Joonhyung Park

Ph.D student at Graduate School of AI, Korea Advanced Institute of Science and Technology (KAIST)

I’m a Ph.D candidate at Machine Learning and Intelligence Lab (MLILAB) in KAIST, advised by Prof. Eunho Yang.

Graph-Driven Modal Understanding
My research interest lies in elucidating unstructured data modalities through the lens of graph structure, e.g. object graphs (3D vision) and knowledge graphs (natural language). I work on building machine learning algorithms that leverage relational information therein, revisiting complex real-world problems within a graph-based framework to provide a structured understanding of unstructured data. Recently, my primary focus has been on cross-modal learning/alignment by means of a graph, oriented towards resolving open-vocabulary challenges.

Learning on 3D Vision
My primary research interest in 3D vision falls into two branches following: 1) Cross-modal 3D understanding. It aims to harness the power of auxiliary data modalities for an in-depth comprehension of complex 3D data. Currently, I’m working on open-vocabulary 3D scene segmentation with object-relational graphs leveraging recent language foundation models’ capabilities. 2) Sim-to-real adaptation for 3D data. My recent research efforts have been dedicated to narrowing the domain gap between synthetic and real-world 3D data. Ranging from developing adaptation strategies to curating 3D photorealistic datasets, my recent objective is to facilitate successful sim-to-real transfer across a broad range of 3D vision tasks.

Conference Publications

  • PC-Adapter: Topology-Aware Adapter for Efficient Domain Adaption on Point Clouds with Rectified Pseudo Label [paper]
    Joonhyung Park, Hyujin Seo, Eunho Yang
    ICCV 2023

  • SGEM: Test-Time Adaptation for Automatic Speech Recognition via Sequential-Level Generalized Entropy Minimization [paper]
    Changhun Kim, Joonhyung Park, Hajin Shim, Eunho Yang
    INTERSPEECH 2023 (Congrats on my mentee’s paper! :tada:)

  • CALA: Connectivity- and Attribute-Aware Logit Adjustment for Class-Imbalanced Graphs
    Joonhyung Park*, Jaeyun Song*, Eunho Yang (* : equal contribution)
    Under Review

  • WeavSpeech: Data Augmentation Strategy for Automatic Speech Recognition via Semantic-Aware Weaving [paper]
    Kyusung Seo, Joonhyung Park, Jaeyun Song, Eunho Yang
    ICASSP 2023 (Congrats on my mentee’s paper! :tada:)

  • TAM: Topology-Aware Margin Loss for Class-Imbalanced Node Classification [paper]
    Jaeyun Song*, Joonhyung Park*, Eunho Yang (*: equal contribution)
    ICML 2022

  • GraphENS: Neighbor-Aware Ego Network Synthesis for Class-Imbalanced Node Classification [paper]
    Joonhyung Park*, Jaeyun Song*, Eunho Yang (*: equal contribution)
    ICLR 2022

  • Saliency Grafting: Innocuous Attribution-Guided Mixup with Calibrated Label Mixing [paper]
    Joonhyung Park, June Yong Yang, Jinwoo Shin, Sung Ju Hwang, Eunho Yang
    AAAI 2022 (oral presentation, 380/9020=4.21%)

  • Graph Transplant: Node Saliency-Guided Graph mixup with Local Structure Preservation [paper]
    Joonhyung Park*, Hajin Shim*, Eunho Yang (*: equal contribution)
    AAAI 2022

Work Experiences

  • Research Intern, University of Virginia, Charlottesville, VA, Jun. 2018 - Aug. 2018
    • Advisor: Prof. Homa Alemzadeh
    • Medical concept extraction in text data for Cognitive Assistant System of emergency medical response (supported by the National Institute of Standards and Technology).
  • Research Intern, Collaborative Robots Research Center, DGIST, Daegu, Jun. 2017 - Aug. 2017
    • Development of treadmill for stroke hemiplegic patients

Education

  • Ph.D. (integrated) in Graduate School of AI, Korea Advanced Institute of Science and Technology (KAIST), Mar. 2021 - Present

  • M.S. in Graduate School of AI, Korea Advanced Institute of Science and Technology (KAIST), Mar. 2020 - Mar. 2021

  • B.E. in Computer Science Engineering, Daegu Gyeongbuk Institute of Science & Technology (DGIST), Mar. 2016 - Feb. 2020 - (Summa Cum Laude)

Projects

  • Sub-task generation based point/regional Out-Of-Distribution detection, Samsung Electronics, Sep. 2020 - Sep. 2025
  • Predicting graph properties with few labels using Graph Neural Networks, Samsung Electronics, Sep. 2020 - Sep. 2025
  • Machine learning model for the prediction of Hypoxaemia during Endoscopic Retrograde Cholangiopancreatography, Yonsei Severance Hospital, Mar. 2020 - Jun. 2020

Acamdeic Services

  • Conference Reviewer
    • International Conference on Machine Learning (ICML)
    • Neural Information Processing Systems (NeurIPS)
    • International Conference on Learning Representations (ICLR)
    • Computer Vision and Pattern Recognition (CVPR)
    • AAAI Conference on Artificial Intelligence (AAAI)
    • International Conference on Acoustics, Speech, and Signal Processing (ICASSP)
    • Learning on Graphs (LoG)
  • Journal Reviewer
    • Transactions on Neural Networks and Learning Systems (TNNLS)