Off-Policy Differentiable Logic Reinforcement Learning

Li Zhang; Xin Li; Mingzhong Wang; Andong Tian

doi:10.1007/978-3-030-86520-7_38

Back

Off-Policy Differentiable Logic Reinforcement Learning

Conference paper

Peer reviewed

Off-Policy Differentiable Logic Reinforcement Learning

Li Zhang, Xin Li, Mingzhong Wang and Andong Tian

Machine Learning and Knowledge Discovery in Databases. Research Track, pp.1-16

European Conference on Machine Learning and Knowledge Discovery in Databases, 2021 (Virtual, 13-Sep-2021 - 17-Sep-2021)

Lecture Notes in Computer Science, 12976, Springer

2021

DOI: https://doi.org/10.1007/978-3-030-86520-7_38

Files and links (2)

url

https://2021.ecmlpkdd.org/wp-content/uploads/2021/07/sub_49.pdfView

Published Version

url

https://doi.org/10.1007/978-3-030-86520-7_38View

Published Version

Abstract

deep reinforcement learning

Interpretable reinforcement learning

Neural-Symbolic AI

In this paper, we proposed an Off-Policy Differentiable Logic Reinforcement Learning (OPDLRL) framework to inherit the benefits of interpretability and generalization ability in Differentiable Inductive Logic Programming (DILP) and also resolves its weakness of execution efficiency, stability, and scalability. The key contributions include the use of approximate inference to significantly reduce the number of logic rules in the deduction process, an off-policy training method to enable approximate inference, and a distributed and hierarchical training framework. Extensive experiments, specifically playing real-time video games in Rabbids against human players, show that OPDLRL has better or similar performance as other DILP-based methods but far more practical in terms of sample efficiency and execution efficiency, making it applicable to complex and (near) real-time domains.

Details

Title: Off-Policy Differentiable Logic Reinforcement Learning
Authors: Li Zhang (Author) - Beijing Institute of Technology
Xin Li (Corresponding Author) - Beijing Institute of Technology
Mingzhong Wang (Author) - University of the Sunshine Coast, Queensland, School of Science, Technology and Engineering
Andong Tian (Author) - Ubisoft China AI & Data Lab
Publication details: Machine Learning and Knowledge Discovery in Databases. Research Track, pp.1-16
Conference details: European Conference on Machine Learning and Knowledge Discovery in Databases, 2021 (Virtual, 13-Sep-2021 - 17-Sep-2021)
Series: Lecture Notes in Computer Science; 12976
Publisher: Springer
DOI: 10.1007/978-3-030-86520-7_38; 10.1007/978-3-030-86520-7
ISSN: 1611-3349
ISBN: 9783030865207
Organisation Unit: University of the Sunshine Coast, Queensland; School of Science, Technology and Engineering
Language: English
Record Identifier: 99571605202621
Output Type: Conference paper

Metrics

11 Record Views