last ten years. Presenting both my work on the subject and a survey of the related literature, I will show how shallow banks of random features are able to match the performance of complicated learned or engineered features on a variety of difficult learning tasks. I will then describe the present challenges in random feature design and our progress towards making these features scalable, parsimonious, and interpretable.