Corpus ID: 214611927

GroSS: Group-Size Series Decomposition for Grouped Architecture Search.

@article{HowardJenkins2020GroSSGS,
  title={GroSS: Group-Size Series Decomposition for Grouped Architecture Search.},
  author={Henry Howard-Jenkins and Y. Li and V. Prisacariu},
  journal={arXiv: Learning},
  year={2020}
}
  • Henry Howard-Jenkins, Y. Li, V. Prisacariu
  • Published 2020
  • Mathematics, Computer Science
  • arXiv: Learning
  • We present a novel approach which is able to explore the configuration of grouped convolutions within neural networks. Group-size Series (GroSS) decomposition is a mathematical formulation of tensor factorisation into a series of approximations of increasing rank terms. GroSS allows for dynamic and differentiable selection of factorisation rank, which is analogous to a grouped convolution. Therefore, to the best of our knowledge, GroSS is the first method to enable simultaneously train… CONTINUE READING

    Topics from this paper.

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 28 REFERENCES
    Aggregated Residual Transformations for Deep Neural Networks
    2538
    DARTS: Differentiable Architecture Search
    857
    Learning Transferable Architectures for Scalable Image Recognition
    1707
    EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
    773