Francis Y. Yan

ABOUT ME

In the fall of 2020, I will join Microsoft Research Redmond as a Senior Researcher in the Mobility and Networking Research Group. My research focuses on practical machine learning (ML) for networking, seeking to create ML algorithms that are deployable on real-world networked systems and build large-scale platforms opened to the research community for training and validating new algorithms. If you are excited about applying ML to systems and networking research, I will be happy to talk to you.

I completed my Ph.D. in computer science at Stanford University in 2020, advised by Keith Winstein and Philip Levis. My dissertation was on the development of platforms and algorithms to achieve practical machine learning for sequential decision problems on the Internet, with a focus on video streaming and congestion control. The two papers composing my dissertation received the Community Award at USENIX NSDI 2020 and the Best Paper Award at USENIX ATC 2018 respectively.

Before that, I graduated from Tsinghua University in 2015, where I received a B.S. in computer science from Yao Class (founded by Turing Award laureate Andrew Chi-Chih Yao) and a B.A. in economics from the School of Economics and Management. In 2014, I spent a semester studying at MIT as an exchange student and took my first machine learning class there.

RECENT RESEARCH

Learning in situ: a randomized experiment in video streaming

Francis Y. Yan, Hudson Ayers, Chenzhi Zhu, Sadjad Fouladi, James Hong, Keyi Zhang, Philip Levis, Keith Winstein

Community Award (“for the best paper whose code and/or data set is made publicly available”)

USENIX Symposium on Networked Systems Design and Implementation (NSDI), February 2020

links website • talk • slides • code • wikipedia • coverage

We built Puffer, a free, publicly accessible website that live-streams television channels and operates as a randomized experiment of adaptive bitrate (ABR) algorithms. As of June 2020, Puffer has attracted 120,000 real users and streamed 60 years of video across the Internet. Using Puffer, we developed an ML-based ABR algorithm, Fugu, that robustly outperformed existing schemes by learning in situ, on real data from its actual deployment environment.

Pantheon: the training ground for Internet congestion-control research

Francis Y. Yan, Jestin Ma, Greg D. Hill, Deepti Raghavan, Riad S. Wahby, Philip Levis, Keith Winstein

Best Paper Award

USENIX Annual Technical Conference (ATC), July 2018

links website • talk • slides • code

Pantheon is a community “training ground” for Internet congestion-control research and has assisted four schemes from other research groups in publishing at NSDI 2018 (Copa and Vivace), ICML 2019 (Aurora), and SIGCOMM 2020 (to appear). It also enabled our own ML-based congestion-control algorithm, Indigo, which was trained to imitate expert congestion-control algorithms we created in emulation and achieved good performance over the real Internet.

SERVICE

Reviewer, IEEE/ACM Transactions on Networking (2019–)
Reviewer, ACM SIGCOMM Computer Communication Review (2019–)
Reviewer, Computer Communications (2019–)

PERSONAL

My current favorite sport is table tennis (skilled), but I also enjoy skiing (blue trail), ice/roller skating (intermediate), playing badminton (average) and tennis (beginner), swimming (novice), kayaking (amateur), etc. When I have no other choice, I'm fine with hiking too, preferably in scenic national parks though.

Life-changing tools (software and hardware):