Work-in-Process: Error-Compensation-Based Energy-Efficient MAC Unit for CNNs
Created by W.Langdon from
gp-bibliography.bib Revision:1.8098
- @InProceedings{Xu:2023:CASES,
-
author = "Xingyu Xu and Qingwen Wei and Yang Zhang and
Hao Cai and Bo Liu",
-
booktitle = "2023 International Conference on Compilers,
Architecture, and Synthesis for Embedded Systems
(CASES)",
-
title = "Work-in-Process: Error-Compensation-Based
Energy-Efficient {MAC} Unit for {CNNs}",
-
year = "2023",
-
pages = "3--4",
-
abstract = "Approximate circuits sacrifice accuracy in exchange
for energy efficiency and have been widely used in
hardware deployment of neural networks (NNs). Since
convolution accounts for most of the power consumption
in NNs, it is necessary to design an approximate
multiplication and accumulation (MAC) unit which
improve the energy efficiency of hardware with
ignorable accuracy loss. an error-compensation-based
energy-efficient MAC unit is proposed in which
approximate multipliers are designed by Boolean matrix
factorization and approximate adders are generated by
Cartesian genetic programming. The proposed MAC unit is
conducted on CIFAR10 using ResNet-18, where PDP is
reduced by 58.8percent with an accuracy loss of
0.81percent.",
-
keywords = "genetic algorithms, genetic programming, Power demand,
Embedded systems, Computer aided software engineering,
Convolution, Artificial neural networks, ANN, Computer
architecture, Approximate computing, Energy efficiency,
Error compensation",
-
URL = "http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=10316208",
-
ISSN = "2643-1726",
-
month = sep,
-
notes = "Also known as \cite{10316208}",
- }
Genetic Programming entries for
Xingyu Xu
Qingwen Wei
Yang Zhang
Hao Cai
Bo Liu
Citations