Slightly belated delivery on Things I Like for June, due to moving house, but better late than never!
This article, while not particularly deep, casts deep neural networks in a Bayesian light. While I think it’s bizarre on its face to claim that deep neural networks are “unprincipled, dumb black boxes that lack elegance” (after all, they’re entirely reducible to extremely concise and elegant expressions in the language of linear algebra), I like the idea of taking both a Bayesian and a non-Bayesian perspective on everything. The article casts Bayesians as possessing the secret weapon of being able to regard non-Bayesian approaches as approximations to more frequentist approaches, but I think the same can be said from the other side. In fact, I have myself described some Bayesian methods as approximations to frequentist methods in the past. In any event, I found a way to regard this article as further confirmation of my belief that the Bayesian-frequentist divide is a lot more fuzzy than it’s often made out to be, and therefore I like it.
8 habits of highly effective data scientists. Normally, I dislike listicles. Also normally, the articles posted on LinkedIn aren’t that great. Anyone who knows me well knows that I’m a big Aristotle fanboy, though, and the fact that this listicle begins with a foundational principle from Aristotle made it catch my eye. I think I’m doing well on some of these already. I’ve set a personal goal of getting better on optimizing my workspace and toolset.