Tractable Large-Scale Deep Reinforcement Learning

12 Pages Posted: 27 May 2022 Last revised: 11 Jul 2022

See all articles by Nima Sarang

Nima Sarang

Concordia University

Charalambos Poullis

Concordia University

Date Written: October 15, 2021

Abstract

Reinforcement learning (RL) has emerged as one of the most promising and powerful techniques in deep learning. The training of intelligent agents requires a myriad of training examples which imposes a substantial computational cost. Consequently, RL is seldom applied to real-world problems and historically has been limited to computer vision tasks, similar to supervised learning. This work proposes an RL framework for complex, partially observable, large-scale environments. We introduce novel techniques for tractable training on commodity GPUs, and significantly reduce computational costs. Furthermore, we present a self-supervised loss that improves the learning stability in applications with a long-time horizon, shortening the training time. We demonstrate the effectiveness of the proposed solution on the application of road extraction from high-resolution satellite images. We present experiments on satellite images of fifteen cities that demonstrate comparable performance to state-of-the-art methods. To the best of our knowledge, this is the first time RL has been applied for extracting road networks. The code is publicly available at: https://github.com/nsarang/road-extraction-rl.

Keywords: Deep Reinforcement Learning, road extraction

Suggested Citation

Sarang, Nima and Poullis, Charalambos, Tractable Large-Scale Deep Reinforcement Learning (October 15, 2021). Available at SSRN: https://ssrn.com/abstract=4121080 or http://dx.doi.org/10.2139/ssrn.4121080

Nima Sarang (Contact Author)

Concordia University ( email )

Canada

Charalambos Poullis

Concordia University ( email )

Canada

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
106
Abstract Views
264
Rank
403,515
PlumX Metrics