Basser Seminar Series

Stan: A probabilistic programming language for Bayesian inference

Speaker: Dr Bob Carpenter
Columbia University, Department of Statistics

When: Monday 30 March 2015, 2-3pm, *Note different day and time to usual.

Where: The University of Sydney, School of IT Building, SIT Lecture Theatre (Room 123), Level 1

Add seminar to my diary

Abstract

I'll describe Stan's probabilistic programming language, and how it's used, including:

  • examples of Stan progams
  • blocks for data, parameter, and predictive quantities
  • transforms of constrained parameters to unconstrained spaces, with automatic Jacobian corrections
  • automatic computation of first- and higher-order derivatives
  • operator, function, and linear algebra library
  • vectorized density functions, cumulative distributions, and random number generators
  • user-defined functions
  • ordinary differential equation solvers

I'll also provide an overview of the underlying algorithms for sampling and optimization:

  • adaptive Hamiltonian Monte Carlo for MCMC
  • L-BFGS optimization and transforms for MLE

I'll also briefly describe the user-facing interfaces:

  • RStan (R), PyStan (Python), CmdStan (command line), Stan.jl (Julia), MatlabStan (MATLAB)

I'll finish with an overview of the what's next:

  • data streaming variational inference
  • data parallel expectation propagation
  • marginal maximum likelihood for empirical Bayes
  • stiff ODE solvers

Speaker's biography

Bob has worked as a professor of computational linguistics (Carnegie
Mellon), an industrial researcher and programmer (Bell Labs, SpeechWorks, LingPipe), and is now back in academia working in statistics (Columbia Univesity).