About me

About me

CV

I’m a Ph.D student of Efficient Computing Lab in the Computer Science & Engineering depeartment at the POSTECH, South Korea.

My main research interest mostly focuses on designing acceleration systems for AI to enable people to take full advantage of AI. Also, I’m interested in core computer vision problems such as representation learning and diffusion model. In particular, I’m interested in the following topics.

• Neural Network Compression
  • Quantization ( NeurIPS23 , CVPR23 ) ,
  • Pruning( CODES+ISSS21).
• Computer Systems
  • GPGPU Kernel (CODES+ISSS21)
• Vision
  • Diffusion Model ( NeurIPS23, ECCV24, CVPR25 )
  • Multi modal learning (Vision + Language) ( NeurIPS23, IEEE Access23 )

News

• [Feb. 25] 1 paper has been accepted to CVPR 2025 .
• [Jul. 24] 1 paper has been accepted to ECCV 2024 .
• [Sep. 23] 2 papers have been accepted to NeurIPS 2023.
• [Jun. 23] 1 paper has been accepted to IEEE Access.
• [Feb. 23] 1 paper has been accepted to CVPR 2023.
• [Mar. 22] I started my graduate student course in POSTECH.
• [Oct. 21] 1 paper has been accepted to CODES+ISSS and ACM TECS.

Publications

Preprint

International Conference

( * co-first author )

• PCM : Picard Consistency Model for Fast Parallel Sampling of Diffuson Models
  Junhyuk So, Jiwoong Shin, Chayeon Jang, Eunhyeok Park
  CVPR 2025, To be appear.

• FRDiff : Feature Reuse for Universal Training-free Acceleration of Diffusion Models
  Junhyuk So*, Jungwon Lee*, Eunhyeok Park
  ECCV 2024, Sep, 2024, Milano.

• Temporal Dynamic Quantization for Diffusion Model
  Junhyuk So*, Jungwon Lee*, Daehyun Ahn, Hyungjun Kim, Eunhyeok Park
  NeurIPS 2023, Dec, 2023, New Orleans.

• Geodesic Multi-Modal Mixup for Robust Fine-Tuning
  Junhyuk So*, Changdae Oh*, Hoyoon Byun, YongTaek Lim, Minchul Shin, Jong-June Jeon, Kyungwoo Song
  NeurIPS 2023, Dec, 2023, New Orleans.

• NIPQ : Noise porxy based Integraded Psuedo Quantization
  Juncheol Shin*, Junhyuk So*, Sein Park, Seungyeop Kang, Sungjoo Yoo, Eunhyeok Park
  CVPR 2023, June, 2023, Vancouver.

• Robust Contrastive Learning with Dynamic Mixed Margin
  Junhyuk So*, YongTaek Lim, Yewon Kim, Changdae Oh, Kyungwoo Song
  IEEE Access , June, 2023.

• Machine-Learning based Optimal Design of a Wireless Power Transfer Coil for Battery-Powered Tram
  Eunchong Noh, Junhyuk So, Seunghwan Lee,
  ICPE 2023-ECCE Asia , May, 2023, Jeju.

• Learning Fair Representation via Distributional Contrastive Disentanglement
  Changdae Oh, Heeji Won, Junhyuk So, Taero Kim, Yewon Kim, Hosik Choi, Kyungwoo Song
  KDD 2022, July, 2022, Washington DC.

• Exploiting Activation Sparsity for Fast CNN Inference on Mobile GPUs
  Chanyoung Oh*, Junhyuk So*, Sumin Kim*, Youngmin Yi
  CODES+ISSS 2021 Oct, 2021, Virtual,
  ACM TECS(Journal Track) , 2021.

Education

• M.S/Ph.D(Combined) in Computer Science and Engineering, POSTECH.
  Mar. 2022 - Current

• B.S. in Electrical and Computer Engineering, University Of Seoul.
  Mar. 2018 - Dec. 2021