Exact natural gradient in deep linear networks and application to the nonlinear case

Alberto Bernacchia, Máté Lengyel, Guillaume Hennequin

Research output: Contribution to journalConference articlepeer-review

Abstract (may include machine translation)

Stochastic gradient descent (SGD) remains the method of choice for deep learning, despite the limitations arising for ill-behaved objective functions. In cases where it could be estimated, the natural gradient has proven very effective at mitigating the catastrophic effects of pathological curvature in the objective function, but little is known theoretically about its convergence properties, and it has yet to find a practical implementation that would scale to very deep and large networks. Here, we derive an exact expression for the natural gradient in deep linear networks, which exhibit pathological curvature similar to the nonlinear case. We provide for the first time an analytical solution for its convergence rate, showing that the loss decreases exponentially to the global minimum in parameter space. Our expression for the natural gradient is surprisingly simple, computationally tractable, and explains why some approximations proposed previously work well in practice. This opens new avenues for approximating the natural gradient in the nonlinear case, and we show in preliminary experiments that our online natural gradient descent outperforms SGD on MNIST autoencoding while sharing its computational simplicity.

Original languageEnglish
Pages (from-to)5941-5950
Number of pages10
JournalAdvances in Neural Information Processing Systems
Volume2018-December
StatePublished - 2018
Event32nd Conference on Neural Information Processing Systems, NeurIPS 2018 - Montreal, Canada
Duration: 2 Dec 20188 Dec 2018

Fingerprint

Dive into the research topics of 'Exact natural gradient in deep linear networks and application to the nonlinear case'. Together they form a unique fingerprint.

Cite this