A few years ago, I decided I needed to learn how to code simple machine learning algorithms. I had been writing about machine learning as a journalist, and I wanted to understand the nuts and bolts. (My background as a software engineer came in handy.
) One of my first projects was to build a rudimentary neural network to try to do what astronomer and mathematician Johannes Kepler did in the early 1600s: analyze data collected by Danish astronomer about the positions of Mars to come up with the laws of planetary motion. I quickly discovered that an —a type of machine learning algorithm that uses networks of computational units called artificial neurons—would require far more data than was available to Kepler. To satisfy the algorithm’s hunger, I generated a decade worth of data about the daily positions of planets using a simple simulation of the solar system.
After many false starts and dead-ends, I coded a neural network that—given the simulated data—could predict future positions of planets. It was beautiful to observe. The network indeed learned the patterns in the data and could prognosticate about, say, where Mars might be in five years.
I was instantly hooked. Sure, Kepler did much, much more with much less—he came up with overarching laws that could be codified in the symbolic language of math. My neural network simply took in data about prior positions of planets and spit out data about their future positions.
It was a black box, its inner workings undecip.