A graph-based evolutionary algorithm: Genetic Network Programming (GNP) and its extension using reinforcement learning

Shingo Mabu, Kotaro Hirasawa, Takayuki Furuzuki

研究成果: Article

251 引用 (Scopus)

抄録

This paper proposes a graph-based evolutionary algorithm called Genetic Network Programming (GNP). Our goal is to develop GNP, which can deal with dynamic environments efficiently and effectively, based on the distinguished expression ability of the graph (network) structure. The characteristics of GNP are as follows. 1) GNP programs are composed of a number of nodes which execute simple judgment/processing, and these nodes are connected by directed links to each other. 2) The graph structure enables GNP to re-use nodes, thus the structure can be very compact. 3) The node transition of GNP is executed according to its node connections without any terminal nodes, thus the pst history of the node transition affects the current node to be used and this characteristic works as an implicit memory function. These structural characteristics are useful for dealing with dynamic environments. Furthermore, we propose an extended algorithm, "GNP with Reinforcement Learning (GNP-RL)" which combines evolution and reinforcement learning in order to create effective graph structures and obtain better results in dynamic environments. In this paper, we applied GNP to the problem of determining agents' behavior to evaluate its effectiveness. Tileworld was used as the simulation environment. The results show some advantages for GNP over conventional methods.

元の言語English
ページ(範囲)369-398
ページ数30
ジャーナルEvolutionary Computation
15
発行部数3
DOI
出版物ステータスPublished - 2007 9

Fingerprint

Network Programming
Genetic Network
Reinforcement learning
Reinforcement Learning
Genetic Programming
Evolutionary algorithms
Evolutionary Algorithms
Graph in graph theory
Vertex of a graph
Dynamic Environment
Genetic algorithms
Data storage equipment
Processing
Implicit Function
Memory Function
Simulation Environment
Network Structure
Reuse

ASJC Scopus subject areas

  • Artificial Intelligence
  • Theoretical Computer Science
  • Computational Theory and Mathematics

これを引用

@article{1f25362d5955444087033e4dfa2b5a03,
title = "A graph-based evolutionary algorithm: Genetic Network Programming (GNP) and its extension using reinforcement learning",
abstract = "This paper proposes a graph-based evolutionary algorithm called Genetic Network Programming (GNP). Our goal is to develop GNP, which can deal with dynamic environments efficiently and effectively, based on the distinguished expression ability of the graph (network) structure. The characteristics of GNP are as follows. 1) GNP programs are composed of a number of nodes which execute simple judgment/processing, and these nodes are connected by directed links to each other. 2) The graph structure enables GNP to re-use nodes, thus the structure can be very compact. 3) The node transition of GNP is executed according to its node connections without any terminal nodes, thus the pst history of the node transition affects the current node to be used and this characteristic works as an implicit memory function. These structural characteristics are useful for dealing with dynamic environments. Furthermore, we propose an extended algorithm, {"}GNP with Reinforcement Learning (GNP-RL){"} which combines evolution and reinforcement learning in order to create effective graph structures and obtain better results in dynamic environments. In this paper, we applied GNP to the problem of determining agents' behavior to evaluate its effectiveness. Tileworld was used as the simulation environment. The results show some advantages for GNP over conventional methods.",
keywords = "Agent, Evolutionary computation, Graph structure, Reinforcement learning, Tileworld",
author = "Shingo Mabu and Kotaro Hirasawa and Takayuki Furuzuki",
year = "2007",
month = "9",
doi = "10.1162/evco.2007.15.3.369",
language = "English",
volume = "15",
pages = "369--398",
journal = "Evolutionary Computation",
issn = "1063-6560",
publisher = "MIT Press Journals",
number = "3",

}

TY - JOUR

T1 - A graph-based evolutionary algorithm

T2 - Genetic Network Programming (GNP) and its extension using reinforcement learning

AU - Mabu, Shingo

AU - Hirasawa, Kotaro

AU - Furuzuki, Takayuki

PY - 2007/9

Y1 - 2007/9

N2 - This paper proposes a graph-based evolutionary algorithm called Genetic Network Programming (GNP). Our goal is to develop GNP, which can deal with dynamic environments efficiently and effectively, based on the distinguished expression ability of the graph (network) structure. The characteristics of GNP are as follows. 1) GNP programs are composed of a number of nodes which execute simple judgment/processing, and these nodes are connected by directed links to each other. 2) The graph structure enables GNP to re-use nodes, thus the structure can be very compact. 3) The node transition of GNP is executed according to its node connections without any terminal nodes, thus the pst history of the node transition affects the current node to be used and this characteristic works as an implicit memory function. These structural characteristics are useful for dealing with dynamic environments. Furthermore, we propose an extended algorithm, "GNP with Reinforcement Learning (GNP-RL)" which combines evolution and reinforcement learning in order to create effective graph structures and obtain better results in dynamic environments. In this paper, we applied GNP to the problem of determining agents' behavior to evaluate its effectiveness. Tileworld was used as the simulation environment. The results show some advantages for GNP over conventional methods.

AB - This paper proposes a graph-based evolutionary algorithm called Genetic Network Programming (GNP). Our goal is to develop GNP, which can deal with dynamic environments efficiently and effectively, based on the distinguished expression ability of the graph (network) structure. The characteristics of GNP are as follows. 1) GNP programs are composed of a number of nodes which execute simple judgment/processing, and these nodes are connected by directed links to each other. 2) The graph structure enables GNP to re-use nodes, thus the structure can be very compact. 3) The node transition of GNP is executed according to its node connections without any terminal nodes, thus the pst history of the node transition affects the current node to be used and this characteristic works as an implicit memory function. These structural characteristics are useful for dealing with dynamic environments. Furthermore, we propose an extended algorithm, "GNP with Reinforcement Learning (GNP-RL)" which combines evolution and reinforcement learning in order to create effective graph structures and obtain better results in dynamic environments. In this paper, we applied GNP to the problem of determining agents' behavior to evaluate its effectiveness. Tileworld was used as the simulation environment. The results show some advantages for GNP over conventional methods.

KW - Agent

KW - Evolutionary computation

KW - Graph structure

KW - Reinforcement learning

KW - Tileworld

UR - http://www.scopus.com/inward/record.url?scp=35748930908&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=35748930908&partnerID=8YFLogxK

U2 - 10.1162/evco.2007.15.3.369

DO - 10.1162/evco.2007.15.3.369

M3 - Article

C2 - 17705783

AN - SCOPUS:35748930908

VL - 15

SP - 369

EP - 398

JO - Evolutionary Computation

JF - Evolutionary Computation

SN - 1063-6560

IS - 3

ER -