I’m excited to say that I’ll be teaching an impromptu course this spring on “Astrophysical Machine Learning”. It’s impromptu because I didn’t expect to teach it, it just worked out that there were a lot of students in my engineering physics course this Fall that got interested when I showed my work in class. Right now there are only about six students enrolled, with several more sitting in. I’m putting the course webpage online here, and the students will be sharing there work (with each other at first) on Github. I haven’t used Github for any of my work yet, so I’m excited to learn as we progress.
A side benefit for me (and the students) is that we’ll be quickly breaking into groups and working on real research problems, many of them centered around ML applications for source-finding and classification of radio sources. I’m the chair of the “Cosmic-Web” Key Science Project for the Evolutionary Map of the Universe survey to be conducted with the ASKAP telescope. Of particular interest to me are source-finding algorithms for diffuse sources (see below), were it is often difficult to find and characterize them when there are imbedded compact sources. Below is an example of a diffuse source (a simple cluster radio halo) with background point-sources imbedded within, taken from a simulation of what the EMU survey will be capable of. Early science for ASKAP is happening right now so the time is right to test some of this out on real data!