On Monday morning the PyMC dev team pushed the first release of PyMC3 , the culmination of over 5 years of collaborative work. We are very pleased to be able to provide a stable version of the package to the Python scientific computing community. For those of you unfamiliar with the history and progression of this project, PyMC3 is a complete re-design and re-write of the PyMC code base, which was primarily the product of the vision and work of John Salvatier. While PyMC 2.3 is still actively maintained and
While PyMC2 relied on Fortran extensions (via f2py ) for most of the computational heavy-lifting, PyMC3 leverages Theano , a library from the LISA lab for array-based expression evaluation, to perform its computation. What this provides, above all else, is fast automatic differentiation, which is at the heart of the gradient-based sampling and optimization methods currently providing inference for probabilistic programming. While the addition of Theano adds a level of complexity to the development of PyMC,
As a point of comparison, here is what a simple hierarchical model (taken from Gelman et al. 's book ) looked like under PyMC 2.3: