English

Accelerating Training of Deep Neural Networks via Sparse Edge Processing

Neural and Evolutionary Computing 2017-11-07 v1 Machine Learning

Abstract

We propose a reconfigurable hardware architecture for deep neural networks (DNNs) capable of online training and inference, which uses algorithmically pre-determined, structured sparsity to significantly lower memory and computational requirements. This novel architecture introduces the notion of edge-processing to provide flexibility and combines junction pipelining and operational parallelization to speed up training. The overall effect is to reduce network complexity by factors up to 30x and training time by up to 35x relative to GPUs, while maintaining high fidelity of inference results. This has the potential to enable extensive parameter searches and development of the largely unexplored theoretical foundation of DNNs. The architecture automatically adapts itself to different network sizes given available hardware resources. As proof of concept, we show results obtained for different bit widths.

Keywords

Cite

@article{arxiv.1711.01343,
  title  = {Accelerating Training of Deep Neural Networks via Sparse Edge Processing},
  author = {Sourya Dey and Yinan Shao and Keith M. Chugg and Peter A. Beerel},
  journal= {arXiv preprint arXiv:1711.01343},
  year   = {2017}
}

Comments

Presented at the 26th International Conference on Artificial Neural Networks (ICANN) 2017 in Alghero, Italy

R2 v1 2026-06-22T22:35:46.977Z