Abstract
Personalized recommendation plays a key role in many successful applications such as search engines, e-commerce, and streaming platforms. With the advances of generative models such as Diffusion Model and Large Language Model, the paradigm of future personalization systems may shift from retrieval-based to generative. I will first talk about how the retrieval-based personalization system works and challenges in modeling noisy and biased user feedback data. Next, I will introduce some ongoing projects on generative personalization systems, where we explore (1) how to enable personalized image generations with Diffusion Models, and (2) how to enhance personalized learning experience in CS Education with LLM. Lastly, I will share some tips for undergraduate students who are interested in doing research.
Biography
Hongyi Wen is an Assistant Professor of Computer Science at NYU Shanghai. He received his BEng in Computer Science from Tsinghua University. He holds a PhD in Information Science from Cornell University, advised by Prof. Deborah Estrin. His research focuses on Recommender Systems, HCI and personalization in real-world applications, where he aims to develop human-centered AI for improving the long-term values of personalizations.