Laplace Approximation
Bayesian machine learning often requires working with posterior distributions that have no known closed form. There are plenty of approximation methods, though many tend to be difficult to understand and implement or too computationally expensive. This post provides an overview of the Laplace Approximation: an exceptionally simple approach that is seeing a resurgence in recent deep learning papers.
The Basics of Bayesian Machine Learning
Bayesian methods can offer capabilities like uncertainty estimates and encoding domain knowledge directly into a model. This post provides an overview of Bayesian methods beginning with a review of probability before showing how it can be applied to the coin flipping problem. By the end, the reader will have a basic understanding of the methods and where to go from here.