BitFlow: Exploiting Vector Parallelism for Binary Neural Networks on CPU

@article{Hu2018BitFlowEV,
  title={BitFlow: Exploiting Vector Parallelism for Binary Neural Networks on CPU},
  author={Yuwei Hu and Jidong Zhai and Dinghua Li and Yifan Gong and Yuhao Zhu and Weiwei Liu and Lei Su and Jiangming Jin},
  journal={2018 IEEE International Parallel and Distributed Processing Symposium (IPDPS)},
  year={2018},
  pages={244-253}
}
Deep learning has revolutionized computer vision and other fields since its big bang in 2012. However, it is challenging to deploy Deep Neural Networks (DNNs) into real-world applications due to their high computational complexity. Binary Neural Networks (BNNs) dramatically reduce computational complexity by replacing most arithmetic operations with bitwise operations. Existing implementations of BNNs have been focusing on GPU or FPGA, and using the conventional image-to-column method that… CONTINUE READING

Figures, Tables, Results, and Topics from this paper.

Key Quantitative Results

  • The results show that BitFlow achieves 83% speedup over unoptimized BNN implementations, and even outperforms a GPU implementation of full-precision DNNs.
  • Based on evaluation on the popular VGG network, we show that BitFlow obtains 83% speedup over unoptimized BNN implementations on a single core machine, and 10% speedup over GPU implementations of full-precision DNNs on a 64 core machine.

Citations

Publications citing this paper.

References

Publications referenced by this paper.
SHOWING 1-10 OF 35 REFERENCES

Deep learning binary neural network on an FPGA

  • 2017 IEEE 60th International Midwest Symposium on Circuits and Systems (MWSCAS)
  • 2017
VIEW 1 EXCERPT

Flexible Network Binarization with Layer-Wise Priority

  • 2018 25th IEEE International Conference on Image Processing (ICIP)
  • 2017
VIEW 3 EXCERPTS

Tensors and dynamic neural networks in python with strong gpu acceleration, 2017

Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan
  • 2017

Accelerating Binarized Neural Networks: Comparison of FPGA, CPU, GPU, and ASIC

  • 2016 International Conference on Field-Programmable Technology (FPT)
  • 2016
VIEW 1 EXCERPT

Similar Papers