Seungjae Shin
I am a Ph.D. student at Applied Artificial Intelligence Lab, advised by Prof. Il-Chul Moon at Korea Advanced Institute of Science and Technology (KAIST). Previously, I received B.S in Industrial Engineering from KAIST in Feb. 2018 and M.S in industrial Engineering from KAIST in Feb. 2020, respectively.
tmdwo0910 [at] kaist [dot] ac [dot] kr // tmdwo0910 [at] gmail [dot] com
Google Scholar / LinkedIn / Twitter / CV
My research has been focused on (but not limited to)
Data-Centric Generalization (e.g. Coreset selection, Dataset Distillation, Domain generalization)
[C8, C9,C12]
ML Efficiency (e.g. Model quantization, Coreset selection)
[C8, C9]
ML Robustness (e.g. Debiasing, Fairness, Class-imbalance, Noisy Label Learning)
[C2, C3, C5, C7, C11]
Representation Learning, Generative Model
[C1, C4, C6, C10]
Publications (C: Conferences, J: Journal, W: Workshop, P: preprint, * denotes the equal contribution. )
1st (or Co-1st) Authored
[C12] Unknown Domain Inconsistency Minimization for Domain Generalization [TBD]
{Seungjae Shin*, Heesun Bae*}, Byeonghu Na, Yoon-yeong Kim and Il-Chul Moon
The Twelfth International Conference on Learning Representations (ICLR) 2024
[C9] Frequency domain-based Dataset Distillation [paper, arxiv, code]
{Donghyeok Shin*, Seungjae Shin*} and Il-Chul Moon
[C8] Loss Curvature Matching for Dataset Selection and Condensation [paper, arxiv, code]
{Seungjae Shin*, HeeSun Bae*}, DongHyeok Shin, Weonyoung Joo, Il-Chul Moon
International Conference on Artificial Intelligence and Statistics (AISTATS) 2023
Winner, Qualcomm Innovation Fellowship Korea 2023
[W2] Improving Group-based Robustness and Calibration via Ordered Risk and Confidence Regularization [paper]
Seungjae Shin, Byeonghu Na, HeeSun Bae, JoonHo Jang, Hyemi Kim, Kyungwoo Song, Youngjae Cho, Il-Chul Moon
[C7] From Noisy Prediction to True Label: Noisy Prediction Calibration via Generative Model [arxiv, paper, code]
{HeeSun Bae*, Seungjae Shin*}, Byeonghu Na, JoonHo Jang, Kyungwoo Song, Il-Chul Moon
Winner, Qualcomm Innovation Fellowship Korea 2022
[J2] Evaluation of optimal scene time interval for out-of-hospital cardiac arrest using a deep neural network [paper]
{Seungjae Shin*, Heesun Bae*}, Giwoon Kim, Youngsoon Cho, Dongwook Lee, Donggil Jeong, HyunJoon Kim, Hyunjung Lee, Hyungjun Moon
American Journal of Emergency Medicine (IF = 3.7)
[C2] Neutralizing Gender Bias in Word Embedding with Latent Disentanglement and Counterfactual Generation [paper]
Seungjae Shin, Kyungwoo Song, Joonho Jang, Hyemi Kim, Weonyoung Joo, and Il-Chul Moon
Empirical Methods in Natural Language Processing (EMNLP) Findings 2020
Co-Authored
[C11] Dirichlet-based Per-Sample Weighting by Transition Matrix for Noisy Label Learning [TBD]
Heesun Bae, Seungjae Shin, Byeonghu Na and Il-Chul Moon
The Twelfth International Conference on Learning Representations (ICLR) 2024
[C10] Make Prompts Adaptable : Bayesian Modeling for Vision-Language Prompt Learning with Data-Dependent Prior [TBD]
Youngjae Cho, HeeSun Bae, Seungjae Shin, YeoDong Youn, Weonyoung Joo, and Il-chul Moon
[C6] Soft Truncation: A Universal Training Technique of Score-based Diffusion Model for High Precision Score Estimation [arxiv, paper, code]
Dongjun Kim, Seungjae Shin, Kyungwoo Song, Wanmo Kang, Il-chul Moon
[C5] ABC : Auxiliary Balanced Classifier for Class-imbalanced Semi-supervised Learning [paper, code]
Hyuck Lee, Seungjae Shin, Heeyoung Kim
[C4] Refine Myself by Teaching Myself : Feature Refinement via Self-Knowledge Distillation [paper, code]
Mingi Ji, Seungjae Shin, Seunghyun Hwang, Gibeom Park, Il-Chul Moon
Conference on Computer Vision and Pattern Recognition (CVPR) 2021
[C3] Counterfactual Fairness with Disentangled Causal Effect Variational Autoencoder [paper]
Hyemi Kim, Seungjae Shin, JoonHo Jang, Kyungwoo Song, Weonyoung Joo, Wanmo Kang, Il-Chul Moon
American Association for Artificial Intelligence (AAAI) 2021
[C1] Bivariate Beta-LSTM [paper, code]
Kyungwoo Song, JoonHo Jang, Seungjae Shin, and Il-Chul Moon
American Association for Artificial Intelligence (AAAI) 2020
[W1] FEWER : Federated Weight Recovery [paper]
Yongjin Shin, Gihun Lee, Seungjae Shin, Se-young Yun, Il-Chul Moon
DistributedML'20: Proceedings of the 1st Workshop on Distributed Machine Learning
[J1] Forecasting the Concentration of Particulate Matter in the Seoul Metropolitan Area Using a Gaussian Process Model [paper]
Joonho Jang, Seungjae Shin, Hyunjin Lee, Il-Chul Moon
Sensors 2020 (IF = 3.576)
[P3] Posterior-Aided Regularization for Likelihood-Free Inference [pdf]
Dongjun Kim, Kyungwoo Song, Seungjae Shin, and Il-Chul Moon
[P2] Adversarial Likelihood-Free Inference on Black-Box Generator [pdf]
Dongjun Kim, Weonyoung Joo, Seungjae Shin, and Il-Chul Moon
[P1] Generalized Gumbel-Softmax Gradient Estimator for Various Discrete Random Variables [pdf]
Weonyoung Joo, Dongjun Kim, Seungjae Shin, and Il-Chul Moon
Education
Ph.D. Candidate in Industrial Engineering, KAIST, Daejeon, Korea (advisor: Prof. Il-Chul Moon) (2020.03 - present)
M.S. in Industrial Engineering, KAIST, Daejeon, Korea (advisor: Prof. Il-Chul Moon) (2018.03 - 2020.02)
B.S. in Industrial Engineering, KAIST, Daejeon, Korea (2013.03 - 2018.02)
Experience
Visiting Student, Department of Electrical and Computer Engineering, National University of Singapore (NUS) (2017.01 ~2017.05)
Computer Science Educator for Juniors (Data Diving, Daejeon EduSci Center, Aura Edu) (2020 ~ )
Honors & Awards
Winner, Qualcomm Innovation Fellowship Korea 2023
Winner, Qualcomm Innovation Fellowship Korea 2022
Best Paper Award (Junior), Korean Software Congress, 2017
1st Prize, KAIST Invention Award, 2017
Honor student scholarship, KAIST, 2016S
Dean's list, KAIST, 2014
Academic Services
Program Committee (Conference Reviewer)
ICML (2022 - )
NeurIPS (2022 - )
ICLR (2024 - )
CVPR (2021 - )
EMNLP (2021 - )
ACL (2023 - )
Other Interests
🥋 Purple belt (one grau) for Brazilian Jiu-Jitsu (BJJ)
🎧 Love listening music (Lo-fi, Hip hop, R&B)