School of Informatics, Computing, and Engineering (SICE)
Intelligent Systems Engineering Colloquium Series (ISE)
Speaker: Shusen Wang, Postdoc, Univ. of California, Berkeley
Where: Luddy Hall 4063
When: Friday, March 9, 2018 at 09:30 AM
Title: Scalable Machine Learning Uncder Different Resource Constraints
Abstract: The modern big data pose big computational challenges for machine learning and data analytics. The scale of problems can easily exceed the memory and computational capacity of any single computer. How can we efficiently learn from big data?
One popular approach to large-scale problems is distributed computing on computer clusters. Distributed computing tremendously reduces the storage and local computational costs; however, the inevitable cost of communications across the network can often be the bottleneck of distributed computing. The speaker will present communication-efficient methods for empirical risk minimization (the most important optimization problem in machine learning).
However, oftentimes people do not afford computer clusters or do not have the distributed programming skills. Randomized approximate algorithms are good choices because they can produce approximate solutions using very limited resources. The speaker will briefly introduce his works on the randomized regression, clustering, dimensionality reduction, and kernel methods.
Biography: Shusen Wang is currently a postdoctoral scholar at Department of Statistics, UC Berkeley. He got both of his bachelor and doctoral degrees from Zhejiang University, China, in 2011 and 2016, respectively. His research interests lie in the domains of machine learning, randomized linear algebra, numerical optimization, and distributed computing. His focus is on computational methods for large-scale machine learning and data analytics. During his Ph.D. study, he has won "Microsoft Research Asia Fellowship" and "Baidu Scholarship", which were the highest scholarships in China.