Energy-efficient machine learning accelerator for binary neural networks

Mao，Wei; Xiao，Zhihua; Xu，Peng; Ren，Hongwei; Liu，Dingbang; Zhao，Shirui; An，Fengwei; Yu，Hao

10.1145/3386263.3407582

2020-09-07

Proceedings of the ACM Great Lakes Symposium on VLSI, GLSVLSI

77-82

Binary neural network (BNN) has shown great potential to be implemented with power efficiency and high throughput. Compared with convolutional neural network (CNN), BNN is trained with binary constrained weights and activations, which are more suitable for edge devices. In this paper, we introduce the BNN characteristics, basic operations and the binarized-network optimization methods. Then we summarize several accelerator designs for BNN hardware implementation by using three mainstream structures, i.e., ReRAM-based crossbar, FPGA and ASIC. Based on the BNN characteristics and hardware custom designs, all these methods achieve massively parallelized computations and highly pipelined data flow to enhance its latency and throughput performance. Besides, the intermediate data with the binary format are stored and processed on chip by the computing-in-memory (CIM) architecture to reduce the off-chip communication costs, including power and latency.

Accelerator Binary neural networks (BNN) Computing in memory (CIM) Convolution neural networks (CNN) ReRAM

第一 ; 通讯

英语

EI

20203909230009

Convolution ; Computing power ; Machine learning ; Computer aided design ; Convolutional neural networks ; Energy efficiency

Energy Conservation:525.2 ; Information Theory and Signal Processing:716.1 ; Data Storage, Equipment and Techniques:722.1 ; Computer Peripheral Equipment:722.2 ; Digital Computers and Systems:722.4 ; Computer Software, Data Handling and Applications:723 ; Artificial Intelligence:723.4 ; Computer Applications:723.5

2-s2.0-85091300397

Scopus

School of Microelectronics Engineering,Research Center of Integrated Circuits for Next-Generation Communications,Ministry of Education Southern University of Science and Technology,Shenzhen,China

Mao，Wei,Xiao，Zhihua,Xu，Peng,et al. Energy-efficient machine learning accelerator for binary neural networks[C],2020:77-82.