Continuous probabilistic model building genetic network programming using reinforcement learning

Xianneng Li, Kotaro Hirasawa

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Recently, a novel probabilistic model-building evolutionary algorithm (so called estimation of distribution algorithm, or EDA), named probabilistic model building genetic network programming (PMBGNP), has been proposed. PMBGNP uses graph structures for its individual representation, which shows higher expression ability than the classical EDAs. Hence, it extends EDAs to solve a range of problems, such as data mining and agent control. This paper is dedicated to propose a continuous version of PMBGNP for continuous optimization in agent control problems. Different from the other continuous EDAs, the proposed algorithm evolves the continuous variables by reinforcement learning (RL). We compare the performance with several state-of-the-art algorithms on a real mobile robot control problem. The results show that the proposed algorithm outperforms the others with statistically significant differences.

Original languageEnglish
Pages (from-to)457-467
Number of pages11
JournalApplied Soft Computing Journal
Volume27
DOIs
Publication statusPublished - 2015

Fingerprint

Reinforcement learning
Evolutionary algorithms
Mobile robots
Data mining
Statistical Models

Keywords

  • Continuous optimization
  • Estimation of distribution algorithm
  • Network programming
  • Probabilistic model building genetic
  • Reinforcement learning

ASJC Scopus subject areas

  • Software

Cite this

Continuous probabilistic model building genetic network programming using reinforcement learning. / Li, Xianneng; Hirasawa, Kotaro.

In: Applied Soft Computing Journal, Vol. 27, 2015, p. 457-467.

Research output: Contribution to journalArticle

@article{739e763c84174f878d453e00c742ffa0,
title = "Continuous probabilistic model building genetic network programming using reinforcement learning",
abstract = "Recently, a novel probabilistic model-building evolutionary algorithm (so called estimation of distribution algorithm, or EDA), named probabilistic model building genetic network programming (PMBGNP), has been proposed. PMBGNP uses graph structures for its individual representation, which shows higher expression ability than the classical EDAs. Hence, it extends EDAs to solve a range of problems, such as data mining and agent control. This paper is dedicated to propose a continuous version of PMBGNP for continuous optimization in agent control problems. Different from the other continuous EDAs, the proposed algorithm evolves the continuous variables by reinforcement learning (RL). We compare the performance with several state-of-the-art algorithms on a real mobile robot control problem. The results show that the proposed algorithm outperforms the others with statistically significant differences.",
keywords = "Continuous optimization, Estimation of distribution algorithm, Network programming, Probabilistic model building genetic, Reinforcement learning",
author = "Xianneng Li and Kotaro Hirasawa",
year = "2015",
doi = "10.1016/j.asoc.2014.10.023",
language = "English",
volume = "27",
pages = "457--467",
journal = "Applied Soft Computing",
issn = "1568-4946",
publisher = "Elsevier BV",

}

TY - JOUR

T1 - Continuous probabilistic model building genetic network programming using reinforcement learning

AU - Li, Xianneng

AU - Hirasawa, Kotaro

PY - 2015

Y1 - 2015

N2 - Recently, a novel probabilistic model-building evolutionary algorithm (so called estimation of distribution algorithm, or EDA), named probabilistic model building genetic network programming (PMBGNP), has been proposed. PMBGNP uses graph structures for its individual representation, which shows higher expression ability than the classical EDAs. Hence, it extends EDAs to solve a range of problems, such as data mining and agent control. This paper is dedicated to propose a continuous version of PMBGNP for continuous optimization in agent control problems. Different from the other continuous EDAs, the proposed algorithm evolves the continuous variables by reinforcement learning (RL). We compare the performance with several state-of-the-art algorithms on a real mobile robot control problem. The results show that the proposed algorithm outperforms the others with statistically significant differences.

AB - Recently, a novel probabilistic model-building evolutionary algorithm (so called estimation of distribution algorithm, or EDA), named probabilistic model building genetic network programming (PMBGNP), has been proposed. PMBGNP uses graph structures for its individual representation, which shows higher expression ability than the classical EDAs. Hence, it extends EDAs to solve a range of problems, such as data mining and agent control. This paper is dedicated to propose a continuous version of PMBGNP for continuous optimization in agent control problems. Different from the other continuous EDAs, the proposed algorithm evolves the continuous variables by reinforcement learning (RL). We compare the performance with several state-of-the-art algorithms on a real mobile robot control problem. The results show that the proposed algorithm outperforms the others with statistically significant differences.

KW - Continuous optimization

KW - Estimation of distribution algorithm

KW - Network programming

KW - Probabilistic model building genetic

KW - Reinforcement learning

UR - http://www.scopus.com/inward/record.url?scp=84919727161&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84919727161&partnerID=8YFLogxK

U2 - 10.1016/j.asoc.2014.10.023

DO - 10.1016/j.asoc.2014.10.023

M3 - Article

AN - SCOPUS:84919727161

VL - 27

SP - 457

EP - 467

JO - Applied Soft Computing

JF - Applied Soft Computing

SN - 1568-4946

ER -