Corpus ID: 6564560

BinaryNet: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1

@article{Courbariaux2016BinaryNetTD,
  title={BinaryNet: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1},
  author={Matthieu Courbariaux and Yoshua Bengio},
  journal={ArXiv},
  year={2016},
  volume={abs/1602.02830}
}
  • Matthieu Courbariaux, Yoshua Bengio
  • Published 2016
  • Computer Science
  • ArXiv
  • We introduce BinaryNet, a method which trains DNNs with binary weights and activations when computing parameters’ gradient. [...] Key Method We wrote a binary matrix multiplication GPU kernel with which it is possible to run our MNIST MLP 7 times faster than with an unoptimized GPU kernel, without suffering any loss in classification accuracy. The code for BinaryNet is available.Expand Abstract
    447 Citations
    ALWANN: Automatic Layer-Wise Approximation of Deep Neural Network Accelerators without Retraining
    • 11
    • PDF
    DeepShift: Towards Multiplication-Less Neural Networks
    • 2
    • PDF
    XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks
    • 1,471
    • Highly Influenced
    • PDF
    How to Train a Compact Binary Neural Network with High Accuracy?
    • 150
    • Highly Influenced
    • PDF
    FINN: A Framework for Fast, Scalable Binarized Neural Network Inference
    • 386
    • Highly Influenced
    • PDF
    Recurrent Neural Networks With Limited Numerical Precision
    • 52
    • PDF
    Deterministic Binary Filters for Convolutional Neural Networks
    • 4
    • PDF
    Bitwise Neural Network Acceleration: Opportunities and Challenges
    • 1
    Stochastic Layer-Wise Precision in Deep Neural Networks
    • 7
    • Highly Influenced
    • PDF
    Loss-aware Binarization of Deep Networks
    • 115
    • Highly Influenced
    • PDF

    References

    SHOWING 1-10 OF 55 REFERENCES
    BinaryConnect: Training Deep Neural Networks with binary weights during propagations
    • 1,541
    • PDF
    Neural Networks with Few Multiplications
    • 232
    • PDF
    ImageNet classification with deep convolutional neural networks
    • 54,895
    • PDF
    Deep Learning using Linear Support Vector Machines
    • Y. Tang
    • Computer Science, Mathematics
    • 2013
    • 551
    • PDF
    Fixed-point feedforward deep neural network design using weights +1, 0, and −1
    • 188
    • PDF
    Learning both Weights and Connections for Efficient Neural Network
    • 2,606
    • PDF
    Improving the speed of neural networks on CPUs
    • 567
    • PDF
    Deeply-Supervised Nets
    • 1,100
    • PDF
    Going deeper with convolutions
    • 20,823
    • PDF