Science Examples#
These examples show how Hyrax can be used in realistic astronomy research settings, where the goal is not just to run a model, but to inspect latent spaces, identify unusual objects, and connect machine learning outputs back to scientific interpretation.
Tip
If you are new to Hyrax, we recommend you go to the Getting Started section before jumping into the science examples.
HSC Unsupervised Galaxies
An end-to-end unsupervised discovery workflow on HSC galaxy cutouts. This notebook trains an autoencoder, builds a UMAP view of the latent space, explores that space interactively, runs nearest-neighbour search through a vector database, and uses distance in latent space to surface unusual objects for follow-up.
PLAsTiCC Supervised Transients
A supervised classification workflow on multi-band light curves from the PLAsTiCC dataset. This notebook defines a custom dataset and a custom 1-D CNN model, trains the classifier on 14 classes of astronomical transients and variables, and evaluates performance with a confusion matrix.
Important
This section will continue to grow. Over time, we plan to add additional science examples covering different data types, surveys, and models. If you have an ML-oriented use case in mind that is not covered here, chances are Hyrax can support it. If you are not sure how to approach it, get in touch with us. We are here to help.