Accelerating Nonnegative-X by extrapolation, where X ∈ {Least Square, Matrix Factorization, Tensor Factorization}
01 Apr, 2019 (Mon)
3:00 pm - 4:00 pm
Room 603, Chow Yei Ching Building

Add to your calendar: iCal, Google Calendar

HKU Campus Map


Mr. Andersen Ang
Mathématique et Recherche opérationnelle
Université de Mons, Belgium


Non-negativity constraint is ubiquitous in machine learning, data mining and signal processing problems. The family of non-negativity constrained problems (NCP) includes non-negative least squares (NNLS), non-negative matrix factorization (NMF) and non-negative tensor factorization (NTF). The goal of these NCPs in general is to fit an given object (vector, matrix, or tensor) by solving an optimization problem with non-negativity constraints. Many algorithms exist to solve the NCPs such as the multiplicative updates (MU), the projected gradient descent (PGD), the alternating non-negative least squares (ANLS), and the hierarchical alternating least squares (HALS). In this talk, we introduce an general extrapolation-based acceleration scheme that can significantly accelerate the algorithms for NNLS, NMF and NTF.  The scheme is in the form of Xk+1 = Xk + βk(Xk – Xk-1) where k is the iteration counter and βis the  extrapolation parameter. When the NCP is a NNLS problem, the scheme reduces to a Nesterov’s style acceleration. When the NCP is a NMF or NTF problem, the scheme is shown to be able to be faster than accelerated method from the literature, such as the extrapolated inexact block coordinate descent method.


Biography of the speaker:
Andersen (Man Shun) Ang is a PhD candidate at the  Mathématique et Recherche opérationnelle, Faculté polytechnique, UMONS, Belgium. Before his pursuit of a PhD degree, he received a Bachelor of Engineering (B.Eng.) in electronics and communication engineering in 2014 and a Master of Philosophy (M.Phil.) in biomedical engineering in 2016, both from the University of Hong Kong, Hong Kong. In 2017, he received the European Research Council (ERC) Scholarship. 

His current research interests are Matrix-Tensor Factorization, continuous optimization and convex geometry.



Prof. R.Y.K. Kwok

Most seminars are open to the general public, free of charge, unless otherwise stated. Registration is not required. Arrangement for car parking facilities on campus please contact us for details.

For enquiries, please contact:
Department of Electrical and Electronic Engineering,
Room 601, Chow Yei Ching Building,
Pokfulam Road, Hong Kong
Tel: (852) 3917 7093