Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

Attention based Curiosity-driven Exploration in Deep Reinforcement Learning

Abstract: 

Reinforcement Learning enables to train an agent via interaction with the environment. However, in the majority of real-world scenarios, the extrinsic feedback is sparse or not sufficient, thus intrinsic reward formulations are needed to successfully train the agent. This work investigates and extends the paradigm of curiosity-driven exploration. First, a probabilistic approach is taken to exploit the advantages of the attention mechanism, which is successfully applied in other domains of Deep Learning. Combining them, we propose new methods, such as AttA2C, an extension of the Actor-Critic framework. Second, another curiosity-based approach - ICM - is extended. The proposed model utilizes attention to emphasize features for the dynamic models within ICM, moreover, we also modify the loss function, resulting in a new curiosity formulation, which we call rational curiosity. The corresponding implementation can be found at https://github.com/rpatrik96/AttA2C/

up
0 users have voted:

Paper Details

Authors:
Patrik Reizinger, Márton Szemenyei
Submitted On:
17 May 2020 - 3:59pm
Short Link:
Type:
Presentation Slides
Event:
Presenter's Name:
Patrik Reizinger
Paper Code:
MLSP-P3.3
Document Year:
2020
Cite

Document Files

ICASSP_presentation.pdf

(18)

Subscribe

[1] Patrik Reizinger, Márton Szemenyei, "Attention based Curiosity-driven Exploration in Deep Reinforcement Learning", IEEE SigPort, 2020. [Online]. Available: http://sigport.org/5386. Accessed: Jul. 14, 2020.
@article{5386-20,
url = {http://sigport.org/5386},
author = {Patrik Reizinger; Márton Szemenyei },
publisher = {IEEE SigPort},
title = {Attention based Curiosity-driven Exploration in Deep Reinforcement Learning},
year = {2020} }
TY - EJOUR
T1 - Attention based Curiosity-driven Exploration in Deep Reinforcement Learning
AU - Patrik Reizinger; Márton Szemenyei
PY - 2020
PB - IEEE SigPort
UR - http://sigport.org/5386
ER -
Patrik Reizinger, Márton Szemenyei. (2020). Attention based Curiosity-driven Exploration in Deep Reinforcement Learning. IEEE SigPort. http://sigport.org/5386
Patrik Reizinger, Márton Szemenyei, 2020. Attention based Curiosity-driven Exploration in Deep Reinforcement Learning. Available at: http://sigport.org/5386.
Patrik Reizinger, Márton Szemenyei. (2020). "Attention based Curiosity-driven Exploration in Deep Reinforcement Learning." Web.
1. Patrik Reizinger, Márton Szemenyei. Attention based Curiosity-driven Exploration in Deep Reinforcement Learning [Internet]. IEEE SigPort; 2020. Available from : http://sigport.org/5386