Christopher Fonnesbeck Probabilistic Programming with PyMC3 PyCon 2017

"Speaker: Christopher Fonnesbeck
Bayesian statistics offers robust and flexible methods for data analysis that, because they are based on probability models, have the added benefit of being readily interpretable by non-statisticians. Until recently, however, the implementation of Bayesian models has been prohibitively complex for use by most analysts. But, the advent of probabilistic programming has served to abstract the complexity of Bayesian statistics, making such methods more broadly available. PyMC3 is a open-source Python module for probabilistic programming that implements several modern, computationally-intensive statistical algorithms for fitting Bayesian models, including Hamiltonian Monte Carlo (HMC) and variational inference. PyMC3’s intuitive syntax is helpful for new users, and the reliance on Theano for much of the computational work has allowed developers to keep the code base simple, making it easy to extend the software to meet analytic needs. PyMC3 itself extends Python's powerful ""scientific stack"" of development tools, which provide fast and efficient data structures, parallel processing, and interfaces for describing statistical models.
Slides can be found at: speakerdeck.com/pycon2017 and github.com/PyCon/2017-slides"

[@inspectahblock1506]

great talk, thank you

[@MarkJay]

Great talk. A bit over my head so I've got some work to do

[@EvanZamir]

Probably not the best place to ask this, but I've recently come across Edward from David Blei's group. It has a Python API and is built on top of TensorFlow. Wondering how does it compare to PyMC3?

[@CharlesLao]

Edward is more focus on variational inference, and the inference of neural network and deep learning (e.g., it has some specific inference algorithm for GAN), PyMC3 is closer to Stan as a more general tool for building probabilistic models. Both PyMC3 and Edward are still rapidly developing and changing so this might change in the future.