Can I read Learning Rates as a Function of Batch Size: A Random Matrix Theory Approach to Neural Network Training on EtoBox?

Learning Rates as a Function of Batch Size: A Random Matrix Theory Approach to Neural Network Training by Granziol, Diego; Zohren, Stefan; Roberts, Stephen is a scholarly article available to read on EtoBox.

What is Learning Rates as a Function of Batch Size: A Random Matrix Theory Approach to Neural Network Training about?

We study the effect of mini-batching on the loss landscape of deep neural networks using spiked, field-dependent random matrix theory. We demonstrate that the magnitude of the extremal values of the batch Hessian are larger than those of the empirical Hessian. We also derive similar results for the Generalised Gauss-Newton matrix approximation of the Hessian. As a consequence of our theorems we derive an analytical expressions for the maximal learning rates as a function of batch size, informing practical training regimens for both stochastic gradient descent (linear scaling) and adaptive algorithms, such as Adam (square root scaling), for smooth, non-convex deep neural networks. Whilst the linear scaling for stochastic gradient descent has been derived under more restrictive conditions, which we generalise, the square root scaling rule for adaptive optimisers is, to our knowledge, completely novel. %For stochastic second-order methods and adaptive methods, we derive that the minimal damping coefficient is proportional to the ratio of the learning rate to batch size. We validate our claims on the VGG/WideResNet architectures on the CIFAR-$100$ and ImageNet datasets. Based on our in

Author

Granziol, Diego; Zohren, Stefan; Roberts, Stephen

Published

2020

Language

EN