Using Machine Learning to Find Changes

The algarithm looks for a large difference between how similar all points are in the series. And so as it goes through, it'll find the point, the pivot, if you will, the split between left and right,. That gives the largest difference, and i'll label that as the most significant change point. So its pretty fast. At some point the n squared will be a problem, and it'll be trivial to take care of because we'll just truncate the data at that point. If it can handle a year's worth of data already, i really don't need two years, you know. I'm always kind of reminded of andrew ing's one second rule