We are pleased to announce our upcoming webinar in April 2026. Dr. Kwang-Sung Jun from Pohang University of Science and Technology (POSTECH) will give a talk at 7pm (ET) on April 22 (Wednesday). Please use the link below to register for the KISS webinar. The webinar title and abstract are as follows.
Date/Time: 7pm – 8pm ET (6pm – 7pm CT; 4pm – 5pm PT) on April 22
Registration link:
https://yonsei.zoom.us/meeting/register/iRRATxrkS12uM9rvvlmeuA
If the link does not open when clicked, please copy and paste it into your browser’s address bar.
Registration is required for this meeting. After registering, you will receive a confirmation email containing information about joining the meeting.
Speaker: Dr. Kwang-Sung Jun from Pohang University of Science and Technology (POSTECH)
Title: Second-order learning in confidence bounds, contextual bandits, and regression
Abstract: Confidence sequence provides ways to characterize uncertainty in stochastic environments, which is a widely-used tool for interactive machine learning algorithms and statistical problems including A/B testing, Bayesian optimization, reinforcement learning, and offline evaluation/learning. In these problems, constructing confidence sequences that are tight and correct is crucial since it has a significant impact on the performance of downstream tasks. In this talk, I will first show how to derive one of the tightest empirical Bernstein-style confidence bounds, both theoretically and numerically. This derivation is done via the existence of regret bounds in online learning, inspired by the seminal work of Raklin & Sridharan (2017). Then, I will discuss how our confidence bound extends to unbounded nonnegative random variables with provable tightness. In offline contextual bandits, this leads to the best-known second-order bound in the literature with promising preliminary empirical results. Finally, I will turn to the $[0,1]$-valued regression problem and show how the intuition from our confidence bounds extends to a novel betting-based loss function that exhibits variance-adaptivity. I will conclude with future work including some recent LLM-related topics.
I am looking forward to seeing you all!
Best regards,
Jeong Hoon Jang
KISS Program Chair Elect
