PICASSO——一个用于模糊决策系统的基于规则的程序设计语言

PICASSO-a Rule-based Programming Language for Fuzzy Decision Support Systems

摘要:

本论文介绍一种新的基于规则的程序设计语言,其名为PICASSO,它为具体执行模糊决策支持系统(Fuzzy Decision SupportSystem)而设计。PICASSO已在美国休斯顿大学计算中心的AT&T公司的3B2OS小型计算机上部分运行。整个系统用Franz Lisp书写。
PICASSO具有以下特征。它使用正向链接(Forward chaining)推理手段。该语言有三种控制策略:面向宽度优先的系统控制模式,用户控制模式和元规则(meta-rule)^*2控制模式。PICASSO支持匹配变量以及其他类型的变量,以便于使用以模式匹配为主的调用方式。在该语言中,不同规则之间的信息能够用信息传送和数据共享两种方法来进行通讯、交换。PICASSO的规则工作在可永久保存的知识库上,这个知识库必须预先用称为符号图形(S-diagram)的一种强有力的数据模式来定义。PICASSO的推理器(inference engine)实施知识库的一致性。知识库中可以保存模糊信息(fuzzy information),这些信息用区间模式的二值法来表示。区间模式允许分配一个概率给一个事实(断言),同时还要表示我们相信这种概率估计的程度,这一点对于模糊决策支持系统来说是重要的,因为这种系统常牵涉到不同程度的经验知识的判定规则。有了对某些事实(断言)的概率估计,又有专家对此概率的相信程度(凭经验),就可较全面地选择某种决策。对于区间模式来说,我们已经提供了一些特定的运算公式,用于对不确定性(uncertainty)知识进行自动推理。PICASSO支持基于规则的程序设计语言和函数型的程序设计语言两者的组合并支持多个基于规则的程序设计语言和函数型的程序设计语言两者的组合,并支持多个基于规则的程序之间的协同操作:可以容易地把PICASSO程序综合成象一般的LISP函数那样的LISP程序,而且PICASSO程序本身也可以任意地调用LISP函数。

Abstract:

The use of rule basedproduction system oriented programming paradigmg has gained a large popularity in the last 5 years,especially for developin computerized expert systems.This paper introduces a new rulebased programming language called PICASSO that has been tailored for the mplementation of fuzzy decision support systems.PICASSO can be characterized as follows:It uses a forward chaining inference approach.The language give its user a choice of three different control strategies:breadth first oriented, user controlled and meta-rule controlled.PICASSO facilitates the formulation of pattern directed invocation schemas by supporting match-variables as well as other sorts of variables.Information can be exchanged between rules using message passing as well as data sharing.PICASSO rules operate on permanent knowledge bases that have to be defined in advance by using a powerful data model called S-diagram.The PICASSO inference engine enforces the consistency of the knowledge base.The knowledge base may contain fuzzy information that is represented using a two-valued approach called interval model that allows to assign a probability to a predicate as well as to express our strength of belief in this estimation,which is important for fuzzy decision support systems in which frequently involves rules that have very different degrees of empirical conformation.Special operators for automating reasoning in uncertainty have been provided for the interval model.PICASSO supports the combination of rule based and functional programming as well as the coorporation between multiple rule based programs:PICASSO-programs can easily be integrated to LISP-programs as ordinary LISP-functions and PICASSO-programs themself can call arbitrary LISP-functions.