Ferenc Huszár • 10/31/2019

The secular Bayesian: Using belief distributions without really believing

The article is a personal narrative about a data scientist's evolution from a rigid, 'religious' adherence to Bayesian statistics to a more pragmatic, 'secular' approach. It discusses the philosophical conflict between Bayesian and frequentist perspectives, the realization of model limitations and misspecification, and the practical utility of Bayesian tools like priors and posteriors without requiring literal belief in the model's existence.

0 comments

#Machine Learning #statistics #Data Science