A ReRAM-Based Computing-in-Memory Convolutional-Macro With Customized 2T2R Bit-Cell for AIoT Chip IP Applications
| Autor: | Feng Zhang, Jianfeng Gao, Xinghua Wang, Yiming Yang, Tian Wang, Yongpan Liu, Fei Tan, Liran Li, Yiming Wang |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Bit cell
Artificial neural network Computer science 020208 electrical & electronic engineering 02 engineering and technology 010501 environmental sciences Chip 01 natural sciences Convolutional neural network Computational science symbols.namesake 0202 electrical engineering electronic engineering information engineering symbols Electrical and Electronic Engineering Macro Edge computing MNIST database 0105 earth and related environmental sciences Von Neumann architecture |
| Zdroj: | IEEE Transactions on Circuits and Systems II: Express Briefs. 67:1534-1538 |
| ISSN: | 1558-3791 1549-7747 |
| DOI: | 10.1109/tcsii.2020.3013336 |
| Popis: | To reduce the energy-consuming and time latency incurred by Von Neumann architecture, this brief developed a complete computing-in-memory (CIM) convolutional macro based on ReRAM array for the convolutional layers of a LeNet-like convolutional neural network (CNN). We binarized the input layer and the first convolutional layer to get higher accuracy. The proposed ReRAM-CIM convolutional macro is suitable as an IP core for any binarized neural networks’ convolutional layers. This brief customized a bit-cell consisting of 2T2R ReRAM cells, regarded ${9 \times 8}$ bit-cells as one unit to achieve high hardware compute accuracy, great read/compute speed, and low power consuming. The ReRAM-CIM convolutional macro achieved 50 ns product-sum computing time for one complete convolutional operation in a convolutional layer in the customized CNN, with an accuracy of 96.96% on MNIST database and a peak energy efficiency of 58.82 TOPS/W. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|