Te-Yuan Huang

Hi! I am currently in my first year of the Ph.D. program in CS, Stanford University.
You can visit my new personal paget at here .

I have recieved my Master's degree in 2008 from EE, National Taiwan University. In NTU EE, I was working in Network and System Laboratory, which is supervised under Prof. Polly Huang. I am also an alumni of CS, National Chaio Tung University and received my B.S. degree in 2006. My undergrad thesis was accomplished with Prof. John K. Zao and friends in Pervasive Computing Laboratory. The project was awarded by National Science Council1 Research Creativity Award in July, 2006.

I am intrested in computer networking, especially network traffic measurement, analysis, and modeling. Peer-to-peer network, artificial intelligence, network security, and online-gaming are also involved in my research interests.

Projects

User-Centric Rate Adaptations for VoIP Traffic:
Is it possible to send as much voice data as the VoIP user desires, while not causing trouble to other best effort traffic on the Internet? In this project, we are trying to design such a sending rate adaptive mechanism that is not only friendly to its VoIP user, but also friendly to other traffice on the Internet, i.e., not causing congestion on the Internet. In this project, we have designed an AIMD rate adaptation mechanism based on USI (User Satisfaction Index). Our preliminary result shows that adapting sending rate based on USI is indeed a promising alternative and could result in better voice communication for users.

Learning User Satisfaction from Conversation Patterns:
User satisfaction is such an intangible thing. However, it is a very important indicator for service providers. In this project, we are trying to find out the conversation patterns from VoIP traces by CHMM (Coupled Hidden Markov Model), a widely exploited machine learning mechanism. We then try to quantify user satisfaction from the conversation patterns learned by CHMM.

Detecting System Abnormality from User Session Time:
Users are always the first ones who discover the system met probelms. Some of them would write a report to the system administration, but MOST of them would just leave the system. Thus, to shorten the detection delay for system administrators, we got the idea to detect system abormality from the session time of users.