안태건 Taegun An

Introduction to Myself, Taegun An, Github page, May 2024
My research interests lie in efficient and explainable neural architecture design, including NAS, optimization, and parameter–performance trade-offs. I like asking how much performance we can squeeze out of an architecture—before simply throwing more parameters at it.
Education
-
[2020 ~ now] Korea University (KU)
Ph.D. Student (Integrated Course of MS & Ph.D), Computer Science & Engineering
-
[2016 ~ 2020] Ulsan National Institute of Science and Technology (UNIST)
Bachelor’s degree, Computer Science & Engineering
-
[2018] University of Arts London (UAL)
short-term exchange course
Selected Publications
- On the Depth Scalability of Logic Gate Networks;
Taegun An, Dohun kim, Haebeom Lee and Changhee Joo*, Arxiv, 2026
- DAG-NAS: Explainable Neural Architecture Search for Reinforcement Learning via Scalar-level DAG Modeling;
Taegun An and Changhee Joo*, Neural Networks, Feb. 2026
- HCF: Hierarchical Cascade Framework for Distributed Multi-Stage Image Compression
Junhao Cai; Taegun An, Chengjun Jin, Sung Il Choi, Juhyun Park, and Changhee Joo*, AAAI 2026. (Oral)
- CycleGANAS: Differentiable Neural Architecture Search for CycleGAN;
Taegun An and Changhee Joo*, CVPR-NAS Workshop, June 2024.
AI for Networks & Communications
- RXC: Runtime Extra Compression for Time-Critical Content Delivery;
J Bae, H Kang, T Kim, T An, S Lee, J Cai, C Joo, and K Lee, IEEE Communications Magazine (ComMag).
- Dynamic downlink interference management in LEO satellite networks without direct communications;
J Yun, T An, H Jo, BJ Ku, D Oh, C Joo, IEEE Access
- Downlink spectrum sharing of heterogeneous communication systems in LEO satellite networks;
J Yun, T An, H Jo, BJ Ku, D Oh, C Joo, Journal of Communications and Networks (JCN)
- Multi-agent adaptive frequency block selection of LEO satellites for interference avoidance;
J Yun, T An, BJ Ku, D Oh, C Joo, Journal of Communications and Networks (JCN)
Honors/Awards
- UNIST Cultural Exchange Scholarship, 2018
- Best Presentation Award, A3 Foresight Program, 2023
- Expert Research Personnel, 2023 ~ Current
Skills
- Mostly Python, occasionally C/C++
- Korean: Native · English: Upper-Intermediate · Japanese: merely communicable
Activities
- UNIST Sudam(手談) (Boardgame Club, Baduk Amateur 初段(1st Dan, Shodan) ⚫⚪)
- UNIST Ensemble (Orchestra Club, Violinist by hobby 🎻)
- KUFIT (Fitness Club, bodybuilding, 💪)
S:155kg B:95kg D:155kg Total: 405kg