Max Shen, PhD Student, Alvin Shi, PhD Student, Carles Boix, PhD Student
Enrollment: Unlimited: Advance sign-up required
Sign-up by 01/08
Attendance: Participants welcome at individual sessions
Recent innovations in computational methods for Bayesian inference, captured in probabilistic programming languages such as Stan and Edward, have made the power of fully Bayesian inference accessible beyond expert statisticians. These methods particularly shine in machine learning settings with small-to-medium size datasets and complex prior/domain knowledge.
This class aims to provide a hands-on introduction to applying probabilistic programming to real-world problems. The ideas behind probabilistic programming will be covered, including automatic differentiation, variational inference, Markov Chain Monte Carlo (MCMC), and other inference methods. The engineering of models will also be emphasized through exercises on debugging, model specification, reparameterization, addressing identifiability issues, and model efficiency.
Coding exercises and sample datasets will be provided. Students are also encouraged to bring in their own datasets. All course details are subject to change.
*Prior experience with Python or R recommended, as well as some experience with statistics. The class is geared towards interested undergraduates and graduate students.
*In addition, the first annual Stan Convention is occurring on January 21st at Columbia University ($50 student registration) and some of us will be attending.
*Please register here: https://goo.gl/forms/6Ovz4ferwITj7ak13
*Course Material here: https://github.com/maxwshen/iap-appbml
Contact: Max Shen, MAXWSHEN@MIT.EDU