Friday, October 28, 2016 3:30 pm - 4:00 pm
It is an exciting time in Machine Learning, with major hardware advances and the recent popularization of Deep Learning. Researchers are still learning quite a bit of what can be done to facilitate machine understanding, and how to do it. Still, a number of patterns have emerged in Deep Learning, even in these early stages. In this talk I will cover some of the patterns and idioms that have become fairly standard in deep learning, including some details on very deep architectures, fine-tuning, language modeling and encoder-decoder networks. I will talk about some of the ways that these techniques have been applied to facilitate machine understanding for Image Processing, Speech Recognition and Natural Language Processing.