Some Matrix Aspects of Memoryless Nonlinear Programming Algorithms
Saman Babaie-Kafaki - Professor, Semnan University
Sun, 23-May-2021 / 18:00 / Link:
Video Slides Poster
Nonlinear programming (NLP) appears in many practical applications such as machine learning, image processing, compressed sensing, curve fitting and neural network training. As known, matrix analysis plays significant roles for improving theoretical features as well as computational performance of the NLP algorithms. Here, we discuss effects of the matrix factors (such as condition number and maximum magnification) on convergence and stability of the memoryless NLP algorithms which are efficient tools for solving large-scale optimization problems. Especially, importance of simple as well as memoryless matrix approximations is discussed. Finally, several recent regression models are introduced for which such matrix approximations can reasonably enhance quality of the estimation.
Saman Babaie-Kafaki is a Professor of Faculty of Mathematics, Statistics and Computer Science of Semnan University (Semnan, Iran). He received his B.Sc. in Applied Mathematics from Mazandaran University, Iran, in 2003, and his M.Sc. and Ph.D. in Applied Mathematics from Sharif University of Technology (Tehran, Iran) in 2005 and 2010, respectively, under supervision of Professor Nezam Mahdavi-Amiri. He also worked as a nonresident researcher of Institute for Research in Fundamental Sciences (IPM) (Tehran, Iran) from 2011 to 2015. His research interests lie within numerical continuous optimization, matrix computations, linear regression, heuristic algorithms and image processing.