Machine learning is a vast field advancing at an alarming rate. I’ll be using this site to highlight some of the resources I found useful in my struggle to keep up and share some work I’ve done to gain a deeper understanding of this fascinating topic.
Gibbs sampling for the uninitiated
The first resource I’d like to highlight is the tutorial paper, Gibbs Sampling for the Uninitiated, by Resnik and Hardisty, which is a masterpiece of exposition. Their main example provides an amazingly clear description of how to build a Gibbs sampler for the very simple Naı̈ve Bayes probabilistic model. I’ll share an implementation of Gibbs sampling in Python and an illustration of how to detect spam with a the sampler. Eventually, I’ll demonstrate the power of Cython by optimizing the sampler to gain a couple of orders of magnitude in performance.
“Let us calculate!”
The title of this site is taken from Leibniz’s The Art of Discovery (1685). This represents for me his vision that the only way to avoid controversy is to make our reasoning as clear as mathematics so that “… we can find our error at a glance, and when there are disputes among persons, we can simply say: Let us calculate [calculemus], without further ado, to see who is right.”