Gaussian Process Modelling is all very well, but when you've got lots of data points - and potentially lots of hyper-parameters - optimizing the values of those hyper-parameters can be very computationally expensive, i.e. slow... In my previous post I was using GPM to extract the rotation period of stars from Kepler data. For that …
Spinning Stars
Stars spin - and the rate at which they spin has a profound affect on their health. So measuring the rotation of different types of stars is an important pass-time for astronomers who like that kind of thing (i.e. stars). Fortunately stellar rotation is something that can even be measured directly from their time dependent …
Predicting the Future
One of the most well known examples of using Gaussian Process Modelling for forward prediction is the application described in Rasmussen & Williams, which shows the prediction for the future of atmospheric CO2 levels. When the book was written, the prediction showed the increase in CO2 concentration continuing at roughly the same rate. However, things …
Gaussian Process Modelling in Python
Non-linear regression is pretty central to a lot of machine learning applications. However, when you don't know enough/anything about the actual physical parametric dependencies of a function it can be a bit of a show-stopper. But... what if you could predict the value of a function at any point based only on its value at …
Gaussian Processes in Python
I'm guessing that most people are pretty comfortable with the concept of uncorrelated Gaussian noise. It's the most frequently assumed noise. Even if you don't realise it, you're probably assuming Gaussian noise. Quick check: Are you using a chi-squared test to fit your data? Yes? Well there you go. Co-variate Gaussian Noise Here I'm going …
ALL YOUR BASE ARE BELONG TO US 