NumFOCUS is pleased to announce PyMC3 as our newest fiscally sponsored project. PyMC3 is a python module for Bayesian statistical modelling and probabilistic programming which focuses on advanced Markov chain Monte Carlo and variational fitting algorithms, such as the No U-Turn Sampler (NUTS) and automatic differentiation variational inference (ADVI), respectively. The generality of these model fitting algorithms, along with the flexibility and extensibility of Python, make PyMC applicable to a large suite of problems across a range of scientific domains. PyMC3’s intuitive syntax makes model specification straightforward. In fact, simple statistical regression models can be coded in as little as a single line of Python code.
The biggest change in this new major release of PyMC is the integration of Theano as the computational backend. Theano is a rich library for the evaluation of mathematical expressions, which allows for dynamic C compilation, computational optimization, and efficient symbolic differentiation. Theano’s features provide the computational backbone that allow developers to more easily implement the newest Bayesian statistical algorithms (many of which have only been in the scientific literature for the past couple of years), particularly those that rely on gradient information in order to make model fitting more efficient.
The PyMC3 project is licensed under the Apache 2.0 license. Please visit the PyMC3 GitHub repository for the latest code, documentation, and examples in order to get started!