Type Slowly

Jruby on Elephants

A few months ago, I gave a talk at the excellent Ruby Manor conference about using Jruby with Mahout to do machine-learning (More specifically, to detect twitter messages containing pictures of kittens). There’s now a video of it online, as well as slides and code.

One small additional note - my explanation of Logistic Regression is a little confused, and is subtly but fundamentally incorrect. In the simple one dimensional example I show, the Y-axis corresponds to the feature to be learned, while the X-axis corresponds to the feature we’re learning from - Watching the video back, It looks like I’m describing the Y-axis as a feature,which would be silly, as then we wouldn’t actually be predicting anything. So, the logistic curve, in this 1D example gives us a means of predicting the classification (Y-high or Y-low) from the (real) value of the X axis. Sorry!