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

TL;DR

This paper provides exact analytical solutions for the nonlinear learning dynamics in deep linear neural networks, revealing phenomena like plateaus, rapid transitions, and depth-independent learning speeds, bridging theory and practice.

Contribution

It introduces new exact solutions to the nonlinear dynamics of deep learning and uncovers conditions for depth-independent learning speed and effective initializations.

Findings

Deep linear networks exhibit nonlinear phenomena similar to nonlinear networks.

Learning speed can remain finite as depth increases under certain initial conditions.

Unsupervised pretraining and orthogonal initializations can achieve depth-independent learning times.

Abstract

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.