This talk has moved to Spring 2017 — Towards practical machine learning with differential privacy and beyond.