Kibok Lee

Ph.D. Candidate
University of Michigan, Ann Arbor

kibok _at_ umich _dot_ edu

I am a Ph.D. candidate in the Computer Science & Engineering department at the University of Michigan, Ann Arbor. My advisor is Honglak LeeMy research interests lie in machine learning and computer vision. I have been working on lifelong learning and generalization in deep learning. I am interested in leveraging unlabeled data in the open world to reduce supervision.
I interned at Uber ATG last summer, advised by Ersin Yumer. I worked on deep learning in 3D vision.


Research Topics
  • Confidence Calibration
  • Out-of-Distribution Detection
  • Continual (Lifelong) Learning
  • Domain Generalization
  • (Generalized) Zero-Shot Learning

Publications (* = equal contribution)

Network Randomization: A Simple Technique for Generalization in Deep Reinforcement Learning
Kimin Lee*, Kibok Lee*, Jinwoo Shin, Honglak Lee
In ICLR, 2020. [paper (arXiv, full version)]
Preliminary version [poster] was presented in NeurIPS Workshop on Deep Reinforcement Learning, 2019. Contributed talk

Overcoming Catastrophic Forgetting with Unlabeled Data in the Wild
Kibok Lee, Kimin Lee, Jinwoo Shin, Honglak Lee
In ICCV, 2019. [paper (arXiv, full version)][GitHub][poster]
[Short version][slides] was presented in CVPR Workshop on Uncertainty and Robustness in Deep Visual Learning, 2019. Spotlight
Comment: Our proposed method outperforms the state-of-the-art even without unlabeled data.

Robust Inference via Generative Classifiers for Handling Noisy Labels
Kimin Lee, Sukmin Yun, Kibok Lee, Honglak Lee, Bo Li, Jinwoo Shin
In ICML, 2019. Long presentation [paper (arXiv, full version)][GitHub]

Automatic Correction of Lithography Hotspots with a Deep Generative Model
Woojoo Sim*, Kibok Lee*, Dingdong Yang, Jaeseung Jeong, Ji-Suk Hong, Sooryong Lee, Honglak Lee
In SPIE Advanced Lithography, 2019. Invited (long presentation) [paper]

A Simple Unified Framework for Detecting Out-of-Distribution Samples and Adversarial Attacks
Kimin Lee, Kibok Lee, Honglak Lee, Jinwoo Shin
In NeurIPS, 2018. Spotlight (168/4856=3.5%) [paper (arXiv, full version)][GitHub]
Comment: Adversarial samples can be used to validate our proposed method without OOD samples; check the right side of the Table 2.

Hierarchical Novelty Detection for Visual Object Recognition
Kibok Lee, Kimin Lee, Kyle Min, Yuting Zhang, Jinwoo Shin, Honglak Lee
In CVPR, 2018. [paper (arXiv, full version)][GitHub][poster]
Comment: Hierarchical novelty detection is a generalization of generalized zero-shot learning, in the sense that it does not require semantic information about zero-shot classes.

Training Confidence-calibrated Classifiers for Detecting Out-of-Distribution Samples
Kimin Lee, Honglak Lee, Kibok Lee, Jinwoo Shin
In ICLR, 2018. [paper (arXiv, full version)][GitHub]

Towards Understanding the Invertibility of Convolutional Neural Networks
Anna Gilbert, Yi Zhang, Kibok Lee, Yuting Zhang, Honglak Lee
In IJCAI, 2017. [paper (arXiv, full version)][slides][poster]

Augmenting Supervised Neural Networks with Unsupervised Objectives for Large-scale Image Classification
Yuting Zhang, Kibok Lee, Honglak Lee

A Flexible Framework for Online Document Segmentation by Pair-wise Stroke Distance Learning
Adrien Delaye, Kibok Lee
Pattern Recognition, 2015. [paper (ScienceDirect) (ResearchGate)]

On the Equivalence of Linear Discriminant Analysis and Least Squares
Kibok Lee, Junmo Kim
In AAAI, 2015. [paper (gdrive)] [supplementary]