Recurrent Neural Networks (RNNs) are becoming increasingly important for time series-related applications which require efficient and real-time implementations. The recent pruning based work ESE suffers from degradation of performance/energy efficiency due to the irregular network structure after pruning. We propose block-circulant matrices for weight matrix representation in RNNs, thereby achieving simultaneous model compression and acceleration. We aim to implement RNNs in FPGA with highest performance and energy efficiency, with certain accuracy requirement (negligible accuracy degradation). Experimental results on actual FPGA deployments shows that the proposed framework achieves a maximum energy efficiency improvement of 35.7× compared with ESE.
@article{arxiv.1803.07661,
title = {Efficient Recurrent Neural Networks using Structured Matrices in FPGAs},
author = {Zhe Li and Shuo Wang and Caiwen Ding and Qinru Qiu and Yanzhi Wang and Yun Liang},
journal= {arXiv preprint arXiv:1803.07661},
year = {2018}
}
Comments
To appear in International Conference on Learning Representations 2018 Workshop Track