No More Pesky Learning Rates

TL;DR

This paper introduces an adaptive method for automatically adjusting multiple learning rates in stochastic gradient descent, eliminating the need for manual tuning and improving performance on various tasks.

Contribution

It presents a novel algorithm that dynamically adjusts learning rates based on local gradient variations, suitable for non-stationary problems.

Findings

Matches the performance of well-tuned SGD and adaptive methods

Removes the need for manual learning rate tuning

Effective on convex and non-convex tasks

Abstract

The performance of stochastic gradient descent (SGD) depends critically on how learning rates are tuned and decreased over time. We propose a method to automatically adjust multiple learning rates so as to minimize the expected error at any one time. The method relies on local gradient variations across samples. In our approach, learning rates can increase as well as decrease, making it suitable for non-stationary problems. Using a number of convex and non-convex learning tasks, we show that the resulting algorithm matches the performance of SGD or other adaptive approaches with their best settings obtained through systematic search, and effectively removes the need for learning rate tuning.