Matrix Profiles for Time Series Data Mining

A highly efficient time series data mining algorithm that is simple, space efficient, has time complexity that is constant in subsequence length, is robust with regard to missing data, and runs in deterministic time. Pretty neat concepts, the math is simple enough to wrap your head around, and it can be implemented in python. Not only that, but it parallelization comes free in the box.

It is also worth noting that the principle can be applied to almost any linear map, not only time series.

https://www.youtube.com/watch?v=1ZHW977t070