For further information, please contact the General chair: Pierre-Emmanuel Gaillardon
NEC Corp., Japan
SX-Aurora Tsubasa, HPC meets machine learning
A new vector computer SX-Aurora TSUBASA is released. Traditional vector computers have been utilized mainly for HPC. In this talk, I will introduce its application to AI / machine learning. Vector computers are faster than general purpose CPUs because they load, calculate, and store many elements of data (e.g. 256) at a time in parallel, which is suitable for machine learning workload. In addition, SX-Aurora TSUBASA provides very high memory bandwidth (1.22 TB/s) to attain high efficiency by feeding enough data to ALU and FPU. To exploit this characteristics, we targeted machine learning algorithms that require high memory bandwidth: statistical machine learning and multilayer perceptron (a kind of deep learning).
Yuichi Nakamura received his B.E. degree in information engineering and M.E. degree in electrical engineering from the Tokyo Institute of Technology in 1986 and 1988, respectively. He received his PhD. from the Graduate School of Information, Production and Systems, Waseda University, in 2007. He joined NEC Corp. in 1988 and he is currently he is currently a vice president at NEC Corp. He is also a guest professor of National Institute of Informatics. He has more than 25 years of professional experience in electronic design automation, signal processing, network on chip.