The Lp-PC of a Matrix, for p<1
Lp-norm Principal-Component Analysis (Lp-PCA) has been solved exactly for p=2 (standard PCA) and, more recently, p=1 (L1-PCA). For general values of p, however, the exact solution remains unknown.
In this work, we show for the first time that, for p<1, the Lp-PC of a matrix can be found exactly through combinatorial optimization.
About the Speaker
Dimitris Chachlakis earned his Diploma in Electronic and Computer Engineering from the Technical University of Crete, Greece, in 2016.
In Fall 2016, he joined the Department of Electrical and Microelectronic Engineering at Rochester Institute of Technology, Rochester, New York, as a PhD student with the Machine Learning Optimization and Signal Processing (MILOS) research group.
His research interests are in the areas of robust multimodal (tensor) data analytics, signal processing, machine learning, and convex/non-convex optimization.
He is a graduate student member of the IEEE Signal Processing Society and the Society for Industrial and Applied Mathematics (SIAM). He has served as a Reviewer to the IEEE TRANSACTIONS ON SIGNAL PROCESSING, IEEE SIGNAL PROCESSING LETTERS, IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, IEEE ACCESS, IEEE PHOTONICS JOURNAL, ELSEVIER DIGITAL SIGNAL PROCESSING, and many IEEE conferences. He was a recipient of the 2018 Gerondelis Foundation Graduate Student Scholarship Award and the 2019 A. G. Leventis Foundation Graduate Student Scholarship Award.