Starting this spring I’ll be working Prof Alexie Leauthaud to use deep learning tools to better identify dwarf galaxies in the Hyper Suprime Cam (HSC) survey.  When that survey’s completed, it’s expected to have >500 TB of imaging data, so our project is going to have to be pretty smart and efficient about how to get the best results in a reasonable amount of time.  Right now the plan is to start with some “simpler” machine learning algorithms (like Random Forests) to filter out the objects that clearly aren’t the right type of galaxies, and then use multi-layered convolutional neural nets for our final classification.

The publically available part of our work can be found here: Not much will happen over the summer, while I’m working in San Francisco, but then I’ll get back into things during the fall and beyond.

I’ll also still be doing my supernovae simulation work with Piero Madau and Mark Krumholz, but it’ll be in parallel with this work.  Things might go a bit slower with those simulations, but we haven’t completely forgotten them!