Research

Overview

My research at National Yang Ming Chiao Tung University (NYCU) focuses on Reinforcement Learning and Restless Multi-Armed Bandits (RMABs), particularly in dynamic and non-stationary environments.

Under the supervision of Prof. Stefano Rini (Institute of Communications Engineering) and Prof. Yu-Chih Huang (Department of Electrical and Computer Engineering), I proposed the Piecewise Stationary Restless Multi-Armed Bandit (PSRMAB) framework — an extension of RMABs that captures environment changes by segmenting time into stationary periods.

The study introduces a low-complexity adaptive algorithm that combines change detection with a restless bandit framework, enabling efficient adaptation without direct state observation.

This work has evolved through multiple stages — a poster presentation, a master's thesis, and a forthcoming conference paper — representing the same line of research refined over time.

Research Focus

Reinforcement Learning Multi-armed Bandit Markov Decision Processes Restless Bandits Change Detection Mathematical Analysis Resource Allocation

An Active Approach for Piecewise Stationary Restless Bandit Problem

Restless Bandit Change Detection Resource Allocation

Advisor: Prof. Yu-Chih Huan ; Collaborators:Prof. Yu-Pin Hsu, Prof. Ping-Chun Hsieh

Abstract: Proposed a novel Piecewise Stationary Restless Multi-Armed Bandit (PSRMAB) model and developed a low-complexity adaptive algorithm for resource allocation in non-stationary communication systems. The method addresses change detection under dynamic environments while maintaining computational efficiency and theoretical performance guarantees.

Poster: Chia-Chun Lin*, Kuan-Ta Li, Ping-Chun Hsieh, Yu-Pin Hsu, and Yu-Chih Huang, “Design and Analysis of Non-Stationary Restless Multi-Armed Bandits,” ICS 2024.