AQFC2015

Optimistic Quadratic Discriminant Analysis Using (Geodesically) Convex Optimization

----------------------------------------------------------------------------------------------------
 
 
             Department of Systems Engineering and Engineering Management
                     The Chinese University of Hong Kong
 
 
----------------------------------------------------------------------------------------------------
 
Date: Friday, November 13, 2020, 16:30 to 17:30
 
Title: Optimistic Quadratic Discriminant Analysis Using (Geodesically) Convex Optimization  
 
Speaker: Prof. Man-Chung Yue
 
Abstract:
 
When evaluating the likelihood of an observation, the nominal distribution for the observation is estimated from data, which makes it susceptible to estimation errors. To alleviate this issue, we propose to replace the nominal distribution with an ambiguity set containing all distributions sufficiently close to the nominal distribution. When this proximity is measured by the Fisher-Rao distance or the KL-divergence, the emerging optimistic likelihood can be calculated efficiently using geodesically or standard convex optimization. We showcase the advantages of our optimistic likelihoods on a classification problem using artificially generated as well as standard benchmark instances. 
 
Biography:
 
Dr. Yue received his B. Sc. degree (2008-2012) in Mathematics from CUHK. He then obtained his Ph.D. degree (2013 - 2017) in Systems Engineering and Engineering Management from CUHK. Currently he is an Assistant Professor at Department of Applied Mathematics, The Hong Kong Polytechnic University. Before taking up his current position, he was an Research Associate (2017 - 2019) at Imperial College Business School, Imperial College London. Dr. Yue's primary research area is optimization, with focuses on the following topics: Large-Scale Algorithms, Non-Convex Optimization, Newton-Type Methods, Robust Optimization.
 
Date: 
Friday, November 13, 2020 - 16:30 to 17:30