17th INTERNATIONAL FORUM ON MPSoC
for software-defined hardware
For further information, please send email to Frédéric Pétrot
Professor at Seoul National University, Korea
Zena: Zero-aware neural network accelerator
It is imperative to accelerate convolutional neural networks (CNNs) due to their ever-widening application areas from server, mobile to IoT devices. Based on the fact that CNNs can be characterized by significant amount of zero values in both kernel weights (under quality-preserving pruning) and activations (when rectified linear units are applied), we propose a novel architecture of hardware accelerator for CNNs which exploits zero values in both weights and activations. We also report a zero-induced load imbalance problem encountered in the zero-aware parallel architecture and present a zero-aware kernel allocation. In our experiments, we designed a cycle-accurate model, RTL and layout designs of the proposed architecture. In our evaluations with two real deep CNNs, pruned AlexNet and VGG-16, our proposed architecture offers 4.4x/2.0x times (AlexNet) and 5.6x/2.4x times (VGG-16) speedup compared with state-of-the-art zero-agnostic/zero activation-aware architectures.
Sungjoo Yoo received Ph.D. from Seoul National University in 2000. He worked as researcher at TIMA laboratory, Grenoble France from 2000 to 2004. He was principal engineer at Samsung System LSI from 2004 to 2008. He was at POSTECH from 2008 to 2015. He joined Seoul National University in 2015 and is now associate professor. His current research interests include low power deep neural networks and near data processing (processing-in-memory and in-storage processing).