摘要:前年的ISSCC 2017大会上,意法半导体做过一个题为“针对智能嵌入式系统、采用28nm FD-SOI工艺的2.9TOPS/W深度卷积神经网络SoC”(A 2.9 TOPS/W Deep Convolutional Neural Network SoC in FD-SOI 28nm for Intelligent Embedded Systems)的演讲。演讲中的这颗SoC应用,当然如其名就是卷积神经网络(CNN),而宣传重点在高效的存储层级,适用于不同深度卷积神经网络拓扑的弹性,输入输出能力,还有与主流的深度学习工具做整合。
注:因版权方要求,不能公开全文,如需全文,请咨询杂志社
热门期刊服务
Journal of Hydrodynamics Journal of Huazhong University of Science and Technology The Journal of China Universities of Posts and Telecommunications Journal of Genetics and Genomics Journal of Iron and Steel Research Journal of Systems Science and Complexity World Journal of Gastroenterology Journal of Computer Science and Technology Applied Mathematics:A Journal of Chinese Universities International Journal of Minerals Metallurgy and Materials International Journal of Mining Science and Technology Journal of Systems Science and Systems Engineering