L4: Practical Loss-based Stepsize Adaptation For Deep Learning

Authors:
Michal Rolinek Max Planck Institute for Intelligent Systems
Georg Martius MPI for Intelligent Systems

Introduction:

The authors propose a stepsize adaptation scheme for stochastic gradient descent.It operates directly with the loss function and rescales the gradient in order to make fixed predicted progress on the loss.The authors demonstrate its capabilities by conclusively improving the performance of Adam and Momentum optimizers.The enhanced optimizers with default hyperparameters consistently outperform their constant stepsize counterparts, even the best ones, without a measurable increase in computational cost.The performance is validated on multiple architectures including dense nets, CNNs, ResNets, and the recurrent Differential Neural Computer on classical datasets MNIST, fashion MNIST, CIFAR10 and others.

Abstract:

We propose a stepsize adaptation scheme for stochastic gradient descent.It operates directly with the loss function and rescales the gradient in order to make fixed predicted progress on the loss.We demonstrate its capabilities by conclusively improving the performance of Adam and Momentum optimizers.The enhanced optimizers with default hyperparameters consistently outperform their constant stepsize counterparts, even the best ones, without a measurable increase in computational cost.The performance is validated on multiple architectures including dense nets, CNNs, ResNets, and the recurrent Differential Neural Computer on classical datasets MNIST, fashion MNIST, CIFAR10 and others.

You may want to know: