approxposterior: Approximate Posterior Distributions in Python
D. P. Fleming, Jake Vanderplas · The Journal of Open Source Software · 2018
This package is a Python implementation of "Bayesian Active Learning for Posterior Estimation" by (Kandasamy, Schneider, & Poczos, 2015) and "Adaptive Gaussian process approximation for Bayesian inference with expensive likelihood functions" (Wang & Li, 2017).These algorithms allows the user to compute approximate posterior probability distributions using computationally expensive forward models by training a Gaussian Process (GP) surrogate for the likelihood evaluation.The algorithms leverage the inherent uncertainty in the GP's predictions to identify high-likelihood regions in parameter space where the GP is uncertain.The algorithms then run the forward model at these points to compute their likelihood and re-trains the GP to maximize the GP's predictive ability while minimizing the number of forward model evaluations.Please read (Kandasamy et al., 2015) and (Wang & Li, 2017) for in-depth descriptions of the respective algorithms.approxposterior is under active development on GitHub and community participation is encouraged.The code is available on GitHub (Fleming, 2018).