ART-AI and University of Bath AI group seminar
We are pleased to have Coralia Cartis, a Professor of Numerical Optimization at the Mathematical Institute, University of Oxford and a Fellow of Balliol College and The Alan Turing Institute for Data Science, join us for this joint ART-AI and University of Bath AI group seminar, entitled ‘Challenges and improvements in optimization algorithms for machine learning’.
This seminar will take place online on Zoom on Tuesday 14th June 2022, 12.15pm-1.15pm (BST). Mohammad Golbabaee will be chairing this seminar. For joining instructions, please e-mail [email protected].
Abstract
We will discuss some key challenges to optimization algorithm development arising from machine learning. In particular, we investigate dimensionality reduction techniques in the variable/parameter domain for local and global optimization; these rely crucially on random projections and allow biased noise, adaptive parameters and have almost sure convergence. We describe and use sketching results that allow efficient projections to low dimensions while preserving using properties, as well as other useful tools from random matrix theory and conic integral geometry. We focus on functions with low effective dimensionality – that are conjectured to provide an insightful proxy for neural networks landscapes. Time permitting we also briefly discuss first- versus second-order optimization methods for training deep neural networks.
Bio
Coralia Cartis is Professor of Numerical Optimization at the Mathematical Institute, University of Oxford and a Fellow of Balliol College and The Alan Turing Institute for Data Science. She received a BSc degree in mathematics from Babesh-Bolyai University, Cluj-Napoca, Romania, and a PhD degree in mathematics from the University of Cambridge, under the supervision of Prof Michael J.D. Powell. Prior to her current post, she worked as a research scientist at Rutherford Appleton Laboratory and as a postdoctoral researcher at Oxford University. Between 2007-2013, she was a (permanent, senior) lecturer in the School of Mathematics, University of Edinburgh. Her research interests include the development and analysis of nonlinear optimization algorithms and diverse applications of optimization from climate modeling to signal processing and machine learning.
Main Photo: samantha-borges-ees69ttpq18-unsplash-2.jpg
