Generalized classifier system: Evolving classifiers with cyclic conditions

Xianneng Li, Wen He, Kotaro Hirasawa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Accuracy-based XCS classifier system has been shown to evolve classifiers with accurate and maximally general characteristics. XCS generally represents its classifiers with binary conditions encoded in a ternary alphabet, i.e., {0,1, #}, where # is a 'don't care' symbol, which can match with 0 and 1 in inputs. This provides one of the foundations to make XCS evolve an optimal population of classifiers, where each classifier has the possibility to cover a set of perceptions. However, when performing XCS to solve the multi-step problems, i.e., maze control problems, the classifiers only allow the agent to perceive its surrounding environments without the direction information, which are contrary to our human perception. This paper develops an extension of XCS by introducing cyclic conditions to represent the classifiers. The proposed system, named generalized XCS classifier system (GXCS), is dedicated to modify the forms of the classifiers from chains to cycles, which allows them to match with more adjacent environments perceived by the agent from different directions. Accordingly, a more compact population of classifiers can be evolved to perform the generalization feature of GXCS. As a first step of this research, GXCS has been tested on the benchmark maze control problems in which the agent can perceive its 8 surrounding cells. It is confirmed that GXCS can evolve the classifiers with cyclic conditions to successfully solve the problems as XCS, but with much smaller population size.

Original languageEnglish
Title of host publicationProceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1682-1689
Number of pages8
ISBN (Print)9781479914883
DOIs
Publication statusPublished - 2014 Sep 16
Event2014 IEEE Congress on Evolutionary Computation, CEC 2014 - Beijing
Duration: 2014 Jul 62014 Jul 11

Other

Other2014 IEEE Congress on Evolutionary Computation, CEC 2014
CityBeijing
Period14/7/614/7/11

Fingerprint

Classifiers
Classifier
Control Problem
Human Perception
Population Size
Ternary
Adjacent
Cover
Binary
Benchmark
Cycle

ASJC Scopus subject areas

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

Cite this

Li, X., He, W., & Hirasawa, K. (2014). Generalized classifier system: Evolving classifiers with cyclic conditions. In Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014 (pp. 1682-1689). [6900457] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CEC.2014.6900457

Generalized classifier system : Evolving classifiers with cyclic conditions. / Li, Xianneng; He, Wen; Hirasawa, Kotaro.

Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014. Institute of Electrical and Electronics Engineers Inc., 2014. p. 1682-1689 6900457.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Li, X, He, W & Hirasawa, K 2014, Generalized classifier system: Evolving classifiers with cyclic conditions. in Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014., 6900457, Institute of Electrical and Electronics Engineers Inc., pp. 1682-1689, 2014 IEEE Congress on Evolutionary Computation, CEC 2014, Beijing, 14/7/6. https://doi.org/10.1109/CEC.2014.6900457
Li X, He W, Hirasawa K. Generalized classifier system: Evolving classifiers with cyclic conditions. In Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014. Institute of Electrical and Electronics Engineers Inc. 2014. p. 1682-1689. 6900457 https://doi.org/10.1109/CEC.2014.6900457
Li, Xianneng ; He, Wen ; Hirasawa, Kotaro. / Generalized classifier system : Evolving classifiers with cyclic conditions. Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014. Institute of Electrical and Electronics Engineers Inc., 2014. pp. 1682-1689
@inproceedings{ae28090aae63443c8fda43f6fbc9d9f6,
title = "Generalized classifier system: Evolving classifiers with cyclic conditions",
abstract = "Accuracy-based XCS classifier system has been shown to evolve classifiers with accurate and maximally general characteristics. XCS generally represents its classifiers with binary conditions encoded in a ternary alphabet, i.e., {0,1, #}, where # is a 'don't care' symbol, which can match with 0 and 1 in inputs. This provides one of the foundations to make XCS evolve an optimal population of classifiers, where each classifier has the possibility to cover a set of perceptions. However, when performing XCS to solve the multi-step problems, i.e., maze control problems, the classifiers only allow the agent to perceive its surrounding environments without the direction information, which are contrary to our human perception. This paper develops an extension of XCS by introducing cyclic conditions to represent the classifiers. The proposed system, named generalized XCS classifier system (GXCS), is dedicated to modify the forms of the classifiers from chains to cycles, which allows them to match with more adjacent environments perceived by the agent from different directions. Accordingly, a more compact population of classifiers can be evolved to perform the generalization feature of GXCS. As a first step of this research, GXCS has been tested on the benchmark maze control problems in which the agent can perceive its 8 surrounding cells. It is confirmed that GXCS can evolve the classifiers with cyclic conditions to successfully solve the problems as XCS, but with much smaller population size.",
author = "Xianneng Li and Wen He and Kotaro Hirasawa",
year = "2014",
month = "9",
day = "16",
doi = "10.1109/CEC.2014.6900457",
language = "English",
isbn = "9781479914883",
pages = "1682--1689",
booktitle = "Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - Generalized classifier system

T2 - Evolving classifiers with cyclic conditions

AU - Li, Xianneng

AU - He, Wen

AU - Hirasawa, Kotaro

PY - 2014/9/16

Y1 - 2014/9/16

N2 - Accuracy-based XCS classifier system has been shown to evolve classifiers with accurate and maximally general characteristics. XCS generally represents its classifiers with binary conditions encoded in a ternary alphabet, i.e., {0,1, #}, where # is a 'don't care' symbol, which can match with 0 and 1 in inputs. This provides one of the foundations to make XCS evolve an optimal population of classifiers, where each classifier has the possibility to cover a set of perceptions. However, when performing XCS to solve the multi-step problems, i.e., maze control problems, the classifiers only allow the agent to perceive its surrounding environments without the direction information, which are contrary to our human perception. This paper develops an extension of XCS by introducing cyclic conditions to represent the classifiers. The proposed system, named generalized XCS classifier system (GXCS), is dedicated to modify the forms of the classifiers from chains to cycles, which allows them to match with more adjacent environments perceived by the agent from different directions. Accordingly, a more compact population of classifiers can be evolved to perform the generalization feature of GXCS. As a first step of this research, GXCS has been tested on the benchmark maze control problems in which the agent can perceive its 8 surrounding cells. It is confirmed that GXCS can evolve the classifiers with cyclic conditions to successfully solve the problems as XCS, but with much smaller population size.

AB - Accuracy-based XCS classifier system has been shown to evolve classifiers with accurate and maximally general characteristics. XCS generally represents its classifiers with binary conditions encoded in a ternary alphabet, i.e., {0,1, #}, where # is a 'don't care' symbol, which can match with 0 and 1 in inputs. This provides one of the foundations to make XCS evolve an optimal population of classifiers, where each classifier has the possibility to cover a set of perceptions. However, when performing XCS to solve the multi-step problems, i.e., maze control problems, the classifiers only allow the agent to perceive its surrounding environments without the direction information, which are contrary to our human perception. This paper develops an extension of XCS by introducing cyclic conditions to represent the classifiers. The proposed system, named generalized XCS classifier system (GXCS), is dedicated to modify the forms of the classifiers from chains to cycles, which allows them to match with more adjacent environments perceived by the agent from different directions. Accordingly, a more compact population of classifiers can be evolved to perform the generalization feature of GXCS. As a first step of this research, GXCS has been tested on the benchmark maze control problems in which the agent can perceive its 8 surrounding cells. It is confirmed that GXCS can evolve the classifiers with cyclic conditions to successfully solve the problems as XCS, but with much smaller population size.

UR - http://www.scopus.com/inward/record.url?scp=84908574593&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84908574593&partnerID=8YFLogxK

U2 - 10.1109/CEC.2014.6900457

DO - 10.1109/CEC.2014.6900457

M3 - Conference contribution

AN - SCOPUS:84908574593

SN - 9781479914883

SP - 1682

EP - 1689

BT - Proceedings of the 2014 IEEE Congress on Evolutionary Computation, CEC 2014

PB - Institute of Electrical and Electronics Engineers Inc.

ER -