Conference season is now in full swing, and Yahoo Labs is kicking things off with two best paper award wins.
At the USENIX Annual Technology Conference (ATC) in Boston, MA, Ben Reed, Flavio Junqueira, Patrick Hunt and Mahadev Konar’s paper, “ZooKeeper: Wait-free Coordination for Internet-scale Systems,” won the best paper award. Reed describes ZooKeeper as a centralized service for maintaining configuration information, naming, providing distributed synchronization, and providing group services. All of these kinds of services are used in some form or another by distributed applications. Each time they are implemented there is a lot of work that goes into fixing the bugs and race conditions that are inevitable. Because of the difficulty of implementing these kinds of services, applications initially usually skimp on them, which makes them brittle in the presence of change and difficult to manage. Even when done correctly, different implementations of these services lead to management complexity when the applications are deployed.
Zookeeper is an Apache open source project that is used by Yahoo as well as by external organizations.
At the International Conference on Machine Learning (ICML), “Hilbert Space Embeddings of Hidden Markov Models” won the best paper award. It was co-authored by Le Song (Carnegie Mellon University), Byron Boots (Carnegie Mellon University), Sajid Siddiqi (Google), Geoffrey Gordon (Carnegie Mellon University) and Alex Smola (Yahoo Labs). In the paper, the team shows that Hidden Markov Models (HMMs) are important tools for modeling sequence data. However, they are restricted to discrete latent states, and are largely restricted to Gaussian and discrete observations. Learning algorithms for HMMs have predominantly relied on local search heuristics, with the exception of spectral methods. The authors propose a non-parametric HMM that extends traditional HMMs to structured and non-Gaussian continuous distributions. Furthermore, they derive a local-minimum-free kernel spectral algorithm for learning these HMMs. They apply their method to robot vision data, slot car inertial sensor data and audio event classification data, and show that in these applications, embedded HMMs exceed the previous state-of-the-art performance.
About USENIX ATC
USENIX ATC brings together leading systems researchers for cutting-edge systems research and opportunities to gain insight into a variety of must-know topics, including virtualization, system administration, cloud computing, security, and networking.
ICML is the leading international machine learning conference, attracting annually some 500 participants from all over the world. ICML is supported by the International Machine Learning Society (IMLS).