Holistic SparseCNN: Forging the Trident of Accuracy, Speed, and Size

@article{Park2016HolisticSF,
  title={Holistic SparseCNN: Forging the Trident of Accuracy, Speed, and Size},
  author={Jongsoo Park and Sheng R. Li and Wei Wen and Hai Li and Yiran Chen and Pradeep Dubey},
  journal={CoRR},
  year={2016},
  volume={abs/1608.01409}
}
We present Holistic SparseCNN, a sparse convolutional neural network design that simultaneously optimizes convolution layers (for classification speed) and fully connected layers (for model size), while maintaining the accuracy. We directly apply convolutions to tensors without bandwidth-wasting lowering step, which is critical for sparse convolution that is more prone to be bandwidth bound than its dense counterpart. Our cross-layer training method balances sparsity among multiple layers to… CONTINUE READING
Related Discussions
This paper has been referenced on Twitter 41 times. VIEW TWEETS

Citations

Publications citing this paper.
Showing 1-7 of 7 extracted citations

References

Publications referenced by this paper.
Showing 1-10 of 23 references

Caffe: Convolutional Architecture for Fast Feature Embedding

ACM Multimedia • 2014
View 7 Excerpts
Highly Influenced

Sparse Convolutional Neural Networks

2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) • 2015
View 17 Excerpts
Highly Influenced

Similar Papers

Loading similar papers…