Less Is More: Towards Compact CNNs

  title={Less Is More: Towards Compact CNNs},
  author={Hao Zhou and Jose M. Alvarez and Fatih Murat Porikli},
We proposed an algorithm to significantly reduce of the number of neurons in a convolutional neural network by adding sparse constraints during the training step. The forward-backward splitting method is applied to solve the sparse constrained problem. We also analyze the benefits of using rectified linear units as non-linear activation function to remove a larger number of neurons. Experiments using four popular CNNs including AlexNet and VGG-B demonstrate the capacity of the proposed method… CONTINUE READING
Highly Cited
This paper has 53 citations. REVIEW CITATIONS



Citations per Year

53 Citations

Semantic Scholar estimates that this publication has 53 citations based on the available data.

See our FAQ for additional information.