|
Hyunseung Kim I am a 3rd year Ph.D. student at Data and Visual Analytics Lab (DAVIAN-LAB), advised by Prof. Jaegul Choo at Korea Advanced Institute of Science and Technology (KAIST). I received my B.S. in Computer Science at Korea University in 2021. My overarching research objective is to develop intelligent agents capable of autonomously acquiring knowledge and adapting their behavior through interactions within the environment. To pursue this goal, my primary areas of research focus: (i) Unsupervised Reinforcement Learning, (ii) Skill Discovery, (iii) World Models, and (iv) Robotics. Feel free to send me an e-mail if you want to chat or collaborate with me! |
|
News |
|
May '24 |
Two papers (Hare&Tortoise and CoIn) are accepted to ICML 2024! |
|
Sep '23 |
Two papers (DISCO-DANCE and PLASTIC) are accepted to NeurIPS 2023! |
Research
|
|
Slow and Steady Wins the Race: Maintaining Plasticity with Hare and Tortoise Networks
Hojoon Lee, Hyeonseo Cho, Hyunseung Kim, Donghu Kim, Dugki Min, Jaegul Choo, Clare Lyle ICML, 2024 Preprint To allow the network to continually adapt and generalize, we introduce Hare and Tortoise architecture, inspired by the complementary learning system of the human brain. |
|
A Simple Convolution Injector for Vision Transformer Towards Effective Adaptation in Visuo-Motor Control
Dongyoon Hwang*, Byungkun Lee*, Hojoon Lee, Hyunseung Kim, Jaegul Choo ICML, 2024 Preprint We introduce CoIn, a lightweight and simple add-on module, which effectively adapts pretrained Vision Transformers for visuo-motor control. |
|
DISCO-DANCE: Learning to Discover Skills through Guidance
Hyunseung Kim*, Byungkun Lee*, Hojoon Lee, Dongyoon Hwang, Sejik Park, Kyushik Min, Jaegul Choo NeurIPS, 2023 project page / paper / code / poster We introduce an Unsupervised Skill Discovery algorithm designed to provide direct guidance to encourage exploration. |
|
PLASTIC: Improving Input and Label Plasticity for Sample Efficient Reinforcement Learning
Hojoon Lee*, Hanseul Cho*, Hyunseung Kim*, Daehoon Gwak, Joonkee Kim, Jaegul Choo SeYoung Yun, Chulhee Yun, NeurIPS, 2023 paper / code / poster We introduce PLASTIC, a simple-to-use algorithm that addresses the loss of plasticity phenomenon and enhances sample efficiency in reinforcement learning. |
|
DraftRec: Personalized Draft Recommendation for Winning in Multi-Player Online Battle Arena Games
Hojoon Lee*, Dongyoon Hwang*, Hyunseung Kim, Byungkun Lee, Jaegul Choo WWW, 2022 paper / code / poster Developed a personalized champion recommendation system in League of Legends with a hierarchical transformer architecture. |
|
This template is a modification to Jon Barron's website. |