Probabilistic programming is based on the insight that probabilistic models and inference algorithms can be viewed as a new kind of software, and that this perspective can lead to radical improvements in developer productivity, application robustness, and scale. This talk will illustrate the paradigms of probabilistic programming using example probabilistic programs written in three languages under active development by the MIT Probabilistic Computing Project:

1. Gen, a probabilistic programming language with fast programmable inference, embedded in Julia and integrated with TensorFlow. Gen allows users to concisely express generative models and perform inference using combinations of state-of-the-art sequential Monte Carlo, gradient-based optimization, and deep learning techniques.

2. BayesDB, a probabilistic programming platform for machine-assisted data science, which lets users without statistics training query probabilistic implications of their data using an SQL-like language. BayesDB automatically builds models by synthesizing probabilistic programs from data, using new Bayesian model discovery techniques.

3. Metaprob, a probabilistic meta-programming language embedded in Clojure. Metaprob makes it possible to distinctive features of other probabilistic programming languages as a short user-space library.

Applications will be drawn from machine perception, such as inferring 3D models of human bodies from single images, and machine-assisted data science, such as discovering structure from econometric time series data and identifying indicators of suicide risk from psychiatric screening questionnaires.

Vikash Mansinghka is a research scientist at MIT, where he leads the Probabilistic Computing Project. Vikash holds S.B. degrees in Mathematics and in Computer Science from MIT, as well as an M.Eng. in Computer Science and a PhD in Computation. He also held graduate fellowships from the National Science Foundation and MIT’s Lincoln Laboratory. His PhD dissertation on natively probabilistic computation won the MIT George M. Sprowls dissertation award in computer science, and his research on the Picture probabilistic programming language won an award at CVPR. He served on DARPA’s Information Science and Technology advisory board from 2010-2012, and currently serves on the editorial boards for the Journal of Machine Learning Research and the journal Statistics and Computation. He was an advisor to Google DeepMind and has co-founded two AI-related startups, one acquired and one currently operational.