Date/Time
Date(s) - 02/12/2025
11:45 am - 12:45 pm

Speaker: (University of Waterloo and Canada Cifar AI Chair, Vector Institute)

Location: ITB 201

Title: Manifold learning from a user’s perspective

Abstract: This talk will briefly introduce manifold learning, and how to navigate the landscape of modern dimension reduction algorithms like Isomap, Diffusion Maps, t-SNE and UMAP. The focus will be on principled methods to safely interpret these algorithms’ results in practice.
Then I will extend this paradigm by asking if it is possible, in the case of scientific data where quantitative prior knowledge  is abundant, to explain a data manifold by new coordinates, chosen from a set of scientifically meaningful functions? Finally, if time permits, I will describe some results from my work on learning flexible models for preference data, such as rankings.

Joint work with Wenyu Bo, Yu-chia Chen, Samson Koelle, Alex Kokot, James McQueen, Dominique-Perrault-Joncas, Anne Wagner, Weicheng Wu, Hanyu Zhang.

 

Biography:
Marina Meila is a Professor with the Cheriton School of Computer Science at the University of Waterloo, and an Affiliate Professor of Statistics at the University of Washington.
Dr. Meila is a Faculty of the Vector Institute and a Canada CIFAR Chair in AI. Her long term interest is in statistical learning, particularly the discovery of geometric and combinatorial structure in data,
efficient algorithms, and developing guarantees and quantitative interpretations for unsupervised structure discovery with realistic knowledge about the data generating process.
She has collaborated with scientists in applied inverse problems, materials science and theoretical chemistry.
Dr. Meila holds a MS degree in Electrical Engineering from the Polytechnic Institute of Bucharest, and a PhD in Computer Science and Electrical Engineering from the Massachusetts Institute of Technology.