Hyeji Kim
Assistant Professor
Electrical and Computer Engineering
University of Texas at Austin
hyeji.kim@austin.utexas.edu
2501 Speedway, EER Building, Room 6.870
Austin, Texas 78712
News
This website will be no longer updated. Please see my new website.
I am thrilled to announce our blog [deepcomm.github.io] where posts on ``Inventing Communication Algorithms via Machine Learning" will be posted.
Together with Stephan Ten Brink and Pramod Viswanath, I organized Deep Learning for Communications I session at Allerton conference, September 2019
I gave an invited talk "Deepcode: Feedback Codes via Deep Learning" at the International Symposium on Information Theory (ISIT), July 2019
Together with Sewoong Oh and Sreeram Kannan, I gave a tutorial "Information Theory and Deep Learning: an Emerging Interface" at the International Symposium on Information Theory (ISIT), June 2018 (Slides , Video recording)
About
I've been with Samsung AI Research Cambridge since 2018. Prior to joining Samsung, I spent two years as a postdoctoral researcher at the University of Illinois at Urbana-Champaign, hosted by Prof. Pramod Viswanath and Prof. Sewoong Oh. I received my PhD from the department of Electrical Engineering at Stanford University in 2016, under the supervision of Prof. Abbas El Gamal.
Research
My research interests are in information theory and machine learning. I am particularly interested in research that creates synergy between these two disciplines.
In applying machine learning to information theory, my mission is to automate the invention of communication algorithms via deep learning.
In applying information theory to machine learning, my mission is to develop new sets of tools to analyze data using novel information theoretic methods.
My research spans both theory and practice. During my PhD, I established several new optimality results on the fundamental limits of communications in networks. More recent research interests also include developing tools and frameworks for designing light-weight high-accuracy neural network models, motivated by practical challenges in the AI industry.