EMERGENT BEHAVIOR IN CLASSIFIER SYSTEMS

被引:11
作者
FORREST, S
MILLER, JH
机构
[1] UNIV CALIF LOS ALAMOS SCI LAB,DIV COMP,LOS ALAMOS,NM 87545
[2] CARNEGIE MELLON UNIV,SANTA FE INST,PITTSBURGH,PA 15213
来源
PHYSICA D | 1990年 / 42卷 / 1-3期
关键词
D O I
10.1016/0167-2789(90)90075-Z
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
The paper presents examples of emergent behavior in classifier systems, focusing on symbolic reasoning and learning. These behaviors are related to global dynamical properties such as state cycles, basins of attraction, and phase transitions. A mapping is defined between classifier systems and an equivalent dynamical system (Boolean networks). The mapping provides a way to understand and predict emergent classifier system behaviors by observing the dynamical behavior of the Boolean networks. The paper reports initial results and discusses the implications of this approach for classifier systems. © 1990.
引用
收藏
页码:213 / 227
页数:15
相关论文
共 25 条
[1]  
ARTHUR B, 1989, UNPUB CLASSIFIER SYS
[2]  
BELEW RK, 1988, MACHINE LEARNING, V3, P192
[3]  
BOOKER LB, IN PRESS ADV NEURAL
[4]   LEARNING AND BUCKET BRIGADE DYNAMICS IN CLASSIFIER SYSTEMS [J].
COMPIANI, M ;
MONTANARI, D ;
SERRA, R .
PHYSICA D, 1990, 42 (1-3) :202-212
[5]  
COMPIANI M, 1989, 3RD P INT C GEN ALG, P298
[6]   EVOLUTION OF OVERLAPS BETWEEN CONFIGURATIONS IN RANDOM BOOLEAN NETWORKS [J].
DERRIDA, B ;
WEISBUCH, G .
JOURNAL DE PHYSIQUE, 1986, 47 (08) :1297-1303
[7]   PHASE-TRANSITIONS IN TWO-DIMENSIONAL KAUFFMAN CELLULAR AUTOMATA [J].
DERRIDA, B ;
STAUFFER, D .
EUROPHYSICS LETTERS, 1986, 2 (10) :739-745
[8]  
FORREST S, 1989, PHYSICA D, V42, P1
[9]  
FORREST S, 1985, THESIS U MICHIGAN
[10]  
Goldberg D. E., 1988, TCGA88002 U AL DEP E