ZTF Summer school 2023: Machine Learning for Time-Domain Astronomy

ZTF Homepage: https://ztf.caltech.edu

In this notebook, we'll take a look at a state-of-the-art machine learning model for time-domain astronomy: BTSbot. Every night, BTSbot runs on every single alert from the Zwicky Transient Facility (images that differed from reference images at a certain position, creating an alert packet sent to the community). It's a deep learning multi-modal model that takes in images (new iamge, reference image, and the subtraction of the two) and metadata (like position, brightness, and other features including some generated by other specialized models!). BTSbot is trained to identify alerts that are of interest to the Bright Transient Survey, which is a survey that is essentially looking for bright transients (like supernovae, kilonovae, and other interesting things that get very bright very quickly).

We'll go through the following steps:

1. Download the dataset
2. Learn about the model and its architecture
3. Build the model
4. Train the model
5. Evaluate the model
6. Hyperparameter tuning, data augmentation, ... anything to get the best out of the model!

We encourage you to make modifications and to try different things. This is a great opportunity to learn about machine learning and time-domain astronomy, and we hope you'll take advantage of it.

Let's get started!