Deep Reinforcement Learning using Cyclical Learning Rates

TL;DR

This paper introduces a novel application of cyclical learning rates in Deep Reinforcement Learning, demonstrating that it can achieve comparable or superior results to traditional hyperparameter tuning methods, reducing manual effort.

Contribution

It is the first to apply cyclical learning rates to DRL, providing a general method that improves training efficiency and performance without extensive hyperparameter tuning.

Findings

Cyclical learning rates match or outperform fixed rates in DRL tasks.

The method reduces the need for manual hyperparameter tuning.

Experimental results validate the effectiveness of cyclical learning in complex DRL problems.

Abstract

Deep Reinforcement Learning (DRL) methods often rely on the meticulous tuning of hyperparameters to successfully resolve problems. One of the most influential parameters in optimization procedures based on stochastic gradient descent (SGD) is the learning rate. We investigate cyclical learning and propose a method for defining a general cyclical learning rate for various DRL problems. In this paper we present a method for cyclical learning applied to complex DRL problems. Our experiments show that, utilizing cyclical learning achieves similar or even better results than highly tuned fixed learning rates. This paper presents the first application of cyclical learning rates in DRL settings and is a step towards overcoming manual hyperparameter tuning.