2016

The Loss Surface of Residual Networks: Ensembles and the Role of Batch Normalization

Littwin, Etai, Wolf, Lior

Understand

Deep Residual Networks present a premium in performance in comparison to conventional networks of the same depth and are trainable at extreme depths.

  • It has recently been shown that Residual Networks behave like ensembles of relatively shallow networks.
  • We show that these ensembles are dynamic: while initially the virtual ensemble is mostly at depths lower than half the network's depth, as training progresses, it becomes deeper and deeper.
  • The main mechanism that controls the dynamic ensemble behavior is the scaling introduced, e.g., by the Batch Normalization technique.

Built on

  • Neural networks: tricks of the trade

    Genevieve B Orr and Klaus-Robert Müller · 2003

    Earlier work this paper cites.

  • Understanding the difficulty of training deep feedforward neural networks

    Xavier Glorot and Yoshua Bengio · 2010

    Earlier work this paper cites.

  • Complexity of random smooth functions on the high-dimensional sphere

    Antonio Auffinger and Gerard Ben Arous · 2013

    Earlier work this paper cites.

  • Random matrices and complexity of spin glasses

    Antonio Auffinger, Gérard Ben Arous, and Jiří Černý · 2013

    Earlier work this paper cites.

Similar

  • The loss surfaces of multilayer networks

    Anna Choromanska, Mikael Henaff, Michaël Mathieu, Gérard Ben Arous, and Yann LeCun · 2015

    Cited alongside, same era.

  • Deep residual learning for image recognition

    Original

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun · 2015

    Cited alongside, same era.

  • Batch normalization: Accelerating deep network training by reducing internal covariate shift

    Sergey Ioffe and Christian Szegedy · 2015

    Cited alongside, same era.

Then

  • Highway networks

    Original

    Rupesh Kumar Srivastava, Klaus Greff, and Jürgen Schmidhuber · 2015

    Later among the works it cites.

  • Identity mappings in deep residual networks

    Original

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun · 2016

    Closest in time.

  • Residual networks behave like ensembles of relatively shallow networks

    Andreas Veit, Michael Wilber, and Serge Belongie · 2016

    Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…