Resources

Tutorials and lectures in astronomical ML

2022 Astro Hack Week

I presented a two-part course on astronomical machine learning during the 2022 Astro Hack Week. There are two Jupyter notebooks with examples and practice problems shown here. The first notebook provides an introduction to machine learning using tabular data. The second notebook presents convolutional neural networks applied to astronomical image cutouts.

2023 KITP Program

I helped coordinate a KITP program on Data-Driven Astronomy (galevo23), which featured some very nice tutorials and talks. We covered topics like simulation-based inference, GNNs, symbolic regression, probabilistic U-nets, and much more. All machine learning tutorials can be accessed on Github.

2023 LSSTC Data Science Fellowship Program

I was a guest lecturer for the [19th Session of the Data Science Fellowship Program (DSFP)]. My first lecture focused on convolutional neural networks. My second lecture introduced graph neural networks and their applications to galaxies, dark matter halos, and large scale structure in cosmological simulations.

Hybrid CNNs with deconvolution layers

In order to predict galaxy spectra from images, I created a CNN with hybrid normalization layers. In the NeurIPS workshop paper, we found that a combination of deconvolution layers and batch normalization can greatly improve results for CNNs trained on astronomical images. Pytorch code for this hybrid CNN can be found on my Github page.

Blog

I sporadically post in my research blog, which showcases some machine learning applications in astronomy.