Home
Group
Publications
Teaching

Donghwan Kim

alt text 

About Me

I am an associate professor in the Department of Mathematical Sciences at KAIST. I am also affiliated with the Graduate School of AI for Math (AI4Math).

Curriculum Vitae, Google scholar entry

Current Research Interests

  • Optimization for Machine Learning

    • Convergence of (Stochastic) Gradient Method in Minimax Optimization, Loss Landscapes, Learning Dynamics

  • Generalization in Machine Learning

    • Implicit Bias of Gradient Methods, Benign Overfitting, Double Descent

  • Application: Generative Models

    • Generative Adversarial Networks (GANs), Flow Matching Models, Language Models

  • Mathematical Reasoning in AI

Recent News

  • 26.03.02: We have three undergraduate students (Woohyun Ahn, Jihoon Seo and Juho Song) doing an individual study during this spring semester.

  • 26.03.02: Eunsung Ji joined our group as an M.S.-Ph.D. student.

  • 26.03.02: Yongseok Kwon and Shiwon Park joined our group as AI4Math M.S.-Ph.D students, co-advised by Prof. Noseong Park and Prof. Changhyun Kwon, respectively.

  • 26.02.20: Jiseok Chae received an outstanding dissertation award from KAIST.

  • 26.02.02: I delivered a tutorial ‘‘From polynomial fitting to generative models: Understanding learning and generalization from first principles’’ at the KMS-NIMS-KRIMS AI Winter School.

  • 26.01.02: We have one undergraduate student (Minhak Song) doing a URP during the winter and spring semesters.

Past News

  • 25.12.02: I attended NeurIPS 2025 in San Diego.

  • 25.11.24: Jiseok Chae successfully defended his thesis.

  • 25.10.23: Kyungjae Lee received the best poster award at the KMS Annual Meeting.

  • 25.10.22: I attended the KMS Annual Meeting, where I organized a session titled ‘‘Mathematics of Machine Learning’’ and gave a talk ‘‘Statistical distances and optimization in generative models’’ in the ‘‘Math for AI’’ session. [Slides]

  • 25.09.23: Our first paper on GAN is accepted to the Optimization for Machine Learning (OPT) Workshop at NeurIPS 2025.

  • 25.09.15: We have one undergraduate student (Gi Tae Kwon) doing individual study this fall semester.

  • 25.09.01: Natdanai Treesakul joined our group as an M.S. student.

  • 25.08.16: I am back at KAIST after my sabbatical year.

  • 25.07.20: I attended International Conference on Continuous Optimization (ICCOPT) 2025 in Los Angeles, and gave a talk under the title ‘‘Strict non-optimality in minimax optimization and its application to GAN.’’ [Slides]

  • 25.07.14: I attended ICML 2025 in Vancouver.

  • 25.06.09: Our first paper on generalization is accepted to the High-dimensional Learning Dynamics (HiLD) Workshop at ICML 2025.

  • 25.04.04: I gave a talk at the University of Wisconsin-Madison Applied and Computational Mathematics Seminar.

  • 25.01.23: I gave a talk at the University of Michigan ECE Communications and Signal Processing (CSP) seminar. [Video] [Slides]

  • 25.01.08: I gave a talk at the UCLA Applied Math Colloquium under the title ‘‘How to make the gradient descent-ascent converge to local minimax optima.’’ [Video]

  • 24.12.10: I attended NeurIPS 2024 in Vancouver.

  • 24.11.05: I gave a talk at the UC Davis Mathematics of Data and Decisions at Davis (MADDD) seminar.

  • 24.09.25: We have one paper accepted to NeurIPS 2024.

  • 24.08.16: I started my sabbatical year at UC Berkeley.

  • 24.07.04: Yejun Kim and Dongchan Shin joined our group as M.S.-Ph.D. students.

  • 24.05.16: I gave a talk at the KIAS under the title ‘‘How to make the gradient descent-ascent converge to local minimax optima.’’

  • 24.05.02: We have two papers accepted to ICML 2024.

  • 24.04.19: I organized the ‘‘Optimization and Machine Learning’’ session at the 2024 KMS Spring Meeting.

  • 24.04.19: Sucheol Lee, the first Ph.D. graduate from our group, received an outstanding dissertation award from the Korean Mathematical Society (KMS).

  • 24.03.04: Jimyeong Kim joined our group as a postdoc.

  • 24.02.01: Kyungjae Lee joined our group as a M.S.-Ph.D. student.

  • 24.01.17: We have one paper accepted to ICLR 2024.

  • 24.01.02: We have one undergraduate student (Minhee Hong) doing individual study this winter semester.

  • 23.07.28: I gave a talk at the KAI-X Mathematics Summer School under the title ‘‘The power of gradient descent in machine learning.’’

  • 23.06.26: We have one graduate student (Keyan Shi, Technische Universitat Munchen) doing KAI-X research internship this summer semester.

  • 23.06.21: We have one paper accepted to COLT (Open problems track) 2023.

  • 23.06.19: We have two undergraduate students (Changmin Kang, Changhyun Ko) doing individual study this summer semester.

  • 23.05.31: I gave a talk at the SIAM Conference on Optimization under the title ‘‘Fast extra gradient methods for smooth structured nonconvex-nonconcave minimax problems.’’

  • 23.05.19: I gave a talk at the KSIAM Spring Conference under the title ‘‘Two-timescale extragradient converges to local minimax points.’’

  • 23.04.28: I gave a talk at the KMS Spring Meeting under the title ‘‘Two-timescale extragradient converges to local minimax points.’’

  • 23.04.24: I gave a talk at the HKUST-KAIST-NUS Joint Workshop on Applied and Computational Mathematics under the title ‘‘Two-timescale extragradient converges to local minimax points.’’

  • 23.04.05: I gave a talk at the KAIST BTM S&T Biz Colloquium under the title ‘‘AI learning and optimization.’’

  • 23.03.31: I gave a talk at the Postech IME seminar about our recent research under the title ‘‘Gradient methods for minimax optimization.’’

  • 23.01.13: We have two undergraduate students (Seong Bae Lim, Yujun Kim) doing individual study this winter semester.

  • 23.01.13: I will now start posting our group's recent news here.

Contact Info

donghwankim (at) kaist (dot) ac (dot) kr