Low-Power Hardware Implementation for Parametric Rectified Linear Unit Function

被引：1

作者：

Wu, Yu-Hsuan ^{[1
]}

Lin, Wei-Hung ^{[1
]}

Huang, Shih-Hsu ^{[1
]}

机构：

[1] Chung Yuan Christian Univ, Dept Elect Engn, Taoyuan, Taiwan

来源：

2020 IEEE INTERNATIONAL CONFERENCE ON CONSUMER ELECTRONICS - TAIWAN (ICCE-TAIWAN) | 2020年

关键词：

Activation Function; Approximate Multiplier; Digital Circuit; Logic Design; Low Power;

D O I：

10.1109/icce-taiwan49838.2020.9258135

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

PReLU (Parametric ReLU) function is an advanced ReLU (Rectified Linear Unit) for improving the accuracy of the neural network However, up to now, there is no special attention to the hardware design of PReLU function. Based on the characteristics of PReLU function, in this paper, we propose a dedicated multiplier for PReLU function. With this dedicated multiplier, we develop a low-power PReLU hardware implementation. Experimental results show that the proposed approach can significantly reduce power consumption with small accuracy loss.

引用

页数：2