The Department of Statistics and Actuarial Science presents the Fall 2026 Data Science Bootcamps, sponsored by Aegon Transamerica. Each week our distinguished presenters will present on topics including data analytics programming, business communication, and observational data analysis.

This week's Data Science Bootcamp presentation:

Dr. Congrui Yi, Machine Learning Engineer, Pinterest Labs

"Introduction to Multi-Armed Bandits"

Multi-armed bandit (MAB) is a simple but powerful modeling framework for making sequential decisions given distributional uncertainty and dynamic environments. Given the right to choose an action at a time and collect corresponding rewards, MAB algorithms aim to maximize cumulative rewards by balancing two competing objectives: exploration of new or less chosen actions, and exploitation of ones shown to be rewarding. This lecture introduces basic concepts including action, reward, regret, exploration-exploitation trade-off, and most common types of MAB algorithms, such as epsilon-greedy, Upper Confidence Bound (UCB), Thompson Sampling (TS), and Boltzmann Exploration. Additionally, it will include a hands-on Python coding exercise and a discussion of real-world applications.