I am a Ph.D. student at ALIN-LAB, advised by Prof. Jinwoo Shin at Korea Advanced Institute of Science and Technology (KAIST).

My research interests focus on trustworthy machine learning for real-world deployment. I mainly focus on robust learning schemes for handling distributional shifts, e.g., out-of-distribution, adversarial examples, and natural corruptions. However, I also have a broad interest in the interpretability and the transparency of deep networks.

Contact: jihoontack at kaist.ac.kr

Publications

(C: Conference, W: Workshop, P: Preprint, *: Equal contribution)

  • [C3/W] Meta-Learning Sparse Implicit Neural Representations
    Jaeho Lee*, Jihoon Tack*, Namhoon Lee, Jinwoo Shin
    Neural Information Processing Systems (NeurIPS) 2021
    Sparsity in Neural Networks workshop (SNN) 2021
    [paper] [code] (stay tuned!)

  • [W3] Consistency Regularization for Training Confidence-Calibrated Classifiers
    Youngbum Hur*, Jihoon Tack*, Eunho Yang, Sung Ju Hwang and Jinwoo Shin
    ICML Workshop on Uncertainty & Robustness in Deep Learning (UDL) 2021
    [paper]

  • [W2] Entropy Weighted Adversarial Training
    Minseon Kim, Jihoon Tack, Jinwoo Shin, Sung Ju Hwang
    ICML Workshop on Adversarial Machine Learning (AdvML) 2021
    [paper] [code]

  • [W1] Consistency Regularization for Adversarial Robustness
    Jihoon Tack, Sihyun Yu, Jongheon Jeong, Minseon Kim, Sung Ju Hwang, Jinwoo Shin
    ICML Workshop on Adversarial Machine Learning (AdvML) 2021, Oral presentation
    [paper] [code] [slide]

  • [C2] CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances
    Jihoon Tack*, Sangwoo Mo*, Jongheon Jeong, Jinwoo Shin
    Neural Information Processing Systems (NeurIPS) 2020
    Qualcomm Innovation Fellowship Korea 2020
    [paper] [code] [slide] [poster]

  • [C1] Adversarial Self-Supervised Contrastive Learning
    Minseon Kim, Jihoon Tack, Sung Ju Hwang
    Neural Information Processing Systems (NeurIPS) 2020
    [paper] [code] [site]

Education

  • M.S. & Ph.D. integrated course in Artificial Intelligence, Korea Advanced Institute of Science and Technology (KAIST), 2019 - present (advisor: Prof. Jinwoo Shin)
  • B.S. in Electrical Engineering and Mathematics (minored), Korea Advanced Institute of Science and Technology (KAIST), 2014 - 2019

Honor & Award

  • Outstanding reviewer award, CVPR 2021
  • Qualcomm Innovation Fellowship Korea 2020
  • Top reviewer award (top 10%), NeurIPS 2020

Academic Services

  • Conference reviewer
    • NeurIPS 2020, 2021
    • ICML 2021 (expert)
    • ICLR 2022
    • CVPR 2021, 2022
    • ICCV 2021
  • Journal reviewer
    • IEEE Transactions on Neural Networks and Learning Systems

Talks

  • “Consistency Regularization for Adversarial Robustness”
    • @ ICML workshop on AdvML (2021): Contribution Talk, Jul 2021
  • “CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances”
    • @ Qualcomm AI Research: Qualcomm Innovation Fellowship Winner Talk, Feb 2021
    • @ NeurIPS Social: ML in Korea, Dec 2020
    • @ Korea Software Congress (KSC): Korea Post-NeurIPS-2020 Workshop, Dec 2020

Teaching

  • Head TA, AI503: Mathematics for AI, KAIST Fall 2021
  • Head TA, AI602: Recent Advances in Deep Learning, KAIST Spring 2021
  • TA, AI504: Programming for AI, KAIST Fall 2020
  • TA, AI-Expert Program, Samsung-DS Summer 2020
  • TA, AI703: Systems and Applications of AI and ML, KAIST Spring 2020
  • TA, EE209: Programming Structure for Electrical Engineering, KAIST Fall 2019

Project

  • Interpretable ML for object detection, 2020 - present
  • Used car project, KB Capital, 2017 - 2020
    Building an AI system that predicts price, depreciation, and sales duration of the used car.
    Currently, my system is deployed at KBchachacha service (only Korean supported).