[Colloquium] CS Distinguished Lecture today at 3 pm, Ry. 251 - Garth Gibson

[Sandra Wallace]

UNIVERSITY OF CHICAGO
DEPARTMENT OF COMPUTER SCIENCE
DISTINGUISHED LECTURE SERIES

http://cs.uchicago.edu/page/distinguished-lecture-series


When:     Thursday, April 16, 2015, 3:00 pm 

Where:    Ryerson 251

Who:       Garth Gibson (Carnegie Mellon University)

Title:       Bounded staleness in distributed machine learning systems: getting the right answer sooner  

Abstract:
Big data, data science and machine learning are a growing application space for large scale computing.  The idea, nicely summarized as “The Unreasonable Effectiveness of Data” by Google’s Halevy, Norvig and Pereira in 2009, is that training and testing your ideas with as much data as possible, and exploring new data for patterns you do not already know, can give you a significant and sometimes overwhelming competitive advantage.  This applies to diverse disciplines including targeted internet advertising, grocery store management, automated document translation, genetic correlation with disease and all sorts of academic research.  The core of these applications is often an iterative numeric computation, a workload that does not well match the strengths of large scale data processing tools like MapReduce.  The computing systems research community has taken to working with machine learning (ML) researchers on distributed systems platforms that optimize the execution of iterative numeric computation.  ML Systems research is perhaps most distinguished from other systems research by its frequent tolerance for data inconsistency and small numeric errors, because the underlying numeric iteration naturally compensates for small variation and error.  This talk will discuss a systematic approach, bounded staleness, for exploiting this error tolerance to reduce synchronization stalls and more fully utilize available hardware.  Bounded staleness offers theoretical convergence assurances and abundant opportunities for hiding communications latency without blocking computation.  We introduce a concept of message importance and use this to differentially schedule communication based on the expected impact on computation convergence.  Then we push this scheduling approach “up the stack” from ordering communication to ordering computation and communication, effectively abandoning the simple notion of a complete iteration in an iterative numeric computation.

Bio:
Garth Gibson is a professor of computer science at Carnegie Mellon University, the co-founder and chief scientist at Panasas Inc, and a Fellow of the ACM and the IEEE. He holds a MS and PhD from the University of California at Berkeley and a B.Math from the University of Waterloo in Canada. His research on Redundant Arrays of Inexpensive Disks (RAID) has been awarded the 1998 SIGMOD Test of Time Award, the 1999 Allan Newell Award for Research Excellence, the 1999 IEEE Reynold B. Johnson Information Storage Award for outstanding contributions in the field of information storage, the 2005 J. Wesley Graham Medal in Computing and Innovation from the University of Waterloo, entrance into the ACM SIGOPS Hall of Fame in 2011, and the 2012 IFIP WG10.4 Jean-Claude Laprie Award in Dependable Computing. Gibson founded CMU's Parallel Data Laboratory in 1992 and was a founding member of the Technical Council of the Storage Networking Industry Association and the USENIX Conference on File and Storage Technology Steering Committee.  At Panasas Gibson led the development of high-performance, scalable, parallel, file-system appliances in use in High-Performance Computing in national labs, academic clouds, energy research, engineered manufacturing and life sciences. Gibson instigated standardizing key features of parallel file systems in NFSv4.1 (parallel NFS), now adopted and deployed in Linux. His 1995 Network-Attached Secure Disks (NASD) led to the ANSI T10 (SCSI) Object Storage Device (OSD) command set. His students have gone on to co-author influential systems such as the Google File System and BigTable, and to lead the technology development of influential products such as EMC's Data Domain. His collaboration with Los Alamos National Laboratory (LANL) led to the Parallel Log-structured File System (PLFS) in 2009 and SC’14’s Best Paper on scaling file system metadata. Recently Gibson has been developing educational courses (Advanced Cloud Computing) and programs (Masters in Computational Data Science) for cloud computing, big data and data science and is an investigator in the Intel Science and Technology Center for Cloud Computing.  His current students are collaborating with the Machine Learning Department at CMU to develop radically asynchronous massive model solvers that trade bounded error for convergence speed in the search for a good predictor of the hidden parameters governing the interpretation of big data sets.


Host:  Haryadi Gunawi



Reception will follow at 4 pm in Ry 255.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20150416/96df1bc8/attachment.htm