Imperial College London

Professor Peter Y. K. Cheung

Faculty of EngineeringDyson School of Design Engineering

Head of the Dyson School of Design Engineering
 
 
 
//

Contact

 

+44 (0)20 7594 6200p.cheung Website

 
 
//

Assistant

 

Mrs Wiesia Hsissen +44 (0)20 7594 6261

 
//

Location

 

910BElectrical EngineeringSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@article{Wang:2020:10.1109/TC.2020.2978817,
author = {Wang, E and Davis, JJ and Cheung, P and Constantinides, GA},
doi = {10.1109/TC.2020.2978817},
journal = {IEEE Transactions on Computers},
pages = {1795--1808},
title = {LUTNet: Learning FPGA Configurations for Highly Efficient Neural Network Inference},
url = {http://dx.doi.org/10.1109/TC.2020.2978817},
volume = {69},
year = {2020}
}

RIS format (EndNote, RefMan)

TY  - JOUR
AB - Research has shown that deep neural networks contain significant redundancy, and thus that high classification accuracy can be achieved even when weights and activations are quantized down to binary values. Network binarization on FPGAs greatly increases area efficiency by replacing resource-hungry multipliers with lightweight XNOR gates. However, an FPGA's fundamental building block, the K-LUT, is capable of implementing far more than an XNOR: it can perform any K-input Boolean operation. Inspired by this observation, we propose LUTNet, an end-to-end hardware-software framework for the construction of area-efficient FPGA-based neural network accelerators using the native LUTs as inference operators. We describe the realization of both unrolled and tiled LUTNet architectures, with the latter facilitating smaller, less power-hungry deployment over the former while sacrificing area and energy efficiency along with throughput. For both varieties, we demonstrate that the exploitation of LUT flexibility allows for far heavier pruning than possible in prior works, resulting in significant area savings while achieving comparable accuracy. Against the state-of-the-art binarized neural network implementation, we achieve up to twice the area efficiency for several standard network models when inferencing popular datasets. We also demonstrate that even greater energy efficiency improvements are obtainable.
AU - Wang,E
AU - Davis,JJ
AU - Cheung,P
AU - Constantinides,GA
DO - 10.1109/TC.2020.2978817
EP - 1808
PY - 2020///
SN - 0018-9340
SP - 1795
TI - LUTNet: Learning FPGA Configurations for Highly Efficient Neural Network Inference
T2 - IEEE Transactions on Computers
UR - http://dx.doi.org/10.1109/TC.2020.2978817
UR - http://arxiv.org/abs/1910.12625
UR - https://ieeexplore.ieee.org/document/9026948
UR - http://hdl.handle.net/10044/1/80013
VL - 69
ER -