Min-hwan Oh

Min-hwan Oh is an Assistant Professor in the Graduate School of Data Science at Seoul National University. His primary research interests are in sequential decision making under uncertainty, reinforcement learning, bandit algorithms, statistical machine learning and their various applications.
Prior to joining SNU, he received a Ph.D. in Operations Research with Data Science specialization at Columbia University, where he was advised by Prof. Garud Iyengar and co-advised by Prof. Assaf Zeevi. During his Ph.D., Min-hwan worked on several projects with the Computational and Statistical Learning Group in IBM Research AI at IBM T.J. Watson Research Center.

Email minoh@snu.ac.kr

Address Bldg #942 Graduate School of Data Science, Seoul National University
1 Gwanak-ro, Gwanak-gu, Seoul, South Korea

Curriculum Vitae Google Scholar

Columbia University.

New York, NY, USA

Ph.D. in Operations Research, 2020

Ph.D. Specialization in Data Science

Columbia University.

New York, NY, USA

B.A. in Mathematics-Statistics, 2015

Summa cum laude
Departmental Honors in Statistics
Phi Beta Kappa

Sparsity-Agnostic Lasso Bandit

with Garud Iyengar and Assaf Zeevi

INFORMS Applied Probability Society Student Paper Award Finalist, 2020

Counting and Segmenting Sorghum Heads

with Peder A. Olsen and Karthikeyan Natesan Ramamurthy

Multinomial Logit Contextual Bandits: Provable Optimality and Practicality

with Garud Iyengar

AAAI Conference on Artificial Intelligence (AAAI), 2021, to appear

Crowd Counting with Decomposed Uncertainty

with Peder A. Olsen and Karthikeyan Natesan Ramamurthy

AAAI Conference on Artificial Intelligence (AAAI), 2020

Thompson Sampling for Multinomial Logit Contextual Bandits

with Garud Iyengar

Neural Information Processing Systems (NeurIPS), 2019

Sequential Anomaly Detection using Inverse Reinforcement Learning

with Garud Iyengar

ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD), 2019

Oral presentation in research paper track (top 9% of total submissions)

Automatic Event Detection in Basketball using Hidden Markov Models with Energy based Defensive Assignment

with Suraj Keshri, Sheng Zhang and Garud Iyengar

Journal of Quantitative Analysis in Sports. 15.2: 141-153. 2019

Adaptive Pattern Matching with Reinforcement Learning for Dynamic Graphs

with Hiroki Kanezashi, Toyotaro Suzumura, Dario Garcia-Gasulla and Satoshi Matsuoka

IEEE International Conference on High Performance Computing, Data, and Analytics (HiPC), 2018

Best Paper Award Winner

Learning Graph Topological Features via GAN

with Weiyi Liu, Pin-Yu Chen, Hal Cooper, Sailung Yeung and Toyotaro Suzumura

IEEE Access, 7, 21834--21843, 133600, 2019

Preliminary version appeared at Workshop on Implicit Models, International Conference on Machine Learning (ICML), 2017

Efficient "Shotgun" Inference of Neural Connectivity from Highly Sub-sampled
Activity Data

with Daniel Soudry, Suraj Keshri, Patrick Stinson, Garud Iyengar and Liam Paninski

PLoS Computational Biology, 11 (10), 2015

Graphical Model for Basketball Match Simulation

with Suraj Keshri and Garud Iyengar

MIT Sloan Sports Analytics Conference, 2015

Finalist in Research Paper Competition (top 2% of total submissions)

Thompson Sampling for Contextual Reinforcement Learning

with Garud Iyengar

Directed Exploration in PAC Model-free Reinforcement Learning

with Garud Iyengar

Preliminary version appeared at Exploration in RL Workshop, International Conference on Machine Learning (ICML), 2018

2nd Place Winner, 2018 INFORMS Annual Meeting Poster Competition

Unsupervised segmentation of neuroanatomy from multispectral images

with Uygar Sümbül, Jeremy Wohlwend, Douglass Roossien Jr., Fei Chen, Nicholas Barry, Adam H. Marblestone, John P. Cunningham, Dawen Cai, Edward S. Boyden and Liam Paninski

Min-hwan Oh

Education

Columbia University.

Columbia University.

Preprint

Sparsity-Agnostic Lasso Bandit

Counting and Segmenting Sorghum Heads

Publications

Multinomial Logit Contextual Bandits: Provable Optimality and Practicality

Crowd Counting with Decomposed Uncertainty

Thompson Sampling for Multinomial Logit Contextual Bandits

Sequential Anomaly Detection using Inverse Reinforcement Learning

Automatic Event Detection in Basketball using Hidden Markov Models with Energy based Defensive Assignment

Adaptive Pattern Matching with Reinforcement Learning for Dynamic Graphs

Learning Graph Topological Features via GAN

Efficient "Shotgun" Inference of Neural Connectivity from Highly Sub-sampled
Activity Data

Graphical Model for Basketball Match Simulation

Working Papers

Thompson Sampling for Contextual Reinforcement Learning

Directed Exploration in PAC Model-free Reinforcement Learning

Unsupervised segmentation of neuroanatomy from multispectral images

Education

Columbia University.

Columbia University.

Preprint

Sparsity-Agnostic Lasso Bandit

Counting and Segmenting Sorghum Heads

Publications

Multinomial Logit Contextual Bandits: Provable Optimality and Practicality

Crowd Counting with Decomposed Uncertainty

Thompson Sampling for Multinomial Logit Contextual Bandits

Sequential Anomaly Detection using Inverse Reinforcement Learning

Automatic Event Detection in Basketball using Hidden Markov Models with Energy based Defensive Assignment

Adaptive Pattern Matching with Reinforcement Learning for Dynamic Graphs

Learning Graph Topological Features via GAN

Efficient "Shotgun" Inference of Neural Connectivity from Highly Sub-sampled Activity Data

Graphical Model for Basketball Match Simulation

Working Papers

Thompson Sampling for Contextual Reinforcement Learning

Directed Exploration in PAC Model-free Reinforcement Learning

Unsupervised segmentation of neuroanatomy from multispectral images

Efficient "Shotgun" Inference of Neural Connectivity from Highly Sub-sampled
Activity Data