Research Seminar - June 18, 2004

Design and Implementation of Distributed Collaborative Filtering

Time: 11am Friday June 18

Computer Science & Software Engineering

Seminar Room 1.24

Speaker:
Mr Han Peng

Abstract:

Collaborative filtering (CF) has proved to be one of the most effective recommendation techniques. However, as the calculation complexity increases quickly both in time and space when the records in the user database accumulate, traditional centralized CF algorithms fail to scale. In this seminar, I will first provide a brief introduction on the background and current research trends of CF techniques and then present a distributed CF algorithm called PipeCF. Using PipeCF, we can perform the user database management and prediction task in a decentralized way. Two novel approaches will also be introduced, i.e. significance refinement (SR) and unanimous amplification (UA), to further improve the scalability and prediction accuracy of PipeCF.

About the Speaker:

Mr. Han Peng is a Ph.D. candidate in the Department of Computer Science and Engineering at Shanghai Jiaotong University, China. Mr. Han’s research area involves the research and application of Artificial Intelligence techniques in e-learning environment. His current work focuses on e-learning user modeling. The main purpose is to find suitable mathematical models to describe the status and behaviors of e-learners in a formal and structured way, and design efficient algorithms to learn this model from observed data which includes log-files, user profiles and so on. The learned model can then be further used to provide personalized services to e-learners and organize e-learners of similar interests into communities to share resources and experiences. His other research interests include Information Retrieval, Data mining and Semantic Web.