Highly Efficient Neuromorphic Computing Systems with Emerging Nonvolatile Memories

Bonan Yan
Ph.D. Candidate in the Department of Electrical and Computer Engineering
Duke University
Webex Meeting: 736 911 002 Password: ECSE
Mon, March 16, 2020 at 4:00 PM

The renaissance of artificial intelligence highlights the tremendous need for computational power as well as higher computing efficiency in both high performance computing and embedded applications. To meet this demand, computing-in-memory (CIM) circumvents the von Neumann bottleneck by integrating computation and memory in the same place with reduced data traffic. CIM processors can be efficiently realized with emerging nonvolatile memories, e.g. memristor crossbar arrays. However, the design overhead of analog/digital interface circuits and the non-ideal properties of memristor devices bring difficulties into large-scale implementation. This talk will start with the architecture of RRAM-based spiking neuromorphic computing systems (NCS), followed by the design challenges and our solutions on implementing such NCS using realistic memristive devices. Designs based on these new architectures have been prototyped and demonstrated in labs to enable both signal processing and neural networks applications with orders of magnitude energy-speed improvement over the CMOS counterparts.

Webex Access:

Meeting number (access code): 736 911 002
Meeting password: ECSE
 
Join meeting

 

Join by phone
Tap to call in from a mobile device (attendees only)
+1-415-655-0001 US Toll
Bonan Yan

Bonan Yan is a Ph.D. candidate supervised by Prof. Hai “Helen” Li and Prof. Yiran Chen at the Department of Electrical and Computer Engineering, Duke University. His research interests are application-specific computing schemes with emerging nonvolatile memories. His expertise is on the VLSI design of controllers and interface circuits for resistive memory devices, e.g. MRAM and RRAM. He has developed a series of monolithically integrated RRAM-CMOS mixed-signal in-memory computing engines for machine learning hardware acceleration since 2015. Before joining Duke University, he received B.S. degree from Beihang University, China in 2014 and M.S. degree from University of Pittsburgh, USA in 2017. He designed the symposium logo for ISVLSI, 2016 and served as the web chair of ICCAD HALO workshop, 2019.