20–27 Mar 2026
Wild View Resorts
Africa/Gaborone timezone

Self-supervised representations for automated astronomical discoveries

27 Mar 2026, 14:45
15m
Wild View Resorts

Wild View Resorts

Plot 80 President Avenue, Kasane, Botswana
In-person - Talk 9 Machine Learning & techniques Science & Engineering

Speaker

Koketso Mohale (University of the Western Cape)

Description

Telescopes such as the Square Kilometre Array (SKA) and Vera C. Rubin Observatory (LSSRT) will produce more data than astronomers can analyse manually. Machine learning, being data-driven, is increasingly being applied in astronomy. Unsupervised machine learning, in particular, is a powerful approach for finding patterns and anomalies automatically, but struggles with high-dimensional data like images. We explore an approach for reducing the dimensionality of the data, called representation learning. We also present novel model-independent methods for measuring the utility of these representations for supervised learning.

Stream Science or Engineering

Primary author

Koketso Mohale (University of the Western Cape)

Co-author

Prof. Michelle Lochner (University of the Western Cape)

Presentation materials

There are no materials yet.