[2018-09-28] Prof. Yen-Huan Li,NTU, “Towards efficient optimization with the logarithmic loss”

Poster:Post date:2018-09-25

Title: Towards efficient optimization with the logarithmic loss
Date: 2018-09-28 14:20-15:30
Location: R103, CSIE
Speaker: Prof. Yen-Huan Li,NTU
Hosted by: Prof. Winston Hsu




The logarithmic loss violates standard conditions---e.g., bounded gradient, bounded Hessian, etc.---in optimization theory, rendering convergence proofs of most existing optimization methods invalid. However, the logarithmic loss appears in several important applications in various fields, such as positron emission tomography (PET), universal portfolio selection (UPS), and quantum state estimation (QSE), and is key to the interior point method for linear programming. In this talk, I will review challenges posed by the non-standard nature of the logarithmic loss, discuss existing achievements regarding addressing the challenges, and present some partial solutions. 




Yen-Huan Li received the B.S.E. degree in electrical engineering in 2008, and the M.S. degree in communication engineering in 2010, both from National Taiwan University. He received the PhD degree in computer science from École Polytechnique Fédérale de Lausanne (EPFL), Switzerland in 2018. From 2011 to 2012, he was a research assistant at Academia Sinica. He is currently an assistant professor at the Department of Computer Science and Information Engineering, National Taiwan University. His research interests include machine learning, convex optimization, high-dimensional statistics, and quantum information.



Last modification time:2018-09-25 AM 11:31

cron web_use_log