Department of Mathematics
 Search | Help | Login | pdf version | printable version

Math @ Duke



Publications [#328865] of Guillermo Sapiro

Papers Published

  1. Sokolić, J; Giryes, R; Sapiro, G; Rodrigues, MRD, Robust Large Margin Deep Neural Networks, Ieee Transactions on Signal Processing, vol. 65 no. 16 (August, 2017), pp. 4265-4280, Institute of Electrical and Electronics Engineers (IEEE) [doi]
    (last updated on 2019/06/17)

    © 2017 IEEE. The generalization error of deep neural networks via their classification margin is studied in this paper. Our approach is based on the Jacobian matrix of a deep neural network and can be applied to networks with arbitrary nonlinearities and pooling layers, and to networks with different architectures such as feed forward networks and residual networks. Our analysis leads to the conclusion that a bounded spectral norm of the network's Jacobian matrix in the neighbourhood of the training samples is crucial for a deep neural network of arbitrary depth and width to generalize well. This is a significant improvement over the current bounds in the literature, which imply that the generalization error grows with either the width or the depth of the network. Moreover, it shows that the recently proposed batch normalization and weight normalization reparametrizations enjoy good generalization properties, and leads to a novel network regularizer based on the network's Jacobian matrix. The analysis is supported with experimental results on the MNIST, CIFAR-10, LaRED, and ImageNet datasets.
ph: 919.660.2800
fax: 919.660.2821

Mathematics Department
Duke University, Box 90320
Durham, NC 27708-0320