Seminars Archive

Squeeze your data! Unsupervised Learning for newbies.

Alejandro Rodriguez Garcia
Department of Mathematics, Informatics and Geosciences (MIGe) of the University of Trieste & International Centre for Theoretical Physics (ICTP)

Abstract
Our society is currently undergoing a significant revolution propelled by the advancement of artificial intelligence (AI), and the field of science is not immune to its influence. As data availability grows exponentially, there's a heightened need for automated analysis using AI tools. However, unlike more common AI applications like image recognition or natural language processing that heavily rely on labeled data, scientific exploration often involves uncovering new phenomena where data lacks predefined labels. So, can we still leverage machine learning to glean insights from such unlabeled data? The answer lies in Unsupervised Machine Learning (UL), a branch of AI that explores the inherent structure within data without explicit guidance. Instead of relying on labeled examples, UL techniques delve into the geometry of the data itself to infer patterns and properties. In this presentation, we'll delve into some popular UL techniques and demonstrate their practical application using real-world examples. By exploring methods such as clustering and dimensionality reduction, attendees will gain an understanding of how UL can be utilized to extract valuable insights from unlabeled scientific data.