Known by many names, sketching techniques allow random projections of data from high to low dimensions while preserving useful properties. This talks explores ways to use sketching, as well as other techniques and tools from random matrix theory and conic integral geometry, so as to improve and analyse the scalability of algorithms for diverse classes of optimization problems and applications, from linear to nonlinear, local to global, derivative-based to derivative-free. Numerical illustrations/results will also be presented.
Bio: Dr Coralia Cartis is Associate Professor in Numerical Optimization at the Mathematical Institute, University of Oxford since 2013, and a Turing fellow at the Alan Turing Institute for Data Science since 2016; previously, she held academic and research positions at University of Edinburgh and Rutherford Appleton Laboratory. She holds a PhD degree in Mathematics, University of Cambridge (supervisor: Prof Mike Powell) and a BSc in Mathematics from Babesh-Bolyai University, Cluj-Napoca, Romania. Her research interests are in the development and analysis of nonlinear optimisation algorithms, with particular emphasis on complexity/global rates of convergence; and in diverse applications of optimisation from climate modelling to signal processing and machine learning.