Spatiotemporally Constrained Action Space Attacks On Deep Reinforcement Learning Agents.pdf

1909.02583.pdf
Preview of Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents
🔗 Source: arxiv.org
📊 Size: 2.32 MB
📄 Pages: 18 pages
⬇️ Downloads: 225

Summary

Researchers propose two attack strategies, Myopic Action Space (MAS) and Look-ahead Action Space (LAS), to compromise Deep Reinforcement Learning (DRL) agents by targeting their action space, which represents physically manipulable actuators in engineering systems. The MAS attack distributes attacks across action space dimensions, while the LAS attack considers the agent's dynamics and distributes attacks across both action and temporal dimensions. Results show that LAS attacks deteriorate agent performance more significantly than MAS attacks using the same resources, highlighting the potential for adversaries to craft effective attacks with limited resources. These attack strategies can also be used to understand a DRL agent's vulnerabilities.

Description

Researchers propose two attack strategies, Myopic Action Space (MAS) and Look-ahead Action Space (LAS), to compromise Deep Reinforcement Learning (DRL) agents...

Technical Information

  • File Format: PDF
  • File Size: 2.32 MB
  • Pages: 18
  • Language: EN
  • Total Downloads: 225
  • Last Updated: 5 days ago

Document Overview

This PDF document about Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents provides comprehensive information and guidance. Whether you're a beginner or advanced user, this resource offers valuable insights into Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents.

Related Topics

If you're interested in Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents, you might also want to explore:

Download Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents eBooks for free and learn more about Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents. These books contain exercises and tutorials to improve your practical skills, at all levels!

Not satisfied with this document? We have related documents to Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents, try searching with similar keywords: Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents, The Outer Space Treaty and Enhancing Space Security, Conference Report, Building the Architecture for Sustainable Space Security, The Outer Space Treaty And Enhancing Space Security Conference Report Building The Architecture For Sustainable Space Security, Spatiotemporally Discrete Urban Traffic Model, Udemy Advanced AI Deep Reinforcement Learning In P, Udemy Advanced AI Deep Reinforcement Learning In Python, Opportunities and Challenges in Applying Deep Reinforcement Learning to Support Production-Inventory Planning at ASML, Deep Learning and Kernel Machines: RBMs, Deep BMs, RKMs, and Generative Models for Explainable AI

You can download PDF versions of the user's guide, manuals and ebooks about Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents, you can also find and download for free A free online manual (notices) with beginner and intermediate, Downloads Documentation, You can download PDF files (or DOC and PPT) about Spatiotemporally Constrained Action Space Attacks on Deep Reinforcement Learning Agents for free, but please respect copyrighted ebooks.