Natural Neural Networks

TL;DR

Natural Neural Networks introduce a reparametrization technique that improves training efficiency by better conditioning the Fisher matrix, demonstrated on large-scale image classification tasks.

Contribution

The paper presents a new family of algorithms with a reparametrization method and a scalable natural gradient-based training algorithm for neural networks.

Findings

Faster convergence in training neural networks.

Effective on both supervised and unsupervised tasks.

Scalable to large datasets like ImageNet.

Abstract

We introduce Natural Neural Networks, a novel family of algorithms that speed up convergence by adapting their internal representation during training to improve conditioning of the Fisher matrix. In particular, we show a specific example that employs a simple and efficient reparametrization of the neural network weights by implicitly whitening the representation obtained at each layer, while preserving the feed-forward computation of the network. Such networks can be trained efficiently via the proposed Projected Natural Gradient Descent algorithm (PRONG), which amortizes the cost of these reparametrizations over many parameter updates and is closely related to the Mirror Descent online learning algorithm. We highlight the benefits of our method on both unsupervised and supervised learning tasks, and showcase its scalability by training on the large-scale ImageNet Challenge dataset.