An almost sure central limit theorem for the overlap parameters in the Hopfield model

被引：7
作者：
Gentz, B
机构：
[1] Inst. für Angewandte Mathematik, Universität Zürich, CH-8057 Zürich
来源：
STOCHASTIC PROCESSES AND THEIR APPLICATIONS | 1996年 / 62卷 / 02期
关键词：
fluctuations; Hopfield model; overlap parameter; neural networks; Laplace's method;
D O I：
10.1016/0304-4149(96)00055-5
中图分类号：
O21 [概率论与数理统计]; C8 [统计学];
学科分类号：
020208 ; 070103 ; 0714 ;
摘要：
We consider the Hopfield model with a finite number of randomly chosen patterns above and below the critical temperature and prove an almost sure conditional central limit theorem for the vector of overlap parameters, For this purpose we analyse the almost sure asymptotic behaviour of the partition function.
引用
收藏
页码：243 / 262
页数：20
相关论文
共 10 条
[1]   RIGOROUS RESULTS FOR THE FREE-ENERGY IN THE HOPFIELD MODEL [J].
ALBEVERIO, S ;
TIROZZI, B ;
ZEGARLINSKI, B .
COMMUNICATIONS IN MATHEMATICAL PHYSICS, 1992, 150 (02) :337-373
[2]   SPIN-GLASS MODELS OF NEURAL NETWORKS [J].
AMIT, DJ ;
GUTFREUND, H .
PHYSICAL REVIEW A, 1985, 32 (02) :1007-1018
[3]   GIBBS-STATES OF THE HOPFIELD MODEL IN THE REGIME OF PERFECT MEMORY [J].
BOVIER, A ;
GAYRARD, V ;
PICCO, P .
PROBABILITY THEORY AND RELATED FIELDS, 1994, 100 (03) :329-363
[4]  
Chung KL., 1974, COURSE PROBABILITY T
[5]  
Ellis R., 2006, ENTROPY LARGE DEVIAT
[6]  
GENTZ B, 1996, THESIS U ZURICH
[7]   NEURAL NETWORKS AND PHYSICAL SYSTEMS WITH EMERGENT COLLECTIVE COMPUTATIONAL ABILITIES [J].
HOPFIELD, JJ .
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA-BIOLOGICAL SCIENCES, 1982, 79 (08) :2554-2558
[8]  
Kuelbs J., 1976, LECT NOTES MATH, V539
[9]   MIXING PROPERTIES, DIFFERENTIABILITY OF FREE-ENERGY AND CENTRAL LIMIT THEOREM FOR A PURE PHASE IN ISING-MODEL AT LOW-TEMPERATURE [J].
MARTINLOF, A .
COMMUNICATIONS IN MATHEMATICAL PHYSICS, 1973, 32 (01) :75-92
[10]  
McCulloch Warren S., 1943, BULL MATH BIOPHYS, V5, P115, DOI 10.1007/BF02478259
1