I really like Sankey Diagrams for visualizing the performance of classification models. I couldn’t find a library to make these plots look the way I wanted them to so I wrote some python code to do the job with matplotlib a couple of years ago. It sat forgotten for a long time, but I recently remembered its existence while working on a paper about classifying stars, galaxies, and AGN.

Galaxies hang out in groups on top of dark matter halos that we can’t see. Figuring out where one group ends and another begins is thus a tricky but critical first step in figuring out how galaxies are affected by their environments. I’ve been playing with interactive plots generated with Plotly to better understand the output of my group-finding efforts.

Libeskind et al. (2016) find that when you look around a close-ish pair of galaxies, you’ll find that their satellites are clumped in the space between them. I’d like to determine how large a sample I’ll need to observe this effect in another dataset and figure out how good photoZs need to be to do this with a sample of pairs selected without specZs.