Mean-variance Based Risk-sensitive Reinforcement Learning with Interpretable Attention

Woo Kyung Kim; Youngseok Lee; Honguk Woo

doi:10.1145/3523111.3523127

Mean-variance Based Risk-sensitive Reinforcement Learning with Interpretable Attention

Woo Kyung Kim
, Youngseok Lee
, Honguk Woo

Department of Computer Science and Engineering

Sungkyunkwan University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

2 Scopus citations

Abstract

Risk-sensitive reinforcement learning (RL) has been studied to address the risk and uncertainty in autonomous systems. While a comprehensive understanding for the behaviors of RL agents plays an important role, interpretability was rarely discussed in the context of risk-sensitivity RL. In this paper, we present an interpretable visualization scheme with attention mechanism in which a saliency map represents the relative influence degree of an input state on the decision-making of mean-variance based risk-sensitive RL. Through 2D navigation experiments, we demonstrate that our scheme provides the interpretability with regard to risk-sensitive levels.

Original language	English
Title of host publication	ICMVA 2022 - 5th International Conference on Machine Vision and Applications
Publisher	Association for Computing Machinery
Pages	104-109
Number of pages	6
ISBN (Electronic)	9781450395670
DOIs	https://doi.org/10.1145/3523111.3523127
State	Published - 18 Feb 2022
Event	5th International Conference on Machine Vision and Applications, ICMVA 2022 - Singapore, Singapore Duration: 18 Feb 2022 → 20 Feb 2022

Publication series

Name	ACM International Conference Proceeding Series

Conference

Conference	5th International Conference on Machine Vision and Applications, ICMVA 2022
Country/Territory	Singapore
City	Singapore
Period	18/02/22 → 20/02/22

Keywords

Explainable Reinforcement Learning
Interpretable Attention Networks
Risk Visualization
Risk-sensitive Reinforcement Learning

Access to Document

10.1145/3523111.3523127

Cite this

@inproceedings{b303c8d80dde49a6b693a379e26b9a3d,

title = "Mean-variance Based Risk-sensitive Reinforcement Learning with Interpretable Attention",

abstract = "Risk-sensitive reinforcement learning (RL) has been studied to address the risk and uncertainty in autonomous systems. While a comprehensive understanding for the behaviors of RL agents plays an important role, interpretability was rarely discussed in the context of risk-sensitivity RL. In this paper, we present an interpretable visualization scheme with attention mechanism in which a saliency map represents the relative influence degree of an input state on the decision-making of mean-variance based risk-sensitive RL. Through 2D navigation experiments, we demonstrate that our scheme provides the interpretability with regard to risk-sensitive levels.",

keywords = "Explainable Reinforcement Learning, Interpretable Attention Networks, Risk Visualization, Risk-sensitive Reinforcement Learning",

author = "Kim, \{Woo Kyung\} and Youngseok Lee and Honguk Woo",

note = "Publisher Copyright: {\textcopyright} 2022 ACM.; 5th International Conference on Machine Vision and Applications, ICMVA 2022 ; Conference date: 18-02-2022 Through 20-02-2022",

year = "2022",

month = feb,

day = "18",

doi = "10.1145/3523111.3523127",

language = "English",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

pages = "104--109",

booktitle = "ICMVA 2022 - 5th International Conference on Machine Vision and Applications",

}

Kim, WK, Lee, Y & Woo, H 2022, Mean-variance Based Risk-sensitive Reinforcement Learning with Interpretable Attention. in ICMVA 2022 - 5th International Conference on Machine Vision and Applications. ACM International Conference Proceeding Series, Association for Computing Machinery, pp. 104-109, 5th International Conference on Machine Vision and Applications, ICMVA 2022, Singapore, Singapore, 18/02/22. https://doi.org/10.1145/3523111.3523127

Mean-variance Based Risk-sensitive Reinforcement Learning with Interpretable Attention. / Kim, Woo Kyung; Lee, Youngseok; Woo, Honguk.
ICMVA 2022 - 5th International Conference on Machine Vision and Applications. Association for Computing Machinery, 2022. p. 104-109 (ACM International Conference Proceeding Series).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Mean-variance Based Risk-sensitive Reinforcement Learning with Interpretable Attention

AU - Kim, Woo Kyung

AU - Lee, Youngseok

AU - Woo, Honguk

PY - 2022/2/18

Y1 - 2022/2/18

N2 - Risk-sensitive reinforcement learning (RL) has been studied to address the risk and uncertainty in autonomous systems. While a comprehensive understanding for the behaviors of RL agents plays an important role, interpretability was rarely discussed in the context of risk-sensitivity RL. In this paper, we present an interpretable visualization scheme with attention mechanism in which a saliency map represents the relative influence degree of an input state on the decision-making of mean-variance based risk-sensitive RL. Through 2D navigation experiments, we demonstrate that our scheme provides the interpretability with regard to risk-sensitive levels.

AB - Risk-sensitive reinforcement learning (RL) has been studied to address the risk and uncertainty in autonomous systems. While a comprehensive understanding for the behaviors of RL agents plays an important role, interpretability was rarely discussed in the context of risk-sensitivity RL. In this paper, we present an interpretable visualization scheme with attention mechanism in which a saliency map represents the relative influence degree of an input state on the decision-making of mean-variance based risk-sensitive RL. Through 2D navigation experiments, we demonstrate that our scheme provides the interpretability with regard to risk-sensitive levels.

KW - Explainable Reinforcement Learning

KW - Interpretable Attention Networks

KW - Risk Visualization

KW - Risk-sensitive Reinforcement Learning

UR - https://www.scopus.com/pages/publications/85130357450

U2 - 10.1145/3523111.3523127

DO - 10.1145/3523111.3523127

M3 - Conference contribution

AN - SCOPUS:85130357450

T3 - ACM International Conference Proceeding Series

SP - 104

EP - 109

BT - ICMVA 2022 - 5th International Conference on Machine Vision and Applications

PB - Association for Computing Machinery

T2 - 5th International Conference on Machine Vision and Applications, ICMVA 2022

Y2 - 18 February 2022 through 20 February 2022

ER -

Mean-variance Based Risk-sensitive Reinforcement Learning with Interpretable Attention

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this