Control Variate Approximation for DNN Accelerators

In this work, we introduce a control variate approximation technique for low error approximate Deep Neural Network (DNN) accelerators. The control variate technique is used in Monte Carlo methods to achieve variance reduction. Our approach significantly decreases the induced error due to approximate...

Full description

Bibliographic Details

Published in:	2021 58th ACM/IEEE Design Automation Conference (DAC) pp. 481 - 486
Main Authors:	Zervakis, Georgios, Spantidi, Ourania, Anagnostopoulos, Iraklis, Amrouch, Hussam, Henkel, Jorg
Format:	Conference Proceeding
Language:	English
Published:	IEEE 05.12.2021
Subjects:	Approximate Computing Arithmetic Circuits Control Variate Convolution Deep learning Deep Neural Networks Design automation Low Power MAC Array Mathematical analysis Monte Carlo methods
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!