Difference between revisions of "SOCR JMM 2026"
m (→Talk: Kime-Phase Analytics: A Mathematical Framework for Complex-Time Representation of Longitudinal Processes) |
(→Program) |
||
| Line 25: | Line 25: | ||
| − | == Program== | + | == [https://meetings.ams.org/math/jmm2026/meetingapp.cgi/Session/13818 Session Program]== |
| − | |||
| + | : '''Location''': Room 204C (Level 2, Walter E. Washington Convention Center) | ||
| + | |||
| + | : '''Session''': [https://meetings.ams.org/math/jmm2026/meetingapp.cgi/Session/13818 AMS Special Session on Mathematical Foundation of Machine Learning, I] | ||
| + | |||
| + | * 8:00 AM: Riemannian Optimization on Manifolds of Low-Rank Tensors, Maryam Bagherian, Idaho State University, POCATELLO, ID | ||
| + | |||
| + | |||
| + | * 8:30 AM: Learning on manifolds without manifold learning, Ryan Michael O'Dowd, claremont Graduate University, Erie, CO and Hrushikesh Mhaskar, claremont Graduate University | ||
| + | |||
| + | |||
| + | * 9:00 AM: Continuous Symmetry Discovery and Enforcement using the Lie Derivative, Benjamin Shaw, Utah State University | ||
| + | |||
| + | * 9:30 AM: Uncovering Latent Structure in Neural Networks through Local CorEx, Thomas Jordan Kerby, Brigham Young University and Kevin Moon, Utah State University, Providence, UT | ||
| + | |||
| + | * '''10:00 AM: Kime-Phase Analytics: A Mathematical Framework for Complex-Time Representation of Longitudinal Processes, Ivo D. Dinov, University of Michigan, Ann Arbor, MI, Yueyang Shen, University of Michigan and Bojko N Bakalov, North Carolina State University''' | ||
| + | |||
| + | * 10:30 AM: Tensor denoising, Harm Derksen, Northeastern University, Boston, MA | ||
| + | |||
| + | * 11:00 AM: An adaptive framework for first order gradient methods with momentum, Yunrong Zhu, Idaho State University, Xiaozhe Hu, Tufts University, Sara Pollock, University of Florida, GAINESVILLE, FL and Zhongqin Xue, Tufts University, Medford, MA | ||
| + | |||
| + | * 11:30 AM: Categorical Foundations of Distributed Optimization and Learning, Tyler Evan Hanks, University of Florida, Matthew Klawonn, Air Force Research Lab, Evan Patterson, Topos Institute, Matthew Hale, Georgia Institute of Technology and James P Fairbanks, University of Florida, Gainesville, FL | ||
==Talk: [https://meetings.ams.org/math/jmm2026/meetingapp.cgi/Paper/51462 Kime-Phase Analytics: A Mathematical Framework for Complex-Time Representation of Longitudinal Processes]== | ==Talk: [https://meetings.ams.org/math/jmm2026/meetingapp.cgi/Paper/51462 Kime-Phase Analytics: A Mathematical Framework for Complex-Time Representation of Longitudinal Processes]== | ||
Revision as of 18:16, 30 December 2025
Contents
SOCR News & Events: 2026 JMM/AMS Special Session on Mathematical Foundation of Machine Learning
Session Overview
- 2026 AMS/JMM: Annual 2026 JMM Congress
- Date: Tuesday, January 6, 2026, 8:00-12:00 PM ET
- Tuesday 01/06/2026, 10:00 - 10:30 AM
- Reference ID: 51462, Title of Paper: Analytics: A Mathematical Framework for Complex-Time Representation of Longitudinal Processes
- Location: Room 204C, Walter E. Washington Convention Center, 801 Allen Y. Lew Place NW, Washington, DC 20001
- Abstract: This special session focuses on the rigorous mathematical foundations underlying modern machine learning. Topics include, but are not limited to, operator theory, functional analysis, optimization, linear/multilinear algebra, metric learning, and approximation theory of neural networks. We welcome contributions that deepen understanding of data-driven algorithms through fundamental mathematical inquiry, emphasizing theoretical rigor in exploring the principles driving machine learning.
Abstract Submission
- Abstract Submission: Abstracts must be submitted through the AMS portal.
- Submission Period: July 10, 2025 -- Tuesday, September 9, 2025
Session Program
- Location: Room 204C (Level 2, Walter E. Washington Convention Center)
- 8:00 AM: Riemannian Optimization on Manifolds of Low-Rank Tensors, Maryam Bagherian, Idaho State University, POCATELLO, ID
- 8:30 AM: Learning on manifolds without manifold learning, Ryan Michael O'Dowd, claremont Graduate University, Erie, CO and Hrushikesh Mhaskar, claremont Graduate University
- 9:00 AM: Continuous Symmetry Discovery and Enforcement using the Lie Derivative, Benjamin Shaw, Utah State University
- 9:30 AM: Uncovering Latent Structure in Neural Networks through Local CorEx, Thomas Jordan Kerby, Brigham Young University and Kevin Moon, Utah State University, Providence, UT
- 10:00 AM: Kime-Phase Analytics: A Mathematical Framework for Complex-Time Representation of Longitudinal Processes, Ivo D. Dinov, University of Michigan, Ann Arbor, MI, Yueyang Shen, University of Michigan and Bojko N Bakalov, North Carolina State University
- 10:30 AM: Tensor denoising, Harm Derksen, Northeastern University, Boston, MA
- 11:00 AM: An adaptive framework for first order gradient methods with momentum, Yunrong Zhu, Idaho State University, Xiaozhe Hu, Tufts University, Sara Pollock, University of Florida, GAINESVILLE, FL and Zhongqin Xue, Tufts University, Medford, MA
- 11:30 AM: Categorical Foundations of Distributed Optimization and Learning, Tyler Evan Hanks, University of Florida, Matthew Klawonn, Air Force Research Lab, Evan Patterson, Topos Institute, Matthew Hale, Georgia Institute of Technology and James P Fairbanks, University of Florida, Gainesville, FL
Talk: Kime-Phase Analytics: A Mathematical Framework for Complex-Time Representation of Longitudinal Processes
- Authors: Ivo D. Dinov (UMich), Yueyang Shen (UMich), and Bojko N Bakalov (NCSU), slidedeck
- Abstract: This talk will present a complex-time (kime) representation framework for modeling repeated measurement longitudinal processes. The induced kime-phase analytics (KPA) offer a mathematical-statistics foundation for developing advanced machine learning and artificial intelligence models of time-varying functional data. By jointly tracking the classical time dynamics and the intrinsic cross-sectional variability of the underlying process, KPA represents temporal data as rich tensor objects, kime-surfaces. These 2D manifolds are parameterized by a complex variable \(\kappa =t e^{i\theta}\), where the kime magnitude \(t=|\kappa |\in \mathbb{R}^+\) is the longitudinal event order (classical time), and the kime phase \(\theta \sim \Phi_{S^1}\) captures the intrinsic stochastic variation of the longitudinal process. Inspired by quantum tomography and grounded in differential geometry, KPA enables reconstruction of latent phase distributions from observable data. As time permits, we will discuss open problems and show biomedical applications.
Translate this page:
