Data Driven Perception and Inference
일시 : 2019년 11월 8일(금), 11:00 ~ 12:00
장소 : 서울대학교제1공학관(301동) 104호
연사 : Tobi Delbruck, Professor of Physics and Electrical Engineering, Institute of Neuroinformatics, University of Zurich and ETH Zurich
Machine vision and audition hardware systems face a fundamental latency-power tradeoff: To get low latency you have to increase frame rate, but then you burn more power. Applications in robotics and human computer interaction demand low system latency, so if they use conventional Nyquist sampling, they always burns lots of power. The event sensors and DNN accelerators developed by our group and others overcome this tradeoff by using brain-inspired data-driven sensing and inference. These systems can always be quick to respond when needed and dissipate little power at other times. I will show results from our lab using our event sensors that demonstrate this principle. I will also discuss our convolutional and recurrent deep neural network accelerators that also exploit activation sparsity to achieve state of the art power efficiency, throughput, and latency.
Tobi Delbruck (IEEE M'99–SM'06–F'13) received the B.Sc. degree in physics from University of California in 1986 and PhD degree from California Institute of Technology in 1993. Currently, he is a Professor of Physics and Electrical Engineering with the Institute of Neuroinformatics, University of Zurich and ETH Zurich, where he has been since 1998. The Sensors group which he leads together with Prof. Shih-Chii Liu focuses on efficient neuromorphic sensory processing and deep neural network theory and hardware accelerators.
He has been awarded 9 IEEE paper awards and was named a Fellow of the IEEE Circuits and Systems Society for his work on neuromorphic sensors and processing. He likes to read, play tennis and basketball, and practice card magic on unwary subjects.
담당교수 : 서울대학교 전기·정보공학부 이 혁 재 / 문의 : 박진우(email@example.com)