Learning with Random Learning Rates

TL;DR

Alrao is a novel optimization method where each network unit gets a randomly sampled learning rate, enabling near-optimal training performance without hyperparameter tuning, thus saving time and computational resources.

Contribution

The paper introduces Alrao, a simple yet effective method assigning random learning rates to units, performing comparably to optimally tuned SGD across various models and tasks.

Findings

Alrao achieves performance close to optimally tuned SGD.

It requires no additional computational cost.

It can quickly evaluate multiple models with minimal tuning.

Abstract

Hyperparameter tuning is a bothersome step in the training of deep learning models. One of the most sensitive hyperparameters is the learning rate of the gradient descent. We present the 'All Learning Rates At Once' (Alrao) optimization method for neural networks: each unit or feature in the network gets its own learning rate sampled from a random distribution spanning several orders of magnitude. This comes at practically no computational cost. Perhaps surprisingly, stochastic gradient descent (SGD) with Alrao performs close to SGD with an optimally tuned learning rate, for various architectures and problems. Alrao could save time when testing deep learning models: a range of models could be quickly assessed with Alrao, and the most promising models could then be trained more extensively. This text comes with a PyTorch implementation of the method, which can be plugged on an existing PyTorch model: https://github.com/leonardblier/alrao .