Math for ML and ML for Math: AltGDmin for Secure Federated LR Learning and MRI & K–12 Math Support скачать в хорошем качестве

Math for ML and ML for Math: AltGDmin for Secure Federated LR Learning and MRI & K–12 Math Support

Title: Mathematical ML and ML for Math: Alternating GD and Minimization (AltGDmin) for Secure Federated Low Rank Matrix Learning for Real-time MRI and ML-enabled K-12 Math Support Speaker: Prof. Namrata Vaswani, Endowed Anderlik Professor, Iowa State University Time: 5:30 PM - 6:30 PM (IST) Date: 23 February 2026 Venue: GJ Hall and Online on Zoom Abstract: This talk will consist of two parts. In the first 40 minutes, I will describe my group’s (mathematical ML) research on the AltGDmin algorithm and its Byzantine-resilient distributed extension. The last 15 minutes will be “ML for Math” where I will describe our CyMath program’s ML-Enabled K-12 Math Tutoring and Support. https://cymath.iastate.edu/ Math for ML: Modern distributed and federated learning systems are vulnerable to various kinds of adversarial attacks. Byzantine attacks are one of the most difficult attacks to deal with, since these are model update poisoning attacks (poison algorithm iterates of the attacked nodes), the adversarial nodes are omniscient, and these nodes can collude. We introduce provably Byzantine-resilient algorithms for solving three different vertically federated learning low-rank (LR) matrix learning problems – LR Matrix Completion, LR Column-wise Sensing, and LR Phase Retrieval – all of which involve solving a partly-decoupled optimization problem, and all involve dealing with data heterogeneity across nodes. These problems find important applications in recommender system design, multi-task representation learning for few-shot learning, federated sketching, accelerated dynamic MRI, and Fourier ptychography. Our proposed algorithms, Byz-AltGDmin, are provably Byzantine-resilient modifications of Alternating GD and minimization (AltGDmin). AltGDmin, introduced in our recent work, is a novel faster, and more communication-efficient, alternative to Alternating Minimization (AltMin) for partly-decoupled optimization problems. These are problems in which the set of optimization variables can be split into two subsets such that the optimization with respect to at least one subset, keeping the other fixed, is decoupled. If time permits, we may also show real-data experimental results on the advantage (speed and generality) of AltGDmin-based methods over the existing state-of-the-art within dynamic MRI. ML for (Cy)Math: Math learning is cumulative; arithmetic fluency is critical for even having the ability to understand basic scalar algebra (solving linear equations in one and two variables for example); this in turn is critical for linear algebra and all SPML and STEM. Fixing the early math skills of students is critical for the future of statistics and all STEM professions. We discuss ways in which STEM students and professionals can help – university or IEEE supported math tutoring for school students, encouraging math practice at home, raising awareness of the need/resources for early math skills, and advocating for better K-12 math teaching policies that re-introduce homework in at least middle school if not earlier. Use of an ML-enabled math learning application (we use ALEKS and Khan Academy for example) by a human makes these tasks easier and less reliant on high quality tutors or math-knowing parents, making our approach scalable and more equitable. Our CyMath https://cymath.iastate.edu/ program uses this. We also argue that some current K-12 policies, based on short-term research, should be critically re-examined from a long-term college STEM student success perspective. For instance removing homework from even older elementary or middle school may have unobservable impact in a 2-3 year study, but no one has done a 10-12 year study of its impact on math skills equity, or earning capacity equity, or workforce success. Reasons are of course it is not easy to do such studies in a statistically sound fashion. However, without long-term impact studies, we need to start relying more on the intuition of college and high school Math and STEM educators Bio: Namrata Vaswani received her Ph.D. from the University of Maryland, College Park, in 2004 and her B.Tech from IIT Delhi in 1999. She has been with Iowa State University since Fall 2005 and is currently the Anderlik Professor of Electrical and Computer Engineering. Her research focuses on data science, particularly Statistical Machine Learning and Signal Processing. She directs the CyMath K–12 Math Tutoring and Support program (https://cymath.iastate.edu/ ). She has served as Associate, Area, or Guest Editor for leading IEEE journals and served on the SPS Board of Governors, including as Chair of Women in Signal Processing. She is a recipient of multiple research awards and is an IEEE Fellow (2019) and AAAS Fellow (2023). ALL ARE WELCOME.