Location visible to members
Our June meeting will feature a talk from Colin Carrol, of Kensho Technologies, about Hamiltonian Monte Carlo sampling with PyMC3.
Hamiltonian Monte Carlo in PyMC3.
Probabilistic programming allows a user to specify a Bayesian model in code and perform inference on that model in the presence of observed data. Markov chain Monte Carlo (MCMC) is a flexible method for sampling from the posterior distribution of these models, and Hamiltonian Monte Carlo is a particularly efficient implementation of MCMC, allowing it to be applied to more complex models. Hamiltonian Monte Carlo was first described 30 years ago, began to be applied to statistics about 20 years ago, started appearing in textbooks 10 years ago, and became easily available in software libraries only in the last few years. This talk will begin with an introduction to MCMC algorithms, then move to the theory and practice of Hamiltonian Monte Carlo algorithms. Examples will be provided using the Python probabilistic programming library PyMC3.
Colin Carroll is a software engineer at Kensho Technologies, a startup applying machine learning to finance. He is a developer on PyMC3, a python library for probabilistic programming. He received his PhD in mathematics from Rice University, where he researched geometric measure theory.
• 6:30pm: Meet and Greet. Networking
• 7:00pm: Talk by Colin Carrol
• 7:45pm: Q&A and general discussion
• 8:00pm: Networking
• 9:00pm: End of the event
This event is sponsored by our friends at Acquia.