Corpus ID: 8569309

Exploiting Cyclic Symmetry in Convolutional Neural Networks

@article{Dieleman2016ExploitingCS,
  title={Exploiting Cyclic Symmetry in Convolutional Neural Networks},
  author={S. Dieleman and J. Fauw and K. Kavukcuoglu},
  journal={ArXiv},
  year={2016},
  volume={abs/1602.02660}
}
  • S. Dieleman, J. Fauw, K. Kavukcuoglu
  • Published 2016
  • Computer Science, Mathematics
  • ArXiv
  • Many classes of images exhibit rotational symmetry. Convolutional neural networks are sometimes trained using data augmentation to exploit this, but they are still required to learn the rotation equivariance properties from the data. Encoding these properties into the network architecture, as we are already used to doing for translation equivariance by using convolutional layers, could result in a more efficient use of the parameter budget by relieving the model from learning them. We introduce… CONTINUE READING
    Deep Rotation Equivariant Network
    23
    Ju n 20 16 Group Equivariant Convolutional Networks
    Group Equivariant Convolutional Networks
    479
    Harmonic Networks: Deep Translation and Rotation Equivariance
    264
    Convolutional Networks for Spherical Signals
    34
    HexaConv
    10

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 33 REFERENCES
    Spatial Transformer Networks
    2864
    Group Equivariant Convolutional Networks
    479
    Deep Symmetry Networks
    166
    Tiled convolutional neural networks
    268
    Transforming Auto-Encoders
    519
    Learning invariant representations and applications to face verification
    54
    Going deeper with convolutions
    19210