Low-Power Neuromorphic Computing with Tunneling Field Effect Transistors (TFETs)
|Event Date:||April 2, 2015|
|Speaker:||Amit Ranjan Trivedi|
|Speaker Affiliation:||Georgia Institute of Technology
Department of Electrical and Computer Engineering
|Contact Name:||Prof. Byunghoo Jung
|Open To:||ACCEPTABLE FOR ECE 694
Electronic computing has advanced through low power platforms such as wearables and IoTs to ‘compute small and compute everywhere’ world. The next evolutionary task for these low power platforms is to facilitate seamless and natural interaction of computing devices with their users/ambience. Traditional computing paradigms lack efficiency in implementing tasks such as recognition/classification imperative for the above objective. Neuromorphic computing, a set of computing paradigms inspired by functioning of biological brain, is being investigated to bring such computing abilities in electronic devices. This talk focuses on power reduction in neuromorphic computing to make these suitable for low power sensors/system applications.
This talk will discuss the role of emerging computing technology, Tunneling field-effect-transistor (TFET), in low power neuromorphic computing. Unique characteristics of TFET, such as its higher gm/IDS and lower off current, is utilized in low power energy-efficient neuromorphic computing. Furthermore, beyond a simple technology replacement, this talk will show non-conventional TFET design, where such a design builds neuromorphic circuits within a single transistor. Novel computing paradigms explore these non-conventional TFETs for area/energy-efficient neuromorphic operation. At the end, the talk also discusses how neuromorphic learning principles can be applied to reduce power of conventional digital computing designs.
Amit completed his undergraduate and graduate degree from Indian Institute of Technology (IIT), Kanpur. Amit was awarded the Academic Excellence Award by IIT Kanpur for standing among the top 5% of his peers. Following this, he was a research staff member at IBM Semiconductor Research and Development Center, where he was involved in compact modeling and characterization of advanced nanometer node transistors/processes. Since, Fall 2010, Amit is pursuing Ph.D. at Georgia Institute of Technology. His research interest are in low power energy-efficient neuromorphic computing with emerging technologies, and particularly, with Tunneling field-effect-transistors. He has published more than ten journals and major conferences during his PhD studies. Amit was awarded IEEE Electron Device Society fellowship of the year, where he was one of the three recipients worldwide. Amit was a research intern at IBM T J Watson research center in summers of 2012, and Intel’s Circuit research lab in summers of 2014.