Computer Architecture in the IoT Cognitive Computing Era
Computing has penetrated all facets of our society ranging from mobile devices connecting people through social media, personalized services such as search engines and health-care powered by cloud computing, automated cars, smarter homes, accurate and timely weather and disaster predictions, and beyond. Today, these technologies are possible because of the increasingly powerful evolution of processors in the past several decades. However, the exponential growth in processor performance has recently slowed down dramatically due to technology scaling much more slowly than it used to for the past several decades. On the other hand, with all of the power of digital computers built on silicon, we still cannot come close to the human brain on many tasks, including pattern recognition, and visual and auditory processing. And even for problems that computers are good at, for example, problems like scheduling and planning that involves remembering facts, computer hardware and software cannot come close in terms of the efficiency requirements that humans require to solve these problems. So perhaps we do not need the technology to continue scaling, but rather we require new architectures and algorithms that dramatically improve the efficiency of computers on such problems. The computing challenge is further exacerbated because today's computational models require optimizing multiple objectives, such as resiliency; security and privacy of computation, while at the same time meet the time-to-solution and energy-to-solution requirements. In this talk, I will outline forward-looking architectural methods for problems pertaining to the broad area of cognitive computing.
Omer Khan is an Associate Professor of Electrical and Computer Engineering at the University of Connecticut. Prior to joining UConn, he was a Postdoctoral Research Scientist at Massachusetts Institute of Technology. He received his Ph.D. in Electrical and Computer Engineering from the University of Massachusetts Amherst. He has more than 15 years of computer architecture research and development experience in academia and industry (Motorola/Freescale and Intel). At UConn, Omer leads the Computer Architecture Group, where he is developing cross-layer methods and system interfaces to improve the efficiency, resiliency, and security of future computing platforms. He has published over 50 papers in competitive journals and conference proceedings. He is a member of the ACM and IEEE.