时 间：2021年10月15日15 : 00 - 17 : 00
地 点： 新光电大楼C117学术报告厅
报告人： 尹首一 教授， 清华大学
邀请人： 刘冬生 教授 邹雪城 教授
报告摘要：
人工智能近年取得了突破性进展，正在深刻改变人类的生产和生活方式。现有的通用计算平台难以实现高能效的神经网络计算，探索新型神经网络计算芯片架构成为研究热点和学科前沿，是世界各国争相发展的战略制高点。国际IT巨头都在竞相研发人工智能计算芯片。本报告综述了深度学习、神经网络及神经网络处理器设计的挑战，详细介绍了清华大学微电子所设计的Thinker系列人工智能芯片。该芯片采用可重构架构和电路技术，突破了神经网络计算和访存的瓶颈，高能效实现了人工智能计算。
Artificial intelligence has made a breakthrough in recent years and is profoundly changing human society and life style. The existing general purpose computing platforms are hard to achieve energy-efficient neural network computing. Exploring a new neural network computing chip architecture has become a research hotspot and discipline frontier, and is the strategic commanding point for the whole world. International IT giants are competing to develop artificial intelligence computing chips. This report summarizes the challenges of deep learning, neural network and neural network processor design, and discusses the opportunities of performance optimization. Then it introduces in detail the thinker series artificial intelligence processors designed by the School of Integrated Circuits in Tsinghua University. Thinker processors adopt reconfigurable architecture and circuit technology, break the bottlenecks of neural network computing and memory access, and achieve high energy efficiency.
报告人介绍：
尹首一，博士，清华大学长聘教授，集成电路学院副院长，国家杰出青年科学基金获得者。研究方向包括可重构计算、人工智能芯片设计等。近5年以第一/通讯作者在IEEE JSSC、TCAS-I等SCI期刊发表论文30余篇，在ISSCC、VLSI、ISCAS等顶级会议发表论文17篇，出版《人工智能芯片设计》专著一部，授权中国专利46项、国际专利4项。科研成果被全球40个国家的300余研究机构（包括MIT、Stanford）跟踪引用。荣获2020中国电子学会技术发明一等奖、2020中国产学研合作创新奖、2015国家技术发明二等奖、2015教育部技术发明一等奖、2015中国发明专利金奖等荣誉。
Yin Shouyi, Ph.D., Professor of Tsinghua University, Deputy Dean of the School of Integrated Circuits. Secretary-General of the Electronic Design Automation Committee of the Chinese Institute of Electronics. Research directions include reconfigurable computing, AI chip design. In the past 5 years, as the first/corresponding author, he published 30 papers in IEEE JSSC, TCAS-I and other SCI journals, published 17 papers in top conferences such as ISSCC, VLSI, ISCAS, and published a book on "Artificial Intelligence Chip Design", authorized 46 Chinese patents and 4 international patents. Scientific research results are tracked and cited by more than 300 research institutions (including MIT, Stanford, UC Berkeley) in 40 countries around the world. In the past five years, the papers have been cited 437 times by SCI. He has won the first prize of the Chinese Institute of Electronics Technology Invention(2020), the second prize of National Technology Invention(2015), the Chinese Invention Patent Gold Award(2015), the first prize of Technology Invention of the Ministry of Education(2015). In 2019, he was selected as the PI of the Beijing Academy of Artificial Intelligence with his research of AI chips. He serves as a member of the technical committee of IEEE DAC, ICCAD, DATE, ASPDAC and A-SSCC. Associate Editor of "IEEE Transactions on Circuits and System I: Regular Papers", "ACM Transactions on Reconfigurable Technology and Systems", "Integration, the VLSI Journal" and Editor Board Member of "Journal of Low Power Electronics".
