Dear all,
This Friday, Rajesh Ranganath (Ph.D. student at Princeton University) will visit the lab and give a talk on his recent work on "deep exponential families". As usual, the talk will start at 14.00 in AA3195.
Hope to see many of you at the talk!
- K
===
Speaker: Rajesh Ranganath, Princeton University
Date/Time: 10 Apr @14.00
Place: AA3195
Title: Deep Exponential Families
Abstract:
We describe deep exponential families (DEFs), a class of latent variable models that are inspired by the hidden structures used in deep neural networks. DEFs capture a hierarchy of dependencies between latent variables, and are easily generalized to many settings through exponential families. We perform inference using recent ``black box" variational inference techniques. We then evaluate various DEFs on text and combine multiple DEFs into a model for pairwise recommendation data. We demonstrate that DEFs find interesting exploratory structure in large data sets, and give better predictive performance than state-of-the-art models.
Bio:
Rajesh Ranganath is a PhD student in machine learning working with David Blei. His primary interests are probabilistic modeling, approximate inference, Bayesian nonparametric statistics, and their applications to medicine and biology. Previously, he worked with Andrew Ng at SAIL on deep learning and Dan Jurafsky as part of the Stanford NLP Group on detecting human social intentions from conversations. Some of his recent work involves black-box inference methods, new types of random measures, and deep stochastic models.