Menu
O Abstract:
In the era of edge AI, where computation meets real-world applications, the demand for high-performance, energy-efficient solutions has never been more pronounced. This seminar introduces an innovative approach to this challenge, introducing the fusion of cutting-edge neural network acceleration and Processing-In-Memory (PIM) technology in the design of advanced Edge AI chips. Traditional AI acceleration methods have paved the way, yet limitations in power consumption and data movement persist. Enter PIM, a revolutionary paradigm that unites processing and memory to redefine the landscape. This seminar delves into the intersection of PIM and neural networks, highlighting the pivotal role of in-memory computing in transforming Edge AI.Through comprehensive exploration, we delve into the intricacies of PIM-based neural network acceleration. We decipher the design considerations, from power-efficient hardware techniques to optimal memory hierarchies, and elucidate the synergistic possibilities when PIM converges with conventional AI accelerators. Real-world applications come to life, illustrating the prowess of PIM-powered neural networks across IoT, healthcare, autonomy, and beyond.
Dr. ChoongHyun Lee holds a Master’s degree and a Ph.D. in materials engineering from the University of Tokyo, Japan, achieved in 2010 and 2013 respectively. His academic pursuits were centered around advancing the frontiers of materials engineering, with a specific focus on interface control of Ge gate stack for high-performance logic applications. Following his doctoral achievements, he was selected for a postdoctoral fellowship by the Japan Society for the Promotion of Science (JSPS). From April 2013 to March 2015, he thrived as a dedicated full-time researcher, contributing to cutting-edge research initiatives during his tenure. In March 2015, Dr. Lee embarked on a new chapter by joining IBM in their research division. Here, his expertise found expression in pioneering work on new materials and processes that underpin the evolution of advanced node semiconductor technology. His contributions significantly impacted the development of 10/7/5 nm logic transistor nodes and beyond, instrumental in shaping the future of high-performance and low-power technologies. In July 2019, he assumed the role of Professor in the College of Information Science and Electronic Engineering at Zhejiang University. His academic pursuits expanded to encompass advanced CMOS devices utilizing high-mobility channel materials, including SiGe, Ge, and III-Vs. Additionally, his expertise extended to non-volatile memories, where he ventured into designing robust synaptic devices tailored for compute-in-memory paradigms. In September 2021, Dr. Lee took the helm as the visionary founder of Pebble-Square. This pioneering initiative is dedicated to the realization of edge AI chips boasting high performance and ultra-low power consumption through the integration of cutting-edge processing-in-memory technologies.
–
Copyright ⓒ 2015 KAIST Electrical Engineering. All rights reserved. Made by PRESSCAT
Copyright ⓒ 2015 KAIST Electrical Engineering. All rights reserved. Made by PRESSCAT
Copyright ⓒ 2015 KAIST Electrical
Engineering. All rights reserved.
Made by PRESSCAT
