Short floating-point representation for convolutional neural network inference
Open Access
- 1 January 2019
- journal article
- research article
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in IEICE Electronics Express
- Vol. 16 (2) , 20180909
Abstract
Convolutional neural networks (CNNs) are being widely used in computer vision tasks, and there have been many efforts to implement CNNs in ASIC or FPGA for power-hungry environments. Instead of the previous common representation, the fixed-point representation, this letter proposes a short floating-point representation for CNNs. The short floating-point representation is based on the normal floating-point representation, but has much less width and does not have complex cases like Not-a-Number and infinity cases. The short floating-point representation, contrary to the common belief, can produce a low-complexity computation logic because the operands of the multiplier logic can be shortened by the exponent concept of the floating-point representation. The exponent can also reduce the total length to reduce the SRAM area. The experimental results show that the short floating-point representation with 8-bit total width achieves less-than-1-percentage-point degradation without the aid of retraining in the top-5 accuracy on very deep CNNs of up to 152 layers and gives more than a 60% area reduction in the ASIC implementation.Keywords
This publication has 5 references indexed in Scilit:
- A Resource-Limited Hardware Accelerator for Convolutional Neural Networks in Embedded Vision ApplicationsIEEE Transactions on Circuits and Systems II: Express Briefs, 2017
- Cambricon: An Instruction Set Architecture for Neural Networks2016
- Eyeriss: A Spatial Architecture for Energy-Efficient Dataflow for Convolutional Neural Networks2016
- ImageNet Large Scale Visual Recognition ChallengeInternational Journal of Computer Vision, 2015
- DianNaoACM SIGPLAN Notices, 2014