Deep learning is a buzz phrase that refers to a branches of machine learning that involve multiple, interactive layers of nonlinear data transformations. It’s deep because there are multiple layers, and it’s deep learning because the layers can communicate with one another. Deep learning techniques have been growing in accuracy and success and are interesting to study in themselves.
As a data scientist, I’m drawn to deep learning because it involves some of my favorite things: multidimensional math, linear algebra in particular; cool algorithms; the chain rule; and surprises – which in the applied math world means “nonlinear effects”.
So my goal is to document adventures in deep learning in the next months, and make this blog a place where interested people can learn along with me.
The problem: Each cell of an 8×8 chessboard is filled with either a 0 or a 1. Prove that if we compute the sums of the numbers in each row, each column, and each of the two diagonals, at least … Continue reading →
In the 80s, in June, when I was in graduate school, a friend called me up and asked if I wanted to make an impromptu trip from Berkeley, California to Vancouver, BC. He wanted to drive up and see his … Continue reading →
Once upon a time in math, I studied metric spaces, spaces for which there exists a way to measure distances between points. At first glance the non-mathematician might say: yuck, this is the genesis and justification of corporate bean counting … Continue reading →
Yep, I like math problems. But just like anyone, I feel frustrated and stupid when I don’t get them. The thing is, the harder the problem, the more satisfying it is to solve it. Here’s a pretty easy one I … Continue reading →