A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning

Reinforcement learning (RL) is one kind of interactive learning methods. Its main characteristics are “trial and error” and “related reward.” A hierarchical reinforcement learning method based on action subrewards is proposed to solve the problem of “curse of dimensionality,” which means that the st...

Full description

Saved in:
Bibliographic Details
Main Authors: Yuchen Fu, Quan Liu, Xionghong Ling, Zhiming Cui
Format: Article
Language:English
Published: Wiley 2014-01-01
Series:The Scientific World Journal
Online Access:http://dx.doi.org/10.1155/2014/120760
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1832550237648977920
author Yuchen Fu
Quan Liu
Xionghong Ling
Zhiming Cui
author_facet Yuchen Fu
Quan Liu
Xionghong Ling
Zhiming Cui
author_sort Yuchen Fu
collection DOAJ
description Reinforcement learning (RL) is one kind of interactive learning methods. Its main characteristics are “trial and error” and “related reward.” A hierarchical reinforcement learning method based on action subrewards is proposed to solve the problem of “curse of dimensionality,” which means that the states space will grow exponentially in the number of features and low convergence speed. The method can reduce state spaces greatly and choose actions with favorable purpose and efficiency so as to optimize reward function and enhance convergence speed. Apply it to the online learning in Tetris game, and the experiment result shows that the convergence speed of this algorithm can be enhanced evidently based on the new method which combines hierarchical reinforcement learning algorithm and action subrewards. The “curse of dimensionality” problem is also solved to a certain extent with hierarchical method. All the performance with different parameters is compared and analyzed as well.
format Article
id doaj-art-b5cd1ef362ba4bef85b26678258514d6
institution Kabale University
issn 2356-6140
1537-744X
language English
publishDate 2014-01-01
publisher Wiley
record_format Article
series The Scientific World Journal
spelling doaj-art-b5cd1ef362ba4bef85b26678258514d62025-02-03T06:07:21ZengWileyThe Scientific World Journal2356-61401537-744X2014-01-01201410.1155/2014/120760120760A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement LearningYuchen Fu0Quan Liu1Xionghong Ling2Zhiming Cui3Suzhou Industrial Park Institute of Services Outsourcing, Suzhou, Jiangsu 215123, ChinaSchool of Computer Science and Technology, Soochow University, Suzhou, Jiangsu 215006, ChinaSchool of Computer Science and Technology, Soochow University, Suzhou, Jiangsu 215006, ChinaSchool of Computer Science and Technology, Soochow University, Suzhou, Jiangsu 215006, ChinaReinforcement learning (RL) is one kind of interactive learning methods. Its main characteristics are “trial and error” and “related reward.” A hierarchical reinforcement learning method based on action subrewards is proposed to solve the problem of “curse of dimensionality,” which means that the states space will grow exponentially in the number of features and low convergence speed. The method can reduce state spaces greatly and choose actions with favorable purpose and efficiency so as to optimize reward function and enhance convergence speed. Apply it to the online learning in Tetris game, and the experiment result shows that the convergence speed of this algorithm can be enhanced evidently based on the new method which combines hierarchical reinforcement learning algorithm and action subrewards. The “curse of dimensionality” problem is also solved to a certain extent with hierarchical method. All the performance with different parameters is compared and analyzed as well.http://dx.doi.org/10.1155/2014/120760
spellingShingle Yuchen Fu
Quan Liu
Xionghong Ling
Zhiming Cui
A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning
The Scientific World Journal
title A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning
title_full A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning
title_fullStr A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning
title_full_unstemmed A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning
title_short A Reward Optimization Method Based on Action Subrewards in Hierarchical Reinforcement Learning
title_sort reward optimization method based on action subrewards in hierarchical reinforcement learning
url http://dx.doi.org/10.1155/2014/120760
work_keys_str_mv AT yuchenfu arewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning
AT quanliu arewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning
AT xionghongling arewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning
AT zhimingcui arewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning
AT yuchenfu rewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning
AT quanliu rewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning
AT xionghongling rewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning
AT zhimingcui rewardoptimizationmethodbasedonactionsubrewardsinhierarchicalreinforcementlearning