Hi, I'm a research scientist at Unknot, a quantitative trading fund.
I received my Master's degree at Seoul National University, and was fortunate to be advised by Gunhee Kim.
My academic interest lies in theoretical deep learning, game optimization, and their intersection.
Simply put, I study how and when multiple neural networks would converge to their (local) Nash equilibrium.
(Résumé →)
Education
Publications
-
On Convergence of Lookahead in Smooth Games
Junsoo Ha, Gunhee Kim.
International Conference on Artificial Intelligence and Statistics (AISTATS), 2022.
[PDF]
-
A Neural Dicihlet Process Mixture Model for Task-Free Continual Learning
Soochan Lee, Junsoo Ha, Dongsu Zhang, Gunhee Kim.
International Conference on Learning Representations (ICLR), 2020.
[PDF]
[Code]
-
Harmonizing Maximum Likelihood with GANs for Multimodal Conditional Generation
Soochan Lee, Junsoo Ha, Gunhee Kim.
International Conference on Learning Representations (ICLR), 2019.
[PDF]
[Code]
Professional Experience
I used to be a software engineer back before my academic career.
I still love writing atomic programs with a single responsibility and minimal side-effects.
You can find most of my mantra from The Bitter Lesson, Unix Philosophy, and The Zen of Python.
-
Unknot (Seoul, Korea. Sep 2021 - Now) / Research Scientist
-
Buzzni (Seoul, Korea. Mar 2017 - July 2017) / Backend Software Engineer
-
Geopia (Seoul, Korea. Apr 2015 - Feb 2017) / Full-Stack Software Engineer
-
Nexol System (Seoul, Korea. Jan 2015 - Mar 2015) / Software Engineering Intern
-
LG Electronics (Seoul, Korea. Jul 2014 - Aug 2014) / Software Engineering Intern
Profiles
You can browse my profiles in the following social media:
Email is the best way to contact me: junsoo.ha@vision.snu.ac.kr