Enoch Hyunwook Kang

News

> September 1, 2025: Best AI papers explained - Apple Podcasts hit 300 subscribers with 100+ daily listeners!

> August 26, 2025: New arXiv preprint "Stability and generalization for Bellman residuals" [arXiv 2508.18741]  is out! This paper proves the first O(1/n) statistical convergence guarantee for gradient-based methods for Offline RL/IRL/Dynamic discrete choice models.

> July 15, 2025: My tutorial lecture at Econometric Society Summer School in Dynamic Structural Econometrics 2025 with John Rust is out!  (YouTube Link) 

Working papers

Machine Learning for structural estimation and pricing

"Empirical risk minimization for Inverse RL and Dynamic Discrete Choice models" by Enoch H. Kang*, Lalit Jain, and Hema Yoganarasimhan, Economics and Computation 2025, [arXiv 2502.14131]

• Invited tutorial talk (John Rust's session, guest lecturer), Econometric Society Summer School in Dynamic Structural Econometrics 2025 (YouTube Link) (Slides)

• The 2025 World Congress of the Econometric Society (ESWC 2025)

• UBC econometric lunch seminar (Planned)

Machine learning for Marketing

Bayesian optimization in language spaces: an eval-efficient AI self-improvement framework, working paper

In Silico Experimentation, working manuscript

AI agents

"Reasonably reasoning AI agents avoid game-theoretic failures in zero-shot, provably", working manuscript

Reinforcement learning theory

"Stability and generalization for Bellman residuals", Enoch H. Kang and Kyoungseok Jang, [arXiv 2508.18741] 

"Fast globally convergent gradient-based offline reinforcement learning", working manuscript 

Publications

Personalized recommendation and experimentation

"Recommender system as an exploration coordinator: a bounded O(1) regret algorithm for large platforms", Enoch H Kang, P. R. Kumar,  ArXiv Link

Preliminary versions of this work have appeared in:

• ICML 2022 Adaptive Experimentation and Active Learning in the Real World (ReALML) workshop 

• RecSys 2022 Causality, Counterfactuals, Sequential Decision-Making & Reinforcement Learning workshop (Selected as the Long Oral presentation)

"Near-equivalence between o(log n) regret and delay robustness in interactive decision-making systems", Enoch H. Kang* and P. R. Kumar, NeurIPS 2024

• NeurIPS 2023 Adaptive Experimentation and Active Learning in the Real World (ReALML) workshop 

Reinforcement Learning with Graph Neural Networks 

1. Enoch H. Kang, Taehwan Kwon, James R. Morrison, and Jinkyoo Park, "Learning NP-Hard Multi-Agent Assignment Planning using GNN: Inference on a Random Graph and Provable Auction-Fitted Q-learning" [Arxiv, submitted June 2019], Latest version Link, NeurIPS 2022

• ICML 2020 GRL+: novel applications, best paper runner up (Oral presentation link)

• Informs Annual Meeting 2020 (Contributed Long Oral talk)

Reviewer Services

Game theory and auctions

Reviewer, IEEE Robotics and Automation Letters (RA-L)

Reviewer, IEEE International Conference on Automation Science and Engineering (CASE)

Machine learning

Reviewer, International Conference on Representation Learning (ICLR)

Reviewer, Neural Information Processing Systems (NeurIPS)

Reviewer, International Conference on Machine Learning (ICML)

Reviewer, International Conference on Artificial Intelligence and Statistics (AISTATS)

Reviewer, IEEE Transactions on Automation Sciences and Engineering (T-ASE)

Economics and Econometrics

Oxford Research Encyclopedia of Economics and Finance

Marketing

Marketing Science

(Very short) Research Summary

There is a strong parallel between the mental models used in research and those in venture investment. My choice of research problems aligns closely with the high-risk, high-return mindset of pre-seed or seed-stage startup investing. Mike Maples Jr. of floodgate is my favorite VC - Pattern Breakers - Podcast - Apple Podcasts