Cheng-Kuang Wu

Current AI systems, which are largely based on deep neural networks, do not have the inherent ability to learn continuously in the dynamic world we live in. This PhD project investigates how deep neural networks can learn continuously from evolving data streams. It aims to develop continual learning methods that allows deep neural networks to adapt efficiently to new tasks while retaining previously acquired core knowledge. A key application focus is human-AI co-improvement, where AI systems learn from ongoing human feedback while also helping users refine decisions, skills, and understanding over time. By advancing continual learning in deep neural networks, the project seeks to contribute to more adaptive AI systems.