Suggested Searches

 

AI/ML STIG Lecture Series

Artificial Intelligence and Machine Learning Science and Technology Interest Group (AI/ML STIG)

AI/ML STIG about AI/ML STIG Lecture Series

Location

Virtual

Dates

23 February 2026
4:00pm ET

Community

AI/ML STIG

Type

Seminar

Build Recurrent Neural Networks from first principles and apply them to real-time classification of astronomical transients. Implement vanilla RNNs, LSTMs, and GRUs in PyTorch and train a classifier on supernova light curves.

Topics Covered

  • Motivation: sequential and time-series data in astronomy
  • Vanilla RNNs: architecture, forward pass, and vanishing gradients
  • Gated architectures: LSTM and GRU cells
  • Uni-directional vs. bi-directional RNNs
  • Building an RNN from scratch in PyTorch
  • Application: supernova light curve classification with LSTMs
View/Download Notebook on GitHub
Session Recording

Presentation

From Theory to Astronomical Light Curve Classification [PDF]

Feb 23, 2026

PDF (2.15 MB)

Meeting Connection

Join the Meeting

News Straight to Your Inbox

Subscribe to your community email news list

We will never share your email address.

Privacy Policy

Sign Up
Angled from the upper left corner to the lower right corner is a cone-shaped orange-red cloud known as Herbig-Haro 49/50. This feature takes up about three-fourths of the length of this angle. The upper left end of this feature has a translucent, rounded end. The conical feature widens slightly from the rounded end at the upper right down to the lower right. Along the cone there are additional rounded edges, like edges of a wave, and intricate foamy-like details, as well as a clearer view of the black background of space. In the upper left, overlapping with the rounded end of Herbig-Haro 49/50, is a background spiral galaxy with a concentrated blue center that fades outward to blend with red spiral arms. The background of space is speckled with some white stars and smaller, more numerous, fainter white galaxies throughout.