Exact solutions to the nonlinear dynamics of learning in deep linear neural networks

By Andrew M. Saxe, James L. McClelland, Surya Ganguli. 2013

2,139 citations recorded by the source corpus.

Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case of deep linear neural networks. Despite the linearity of their input-output map, such networks have nonlinear gradient descent dynamics on weights that change with the addition of each new hidden layer. We show that deep linear networks exhibit nonlinear learning phenomena similar to those seen in simulations of nonlinear networks, including long plateaus followed by rapid transitions to lower error solutions, and faster convergence from greedy unsupervised pretraining initial conditions than from random initial conditions. We provide an analytical description of these phenomena by finding new exact solutions to the nonlinear dynamics of deep learning. Our theoretical analysis also reveals the surprising finding that as the depth of a network approaches infinity, learning speed can nevertheless remain finite: for a special class of initial conditions on the weights, very deep networks incur only a finite, depth independent, delay in learning speed relative to shallow networks. We show that, under certain conditions on the training data, unsupervised pretraining can find this special class of initial conditions, while scaled random Gaussian initializations cannot. We further exhibit a new class of random orthogonal initial conditions on weights that, like unsupervised pre-training, enjoys depth independent learning times. We further show that these initial conditions also lead to faithful propagation of gradients even in deep nonlinear networks, as long as they operate in a special regime known as the edge of chaos.

Subfields: Artificial Intelligence, Statistical and Nonlinear Physics, Optimization & Training, Scaling & Learning Theory

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

  1. Yoshua Bengio
  2. Mohit Bansal
  3. Sergey Levine
  4. Jianfeng Gao
  5. Quoc V. Le
  6. Percy Liang
  7. Ruslan Salakhutdinov
  8. Kyunghyun Cho

In Statistical and Nonlinear Physics

  1. Greg Yang
  2. Yoshua Bengio
  3. Sergey Levine
  4. Sam Schoenholz
  5. Roger Baker Grosse
  6. Yuanzhi Li
  7. Jeffrey Pennington
  8. Pieter Abbeel

In Optimization & Training

  1. Yoshua Bengio
  2. Quoc V. Le
  3. Soham De
  4. Roger Baker Grosse
  5. Razvan Pascanu
  6. Simon Lacoste-Julien
  7. Stefano Soatto
  8. Yann Dauphin

In Scaling & Learning Theory

  1. Greg Yang
  2. Yuanzhi Li
  3. Tengyu Ma
  4. Simon Shaolei Du
  5. Jason D. Lee
  6. Sam Schoenholz
  7. Yoshua Bengio
  8. Matthieu Wyart

Citations in the corpus