- April 9th (Thursday)
- Room 1212 (Lecture room 1), National Institute of Informatics
- Professor Xiaoming Yuan
Department of Mathematics, Hong Kong Baptist University
- Accuracy v.s. Implementability in Algorithmic Design
---- An Example of Operator Splitting Methods for Convex Optimization
- Accuracy and implementability are two common yet usually conflicted objectives for developing an efficient algorithm. In this talk, I will focus on the context of convex optimization models with separable structures to show how to make a trade-off between these two objectives for some operator splitting methods originated from the PDE literature (e.g., the Douglas-Rachford and Peaceman-Rachford schemes) The resulting algorithms could be applicable to large-scale dataset; and their efficiency will be demonstrated by some specific applications in statistical learning and image processing (e.g., the LASSO and TV-deblurring models). Some theoretical results such as the convergence rates of these algorithms will also be mentioned briefly.
*Please replace [at] with @.