CompAct: on-chip com pression of act ivations for low power systolic array based CNN acceleration
This paper addresses the design of systolic array (SA) based convolutional neural network
(CNN) accelerators for mobile and embedded domains. On-and off-chip memory accesses …
(CNN) accelerators for mobile and embedded domains. On-and off-chip memory accesses …
[PDF][PDF] CompAct: On-chip Compression of Activations for Low Power Systolic Array Based CNN Acceleration
JJUN ZHANG, P RAJ, S ZARAR, A AMBARDEKAR… - 2019 - researchgate.net
Deep neural networks (DNN) provide best-in-class accuracy for a range of machine learning
tasks including text, speech, image and video [18, 21, 32]. DNNs are composed of multiple …
tasks including text, speech, image and video [18, 21, 32]. DNNs are composed of multiple …
Compact: On-chip compression of activations for low power systolic array based CNN acceleration
J Zhang, P Raj, S Zarar… - ACM Transactions on …, 2019 - nyuscholars.nyu.edu
This paper addresses the design of systolic array (SA) based convolutional neural network
(CNN) accelerators for mobile and embedded domains. On-and off-chip memory accesses …
(CNN) accelerators for mobile and embedded domains. On-and off-chip memory accesses …