DESCRIPTION
Learning is fundamentally rooted in data representation. Starting with a Harmonic Analytic perspective where canonical bases (Fourier, wavelets and others) are used to represent data to create inference and decision making algorithms whose optimality is typically a result of an assumed associated statistical model, one can proceed to a more data-driven alternative including Principal Component Analysis and Manifold Learning when the reduced dimension of the information space is adopted in comparison to the ambient space. These insights turn out to afford sufficient insight to develop a so-called a Union of Subspace (or subspace clustering) which in turn, are a bridge to Deep Structure learning and Deep learning in all its variations including Transformers. All these techniques will be presented in a unified framework where the evolution is gradual and analytically tractable.
DETAILS
Course type: Tutorial (in person delivery)
Duration: 3-4 hours
Level: Postgraduate
Institution of lecturer: North Carolina State University
Notes: Creatively apply the methodologies on proposed application problems, rather than exams.
LECTURER
Prof. Hamid Krim
Hamid Krim (ahk@ncsu.edu) received his BSc. MSc. And Ph.D. degrees in EE. He was a Member of Technical Staff at AT&T Bell Labs, where he has conducted research and development in the areas of telephony and digital communication systems/subsystems. Following an NSF postdoctoral fellowship at Foreign Centers of Excellence, LSS/University of Orsay, Paris, France, he joined the Laboratory for Information and Decision Systems, Massachusetts Institute of Technology, Cambridge, MA as a Research Scientist and where he was performing and supervising research. He is presently Professor of Electrical Engineering in the ECE Department, North Carolina State University, Raleigh, leading the Vision, Information and Statistical Signal Theories and Applications group, and has recently held a rotation as an IPA at Army Research Office in the Research Triangle Park, NC. His research interests are in statistical signal and image analysis and mathematical modeling, Machine Learning and AI problems, with a keen emphasis on applied problems in classification and recognition using geometric and topological tools. He was recently awarded (together with M. Viberg) the Best Sustained Impact Paper Award by the IEEE Signal Processing Society for the paper which appeared over 25 years ago in the IEEE Signal Processing Magazine.