A LOWER BOUND FOR THE MEMORY CAPACITY IN THE POTTS-HOPFIELD MODEL

被引：9

作者：

FERRARI, PA

MARTINEZ, S

PICCO, P

机构：

[1] UNIV CHILE,DEPT INGN MATEMAT,SANTIAGO 3,CHILE

[2] CNRS,CTR PHYS THEOR,LAB PROPRE 7061,F-13288 MARSEILLE 9,FRANCE

来源：

JOURNAL OF STATISTICAL PHYSICS | 1992年 / 66卷 / 5-6期

关键词：

NEURAL NETWORK; HOPFIELD MODEL; POTTS MODEL;

D O I：

10.1007/BF01054440

中图分类号：

O4 [物理学];

学科分类号：

0702 ;

摘要：

We consider Potts-Hopfield networks of size N. We prove the result: reversed capital E-alpha(c) > 0 such that for all 0 < alpha < alpha(c) we can find delta, epsilon > 0 in such a way that, when N --> infinity, we can store alpha-N patterns, all of them being sorrounded by epsilon-energy barriers at distance delta.

引用

页码：1643 / 1652

页数：10

共 8 条

[1] SPIN-GLASS MODELS OF NEURAL NETWORKS [J].