Expert Systems

28
Expert Systems Expert Systems Directors : Prof. Zixing Cai &Miss WenSha Central South University College of Information Science and Engineering

description

 

Transcript of Expert Systems

Page 1: Expert Systems

Expert SystemsExpert Systems

Directors : Prof. Zixing Cai &Miss WenSha

Central South UniversityCollege of Information Science and Engineering

Page 2: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Experts are people who are very familiar with solving specific types of problems.Expert System

Until now, no unified definition has been given.Knowledge-based system

The fundamental function of the expert system depends upon its knowledge, therefore, the expert system is sometimes called knowledge-based system.

What is an Expert System?What is an Expert System?

Page 3: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Definition 1: ES can handle real-world complex problems which need an expert’s interpretation and solve problems by using a computer model of human expert reasoning to reach the same conclusions that the human expert would do if he or she faces with a comparable problem.Definition 2: ES is an intelligent computer program that uses knowledge and inference procedures to solve problems that are difficult enough to require significant human expertise for their solutions.

What is an Expert System(ES)?What is an Expert System(ES)?

In short, an ES is an intelligent computer program that can perform special and

difficult task(s) in some field(s) at the level of human

experts.

Page 4: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Architecture of ideal expert Architecture of ideal expert systemsystem

Architecture of an ideal expert system

User

Communication Interface Knowledge

Base

Interpreter

Blackboard

Plan

Agenda

Solution

ReasoningMachine

Coordinator

Planner

Adjuster

Page 5: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

ES-Knowledge Base(1)ES-Knowledge Base(1)

Knowledge Base To store knowledge from the

experts of special field(s). It contains facts and feasible operators or rules for heuristic planning and problem solving.

The other data is stored in a separate database called global database, or database simply.

Page 6: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

ES-Reasoning Machine(2)ES-Reasoning Machine(2)

Reasoning Machine To memorize the reasoning rules

and the control strategies applied. According to the information

from the knowledge base, the reasoning machine can coordinate the whole system in a logical manner, draw inference and make a decision.

Page 7: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

ES- User Interface (3)ES- User Interface (3)

User Interface To communicate between the

user and the expert system. The user interacts with the expert

system in problem-oriented language such as in restricted English, graphics or a structure editor. The interface mediates information exchanges between the expert system and the human user.

Page 8: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

ES- Interpreter(4)ES- Interpreter(4)

Interpreter Through the user interface,

interpreter explains user questions, commands and other information generated by the expert system, including answers to questions, explanations and justifications for its behavior, and requests for data.

Page 9: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

ES-Blackboard (5)ES-Blackboard (5)

Blackboard To record intermediate

hypotheses and decisions that the expert system manipulates.

Page 10: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

ES-NoteES-Note

Note: Almost no exiting expert system

contains all the components shown above, but some components, especially the knowledge base and reasoning machine, occur in almost all expert systems.

Many ESs use global database in place of the blackboard. The global database contains information related to specific tasks and the current state.

Page 11: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Building Expert SystemBuilding Expert System

The key for successfully building an expert system is to begin it from a smaller one, and extend and test it step by step, make it into a larger-scale and more perfect system.The general procedure for building ESs :

Design of initial Knowledge BaseDevelopment & test for prototype 原型

systemImprovement & induction 归纳 for the

knowledge

Page 12: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Design of initial Knowledge Design of initial Knowledge Base Base

Problem identificationKnowledge conceptualizationConcept formulizationRule formulationRule validation

Page 13: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Questions

Indentifi-cation

Indentifi-cation

Conceptu-alization

Conceptu-alization

Formali-zation

Formali-zation

RuleFormalization

RuleFormalization ValidationValidation

Knowledge

Concepts

StructureRules

ConclusionRepresentation

Refinements

Re-designment

Stages for designing knowledge base

Stages for Designing KBStages for Designing KB

Concepts

identify what the problem is , how to define it , can

we divide it into some sub problems

define key concept of the knowledge ,for example :

type of data structure , conditions that have known, the goal state, assumption

and control strategy.

use knowledge representation

method to represent the knowledge.

change the knowledge to programming language that can be identified by

the computer.

check the correctness of

rules or knowledge

Page 14: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Category Problem Addressed

InterpretationPredictionDiagnosisDesignPlanningMonitoringDebuggingRepairInstruction

Control

Inferring situation descriptions from sensor dataInferring likely consequences of given situationInferring system malfunction from observationConfiguring objects under constrainsDesigning actionsComparing observation to plan vulnerabilitiesPrescribing remedies for malfunctionExecuting a plan to administer a prescribed remedyDiagnosing, debugging and repairing student behaviorInterpreting, predicting, repairing and monitoring system behavior

Types of Expert System (ES)

Page 15: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Expert Control SystemsExpert Control Systems

Important differences between expert systems and expert control systems:

Expert systems simply complete consultative function for problems of special domains and aid users to work.

Expert control systems need to make decisions to control action independently and automatically.

Expert systems usually work in off-line mode. Expert control systems need to acquire dynamic information in on-line mode and make real-time control for the system.

Page 16: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Two main types of expert Two main types of expert controlcontrol

Two main types of expert control: Expert control system With a more complex structure, higher

cost, better performance, and used to plants or processes where higher technical requirements are needed.

Expert controller With a simpler structure, lower cost and

has a performance that can meet the general requirements for the industrial process control.

Page 17: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Structures of Expert Control Structures of Expert Control SystemSystem

Reasoning Machine

Operator Interface

Knowledge Base Digital Processing

DomainDate

A/D

SensorsProcessActuators

D/A

ControlAlgorithms

Controller

A typical structure of expert control system

Page 18: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Tasks of Expert Control System Tasks of Expert Control System

The expert control system should execute following tasks:

Supervise the operation of the plant (process) and controller.

Examine possible failure or fault of the system components, replace these faulty components or revise control algorithms to keep the necessary performance of the system.

In special cases, select suitable control algorithm to adapt the variation of the system parameters and environment.

Page 19: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Knowledge Base (KB)

Inference Engine (IE)

Set ofControl Rules

Feature Recognitio

n Informatio

n Processing

Plant

Sensor(s)

e

u

R-

S I

K G

U Y

Expert Controller

Store the domain knowledge of industrial

process control,experience of experts(expertise) and

factsExtract and process information, provide

control strategy and learn adaptation with foundation

Sum up every control pattern and control experience of the controlled process

Use the forward chaining reasoning to judge the

conditions of every rule in the sequence

Industrial expert controller

Page 20: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Expert system-MYCINExpert system-MYCIN

An early expert system developed in early 1970s at Stanford UniversityWrote by Lisp LanguageAuthor: Bruce G. Buchanan & Edward H. Shortliffe

<<Rule-Based Expert Systems:The MYCIN Experiments of the Stanford Heuristic Programming Project >>

This expert system was designed to identify bacteria causing severe infections

Page 21: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Page 22: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Reasoning & Problem solving Reasoning & Problem solving strategystrategy

MYCIN could use backward chaining to find out whether a possible bacteria was to blame.“Certainty factor” is used for an assessment of the likelihood 可能性评估 of one bacteria.MYCIN’s problem solving strategy was simple:

For each possible bacteria: Using backward chaining, try to prove that it is the case, finding the certainty.Find a treatment which ” covers” all the bacteria above some level of certainty.

Page 23: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

MYCIN: Problem SolvingMYCIN: Problem Solving

When trying to prove a goal through backward chaining, system could ask user certain questions.Certain facts are marked as “askable”, so if they couldn’t be proved, ask the user.The ask procedure is carried out in following style of dialogue:

MYCIN: Has the patient had neurosurgery?USER: No.MYCIN: IS the patient a burn patient?USER: No.…MYCIN: It could be Diplococcus..

Page 24: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Modeling Human Diagnostic Modeling Human Diagnostic StrategiesStrategies

Problem Solving Strategy used in MYCIN only works when small number of hypotheses (e.g., bacteria).For hundreds of possible diseases, need a better strategy.Later medical diagnostic systems used an approach based on human expert reasoning.

Page 25: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Diagnostic Reasoning: InternistDiagnostic Reasoning: Internist

Internist is a medical expert system for general disease diagnosis.Knowledge in system consists of disease profiles 概况 , giving symptoms症状 associated with disease and strength of association.

Page 26: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Problem Solving in InternistProblem Solving in Internist

Use initial data (symptoms) to suggest, or trigger 引发 possible diseases.Determine what other symptoms would be expected to confirm these diseases.Gather more data to differentiate 区分 between these hypotheses. Either:

If one hypothesis most likely, try to confirm it.If many possible hypotheses, try to throw some out.If a few hypotheses, try to discriminate 区别 between them.

Page 27: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Medical Expert Systems TodayMedical Expert Systems Today

Medical expert systems were quite effective in evaluations comparing their performance with human experts.Support the physicians 医生 decisions, rather than doing the whole diagnosis.Include many useful support materials 辅助材料 , such as report generating tools, reference material etc.

Page 28: Expert Systems

Central South University Artificial IntelligenceCentral South University Artificial Intelligence

                                                                                                                                                                                                                                                                                                      

                  

CISIC

CISIC

Summary: Expert SystemsSummary: Expert Systems

Effective systems have been developed that capture expert knowledge in areas like medicine.Typically combine rule-based approaches, with additional certainty/probabalistic reasoning, and some top level control of the problem solving process.Not a huge take-up of systems, perhaps due to failure to adequately consider how they would be integrated into current practice.