Here are some fun projects I’ve done for classes or just to help me understand machine learning and data mining techniques!
Visualizing star spots
These were created for a lunch talk I gave at the PSU astronomy department showing the difference between the dataset used by a previous work versus the dataset I was creating and using for my radial velocity work.
Github here.
Spherical clustering
Github here.
Gaussian processes regression
In general, GPs are a flexible scheme for modeling and performing Bayesian inference on stochastic (i.e. random) processes of where we don’t know their functional forms. GP regression allows us to model stochastic processes by describing the covariance between datapoints with a kernel function instead of the data themselves. GPs are cool because they are really good at fitting extremely weird functions, lend themselves towards robust error propagation, and can have kernel functions that tell us about the underlying process being fitted.
I created a nice Jupyter notebook walking through the basics of creating and drawing GPs for fitting functions of unknown functional form here.
Here are some figures!
This is what finding the best hyperparameters on real data can look like