Jongin Lim

I'm a Staff Researcher at Samsung AI Center (formerly SAIT AI Research Center), developing robust and efficient machine learning models tailored for industrial applications. My work has focused on AI-driven automation in manufacturing, spanning diverse modalities including vision, language, structured data (e.g., graph, tabular), and time-series domains.

I received my Ph.D. (through an integrated M.S./Ph.D. program) in Electrical and Computer Engineering from Seoul National University in 2022, under the supervision of Prof. Jin Young Choi. I received my B.S. in Electrical and Computer Engineering from Seoul National University in 2016.

Email  /  CV  /  Google Scholar  /  Github

News

• [2025.07] 📢 I will serve as Program Committee for AAAI 2026.
• [2025.07] ✈️ I will attend ICML 2025 in Vancouver, Canada 🇨🇦
• [2025.05] 📝 Two papers were accepted to ICML 2025.
• [2025.01] 📝 One paper was accepted to ICLR 2025.

Research

My research focuses on developing generalizable and transferable ML models for real-world applications. Recently, I have worked on improving model robustness under data imbalance (e.g., IB, PRIME), distribution shifts (e.g., BiasAdv), and noisy labels (e.g., SLC), but I am open to exploring a broader range of topics. I have also investigated deep metric learning (e.g., HIST), graph-based learning (e.g., CAD), and realistic human motion generation (e.g., PMnet). First-author papers are highlighted.

We propose Proxy-based Representation learning for IMbalanced rEgression (PRIME), a novel framework that leverages learnable proxies to construct a balanced and well-ordered feature space for imbalanced regression.

We address learning tasks on graphs with missing features, enhancing the applicability of graph neural networks to real-world graph-structured data.

We introduce a new challenging task to model the interactions of OOD nodes in a graph, termed spreading OOD detection, where a newly emerged OOD node spreads its property to neighboring nodes. We present a realistic benchmark setup with a new dataset and propose a novel aggregation scheme for the new task.

We propose a novel learning framework that selectively suppresses noisy samples while avoiding underfitting clean data. Our framework incorporates label confidence as a measure of label noise, enabling the network model to prioritize the training of samples deemed to be noise-free.

We propose a novel data augmentation approach termed Bias-Adversarial augmentation (BiasAdv) that supplements bias-conflicting samples with adversarial images. Our key idea is that an adversarial attack on a biased model that makes decisions based on spurious correlations may generate synthetic bias-conflicting samples, which can then be used as augmented training data for learning a debiased model.

We formulate deep metric learning as a hypergraph node classification problem to capture multilateral relationship by semantic tuplets beyond previous pairwise relationship-based methods.

We propose a new loss function for imbalanced visual classification, which alleviates the influence of samples that cause an overfitted decision boundary.

We propose a novel stochastic process, called Class-Attentive Diffusion (CAD), that strengthens attention to intra-class nodes and attenuates attention to inter-class nodes. In contrast to the previous diffusion methods, CAD considers both the node features and the graph structure with the design of our class-attentive transition matrix that utilizes a classifier.

We propose a novel fine-tuning strategy that allows low-level layers to be sufficiently trained by rolling back the weights of high-level layers to their initial pre-trained weights. Our strategy alleviates the problem of gradient vanishing in low-level layers.

We propose a deep learning framework for unsupervised motion retargeting. In contrast to the existing method, we decouple the motion retargeting process into two parts that explicitly learn poses and movements of a character, reducing the motion retargeting error (average joint position error) from 7.68 (sota) to 1.95 (ours).

We propose a novel deep conditional generative model for human pose transforms. To generate the desired pose-transformed images from a single image, a variational inference model is formulated to disentangle human posture semantics from image identity (human personality, background etc.) in variational auto-encoded latent space.

We propose a selective ensemble deep network architecture for crowd density estimation and people counting. In contrast to existing deep network-based methods, the proposed method incorporates two sub-networks for local density estimation: one to learn sparse density regions and one to learn dense density regions.

We propose a moving object detection algorithm adapting to various scene changes in a moving camera. Our method adapts itself to the dynamic scene changes and outperforms the state-of-the art methods.

We propose a moving object detection algorithm for a monocular dashcam mounted on a vehicle. To deal with dynamic changes of the scene from the dashcam, we propose a new scheme inspired by human-attention inclination for change detection.

Patents

Talks

• [2024.10] 📹 An interview about life as a researcher at SAIT, Samsung Advanced Institute of Technology (SAIT)
• [2022.08] 🎤 Invited talk on Graph-based Representation Learning for Class Discriminative Feature Embedding, Seoul National University (SNU) AI Summer School 2022

Honors & Awards

• [2019.01] 🎖️ Samsung SAIT Scholarship
  Awarded a scholarship from Samsung as part of an industry-academia collaboration program, receiving research funding during the Ph.D. program.
• [2018.09] 🏆 Best Presentation Award, Samsung AI Forum
  Recognized for outstanding presentation in the poster session and awarded a Galaxy Note 9 as a prize. [link]
• [2013.03] 🎖️ Kwanjeong Educational Foundation Scholarship
  Awarded a prestigious scholarship selecting top talent in South Korea, providing full tuition and financial support for two years.
• [2012.09] 🎓 SNU Eminence Scholarship
  Awarded a full tuition scholarship in recognition of academic excellence.
• [2011.09] 🎓 SNU Superior Academic Performance Scholarship
  Awarded a partial tuition scholarship in recognition of academic excellence.

Feel free to steal this website's source code. Do not scrape the HTML from this page itself, as it includes analytics tags that you do not want on your own website — use the github code instead. Also, consider using Leonid Keselman's Jekyll fork of this page.