学术报告第201513期

报告题目：Memristive  circuits and systems for mimicking human brains

报 告 人：Professor Kyeong-Sik MIN  (Kookmin University, Seoul,  Korea)

报告时间：2015年4月17日（周五）上午9：00（临时因故延至上午10：00开始）

报告地点：南一楼中311室

邀 请 方：“多谱信息处理技术”国家级重点实验室

Abstract: 

   For last more than five decades, computing technology has been advancing further  and further according to Moore’s law. The device scaling scenario of the device  integration doubled every one and half year has been acting as a rule of law in  all the IT areas, since Moore’s first prediction in 1965. As the device  dimensions get closer to the physical limit, however, another driving force that  can make it possible the current phase of improvement in power efficiency and  computing performance should be considered now more than ever.

   As one way of searching for new driving force, neuromorphic studies were started  to mimic human brains, because human brains are known 10X more energy-efficient,  compared to the state-of-the-art Silicon-based systems. In terms of device and  process technologies, many new devices and materials are emerging now for  preparing some days of Moore’s law ending.

   Among many new approaches to brain-like neuromorphic computing systems,  recently, memristor-based neuromorphic circuits have  been regarded one of strong candidates of promising neuromorphic architecture,  because they can be implemented by Silicon-compatible technologies with high  density and low cost. Moreover, memristors can emulate synaptic plasticity with  high energy efficiency and high computing performance. Thus they can be used in  realizing future neuromorphic circuits and systems.

   In this presentation, I would introduce memristor’s physics and explain how to  read and write memristors with mentioning the sneak-path leakage problem. And, I  would talk about the memristor crossbar architecture that can be used in  neuromorphic applications such as image and speech recognition. After that, I  would mention Cellular Neural Network circuits which are realized with  memristors and discuss about the operation, applications, possibility, etc of  the memristor-based Cellular Neural Networks.

Speaker’s  Short Bio: 

   Kyeong-Sik Min received the B.S. degree in Electronic and Computer Engineering  from Korea University, Seoul, Korea, in 1991, and the M.S.E.E. and Ph. D.  degrees in Electrical Engineering from Korea Advanced Institute of Science and  Technology (KAIST), Daejeon, Korea, in 1993 and 1997, respectively. In 1997, he  joined Hynix Semiconductor Inc., where he was engaged in developing low-power and high-speed DRAM circuits. From 2001 to 2002, he was a  research associate at University of Tokyo, Tokyo, Japan, where he designed  low-leakage memories and logic circuits. In September 2002, he joined the  faculty of Kookmin University, Seoul, Korea, where he is currently a Professor  at School of Electrical Engineering. At Kookmin Univ., he is working in  low-power VLSI design, new emerging memory modeling and circuit design, and  developing power management IC for energy harvesting. Prof. Min served on the  technical program committees of Asian Solid-State Circuits Conference (A-SSCC),  Korean Conference on Semiconductors (KCS), International SoC Design Conference  (ISOCC), etc. He is a member of IEEE, IEICE, and IEEK.