Matthew Johnson

Matthew James Johnson mattjj@csail.mit.edu mattjj@google.com @SingularMattrix aboutrésumépublicationsteachingcode About matt I'm a senior research scientist at Google Brain. My research interests lie broadly in machine learning, statistics, and optimization. I was a postdoc with Ryan Adams in the Harvard Intelligent Probabilistic Systems (HIPS) group and Bob Datta in the Datta Lab in Neurobiology at the Harvard Medical School. I was a Ph.D. student in EECS at MIT. I worked with Alan Willsky in the Stochastic Systems Group, which is part of LIDS and also strongly connected to CSAIL. My thesis committee included Ryan Adams, David Blei, and Tommi Jaakkola. I was an undergraduate in EECS at UC Berkeley, where I worked with Martin Wainwright and Dan Garcia. I am married to the wonderful and lovely Lily Johnson (that's us at the top!). Research highlights link to section Composing graphical models with neural networks for structured representations and fast inference (NIPS 2016) link to section Dependent multinomial models made easy: stick-breaking with the Polya-gamma augmentation (NIPS 2015) link to section Natural gradient stochastic variational inference for Bayesian nonparametric HMMs and HSMMs (ICML 2014) link to section Analyzing Hogwild parallel Gibbs sampling in Gaussian models (NIPS 2013)