关于知识的推理

近年来,关于知识的推理(Reasoning aboutKnowledge)正在成为计算机科学,特别是人工智能研究中的最重要的分支之一。自从八六年在美国加州召开首届关于知识的推理理论问题国际会议(简称 TARK:Theoretical Aspects of Reasoni-ng about Knowledge)以来,每两年将召开一次TARK 会议。今年三月份召开的 TARK90已是第三次盛会了。关于知识的推理的有关技术已在诸多的领域中得到应用,主要有,经济学,语言学,人工智能及计算机科学。这里,将简要地介绍关于知识的推理的有关基本知识、基本技术及其发展动向。一、关于知识的经典模型关于知识的逻辑推理(通常被称为认识逻辑(Epistemic Logics))的研究,最早由语言哲学家 J.Hintikka 在题为《知识和信念》书中提出了可能世界模型,其基本思路是     

知识与常识的表示和推理

一、引言无疑,常识(Common Knowledge)的表示和推理是知识领域中极其关键的研究问题,因为常识的特殊性质使它在多主体(Multi-agent)协同推理、通讯中起着重要的作用。可以看到,对于多主体的问题如三个聪明人问题,Conway问题乃至更一般的博奕理论,分布式处理等,都将涉及到超越个体知识和事实之上的更高层次的知识,这就是常识。常识     

面向知识图谱的知识推理研究进展

近年来,随着互联网技术和应用模式的迅猛发展,引发了互联网数据规模的爆炸式增长,其中包含大量有价值的知识.如何组织和表达这些知识,并对其进行深入计算和分析,备受关注.知识图谱作为丰富直观的知识表达方式应运而生.面向知识图谱的知识推理是知识图谱的研究热点之一,已在垂直搜索、智能问答等应用领域发挥了重要作用.面向知识图谱的知识推理旨在根据已有的知识推理出新的知识或识别错误的知识.不同于传统知识推理,由于知识图谱中知识表达形式的简洁直观、灵活丰富,面向知识图谱的知识推理方法也更加多样化.本文将从知识推理的基本概念出发,介绍近年来面向知识图谱知识推理方法的最新研究进展.具体地,本文根据推理类型划分,将面向知识图谱的知识推理分为单步推理和多步推理,根据方法的不同,每类又包括基于规则的推理、基于分布式表示的推理、基于神经网络的推理以及混合推理.本文详细总结这些方法,并探讨和展望面向知识图谱知识推理的未来研究方向和前景.     

基于神经网络的知识推理研究综述

知识推理是知识图谱补全的重要手段,一直以来都是知识图谱领域的研究热点之一。随着神经网络不断取得新的发展,其在知识推理中的应用在近几年逐渐得到广泛重视。基于神经网络的知识推理方法具备更强的推理能力和泛化能力,对知识库中实体、属性、关系和文本信息的利用率更高,推理效果更好。简要介绍知识图谱及知识图谱补全的相关概念,阐述知识推理的概念及基本原理,从语义、结构和辅助存储三个维度展开,综述当下基于神经网络的知识推理最新研究进展,总结了基于神经网络的知识推理在理论、算法和应用方面存在的问题和发展方向。     

大气污染总量控制规划智能决策支持系统中的知识表示和推理

在大气污染总量控制规划智能决策支持系统（ＩＳＡＰＥＣ）中知识表示方法采用产生式表示法,为提高推理的效率,将ＩＳＡＰＥＣ知识库设计为包含领域级和元级的两级知识库。为了便于用户和领域专家理解,又便于推理的实现,对这两级知识库中的知识都设计内、外两种形式,并且外部形式的知识可通过编译转化为内部形式的知识。针对领域级知识库分别设计了综合应用正向推理和逆向推理优点的目标推理机以及元推理机,此外还设计了推理中     

基于智能体的混合知识自适应推理控制

本文将传统的混合知识表示法推广，建立了一个将一般知识、案例知识、模型知识、模型知识及神经网络知识有机集成的结构．该种知识表示结构有助于知识的搜索、匹配和推理控制，解决了复杂问题的知识表示．为适应推理方法的需要，将各种不同的推理方法有机融合与集成，提出一种基于智能体的自适应推理控制结构，该种自适应推理控制结构对于改进解决复杂问题的效果以及提高解决复杂问题的效率具有重要意义。     

基于浅层知识和深层知识推理的故障诊断专家系统工具

以专家系统和诊断问题求解为基础，建立了基于浅层知识和深层知识推理相结合的故障诊断专家系统工具。文中论述了工具的结构、知识表示、参数设计、推理机制、黑板控制、学习机制和其它支持环境。     

一种基于关系型知识构造的高效推理平台

本文针对工程类知识的多样化特点及实时性、规模、易维护性和综合问题求解等系统实用化指标，研究并实现了一种基于关系型构造的新型工程知识推理平台ＲＥＫＰ。文中论述的工程短短的关系模型及相应的知识群机制，使系统既便于知识维护和实现海量存储，又使问题求解达到高效率。性能测试与综合比较表明ＲＥＫＰ在实时性、存储、知识管理和易维护性等方面具有综合优势。     

关于知识的推理新进展

本文由上篇和下篇两个部分组成。上篇主要阐述人工智能理论研究的意义和作用，以消除人们通常对理论性研究所产生的一些误解；同时也进一步对关于知识的推理研究的意义进行分析。下篇主要介绍最近召开的关于知识推理理论问题的国际会议情况，着重其在计算机科学和人工智能领域方面的国际新动态和新进展。     

可重构知识管理系统中案例与规则的集成推理方法

传统上的知识管理工具往往在设计时就固定了知识结构,这样的系统不但缺乏通用性,而且也限制了知识检索与推理的效率。介绍一个基于本体的可重构知识管理系统,知识作为本体概念的对象实例,利用本体模型的可定制性,解决了以往知识类型不能扩展的问题。详细阐述了结合案例推理与规则推理的集成推理方法,通过规则学习算法支持了规则库的动态扩充与调整,并将本体类的语义关系应用于推理方法,进一步提高了推理的效率。最后介绍了该系统在某飞机设计研究院的应用情况和今后的研究方向。     

Extending Answer Sets for Logic Programming Agents

We present systems of logic programming agents (LPAS) to model the interactions between decision-makers while evolving to a conclusion. Such a system consists of a number of agents connected by means of unidirectional communication channels. Agents communicate with each other by passing answer sets obtained by updating the information received from connected agents with their own private information. We introduce a credulous answer set semantics for logic programming agents. As an application, we show how extensive games with perfect information can be conveniently represented as logic programming agent systems, where each agent embodies the reasoning of a game player, such that the equilibria of the game correspond with the semantics agreed upon by the agents in the LPAS.     

A characterization of answer sets for logic programs

Checking if a program has an answer set, and if so, compute its answer sets are just some of the important problems in answer set logic programming. Solving these problems using Gelfond and Lifschitz's original definition of answer sets is not an easy task. Alternative characterizations of answer sets for nested logic pro- grams by Erdem and Lifschitz, Lee and Lifschitz, and You et al. are based on the completion semantics and various notions of tightness. However, the notion of tightness is a local notion in the sense that for different answer sets there are, in general, different level mappings capturing their tightness. This makes it hard to be used in the design of algorithms for computing answer sets. This paper proposes a characterization of answer sets based on sets of generating rules. From this char- acterization new algorithms are derived for computing answer sets and for per- forming some other reasoning tasks. As an application of the characterization a sufficient and necessary condition for the equivalence between answer set seman- tics and completion semantics has been proven, and a basic theorem is shown on computing answer sets for nested logic programs based on an extended notion of loop formulas. These results on tightness and loop formulas are more general than that in You and Lin's work.     

Probabilistic logic programming

Of all scientific investigations into reasoning with uncertainty and chance, probability theory is perhaps the best understood paradigm. Nevertheless, all studies conducted thus far into the semantics of quantitative logic programming have restricted themselves to non-probabilistic semantic characterizations. In this paper, we take a few steps towards rectifying this situation. We define a logic programming language that is syntactically similar to the annotated logics of Blair et al., 1987 and Blair and Subrahmanian, 1988, 45–73) but in which the truth values are interpreted probabilistically. A probabilistic model theory and fixpoint theory is developed for such programs. This probabilistic model theory satisfies the requirements proposed by Fenstad (in “Studies in Inductive Logic and Probabilities” (R. C. Jeffrey, Ed.), Vol. 2, pp. 251–262, Univ. of California Press, Berkeley, 1980) for a function to be called probabilistic. The logical treatment of probabilities is complicated by two facts: first, that the connectives cannot be interpreted truth-functionally when truth values are regarded as probabilities; second, that negation-free definite-clause-like sentences can be inconsistent when interpreted probabilistically. We address these issues here and propose a formalism for probabilistic reasoning in logic programming. To our knowledge, this is the first probabilistic characterization of logic programming semantics.     

On finite-state approximants for probabilistic computation tree logic

We generalize the familiar semantics for probabilistic computation tree logic from finite-state to infinite-state labelled Markov chains such that formulas are interpreted as measurable sets. Then we show how to synthesize finite-state abstractions which are sound for full probabilistic computation tree logic and in which measures are approximated by monotone set functions. This synthesis of sound finite-state approximants also applies to finite-state systems and is a probabilistic analogue of predicate abstraction. Sufficient and always realizable conditions are identified for obtaining optimal such abstractions for probabilistic propositional modal logic.     

多Agent信念逻辑及其在概率意义下的推广

首先建立了一种多Agent信念逻辑MBL(multi-agentbelieflogic),在经典信念逻辑基础上增加了普遍信念算子和公共信念算子,给出MBL的Kripke语义与广义Aumann语义,讨论了两者的等价性,证明了MBL对于上述两种语义的可靠性和完备性.其次,建立了一种多Agent概率信念逻辑MPBL(multi-agentprobabilisticbelieflogic),通过在广义Aumann语义基础上引入概率空间,给出了MPBL的概率Aumann语义,证明了它的可靠性,并给出MPBL的一些推论.     

集合代数是经典命题演算形式系统的语义解释

经典命题演算形式系统(CPC)中的公式只是一些形式符号,其意义是由具体的解释给出的.逻辑代数和集合代数都是布尔代数,都是CPC的解释.集合代数是CPC的集合语义,其中对联结词的解释就是集合运算;对形式公式的解释就是集合函数;对逻辑蕴涵.逻辑等价的解释就是集合包含和集合相等=.标准概率逻辑是在标准概率空间上建立的逻辑体系,命题表示随机事件,随机事件是集合,概率空间中的事件域是集合代数,概率逻辑就是CPC集合语义的实际应用.CPC完全适用于概率命题演算.     

A new approach to hybrid probabilistic logic programs

This paper presents a novel revision of the framework of Hybrid Probabilistic Logic Programming, along with a complete semantics characterization, to enable the encoding of and reasoning about real-world applications. The language of Hybrid Probabilistic Logic Programs framework is extended to allow the use of non-monotonic negation, and two alternative semantical characterizations are defined: stable probabilistic model semantics and probabilistic well-founded semantics. These semantics generalize the stable model semantics and well-founded semantics of traditional normal logic programs, and they reduce to the semantics of Hybrid Probabilistic Logic programs for programs without negation. It is the first time that two different semantics for Hybrid Probabilistic Programs with non-monotonic negation as well as their relationships are described. This proposal provides the foundational grounds for developing computational methods for implementing the proposed semantics. Furthermore, it makes it clearer how to characterize non-monotonic negation in probabilistic logic programming frameworks for commonsense reasoning. An erratum to this article can be found at     

Exogenous Probabilistic Computation Tree Logic

We define a logic EpCTL for reasoning about the evolution of probabilistic systems. System states correspond to probability distributions over classical states and the system evolution is modelled by probabilistic Kripke structures that capture both stochastic and non–deterministic transitions. The proposed logic is a temporal enrichment of Exogenous Probabilistic Propositional Logic (EPPL). The model-checking problem for EpCTL is analysed and the logic is compared with PCTL; the semantics of the former is defined in terms of probability distributions over sets of propositional symbols, whereas the latter is designed for reasoning about distributions over paths of possible behaviour. The intended application of the logic is as a specification formalism for properties of communication protocols, and security protocols in particular; to demonstrate this, we specify relevant security properties for a classical contract signing protocol and for the so–called quantum one–time pad.