Professional Documents
Culture Documents
CONTENTS
• But GE wanted
• A more effective and dependable way to disseminate expertise.
• To prevent valuable knowledge from retiring.
• To minimize extensive travel or moving the locomotives.
• The expert system solution
• To model the way a human troubleshooter works
Months of knowledge acquisition
3 years of prototyping
• A novice engineer or technician can perform at an expert’s level
On a personal computer
Installed at every railroad repair shop served by GE
An expert system (ES) is a computer program that uses knowledge and inference procedures of an
expert to solve problems.
• Expertise is the extensive, task-specific knowledge acquired from training, reading, and
experience.
Includes:
• theories about the problem area
• rules and procedures
• heuristics
• global strategies
• meta-knowledge
• facts
• Heuristics are rules of experience that characterise expert-level decision making in the field.
Basic Concepts of Expert Systems
Knowledge engineering - the task of eliciting and modelling the knowledge and building a
computer system
•
• Transferring expertise:
expert computer non-expert
• knowledge acquisition (from experts and other sources)
• knowledge representation (in the computer)
• knowledge inferencing
• knowledge transfer to the user
•
• The human elements in expert systems:
• expert
• knowledge engineer
• user
• Explanation:
• HOW was a particular conclusion reached?
• WHY the program asks the user a particular question?
• TRACE displays all rules that are tried (not only those that are executed).
• WHAT-IF explains what will happen if a certain value or rule is changed.
Production Systems
example
The Water Jug Problem
two jugs: 4 gallon, 3 gallon
task: to get exactly 2 gallons of water in the 4 gallon jug
State space
(x, y) where x =0, 1, 2, 3, 4
y=0, 1, 2, 3
initial state: (0, 0)
goal state (2, y)
production rule:
IF there are less than 4 gallons in the 4 gallon jug
THEN fill the 4 gallon jug
IF (x, y) where x < 4 THEN (4, y)
(1) Working memory holds the symbol structure which represents the states of the problem
space.
(2) Production memory (long-term memory) holds the production rules.
(3) Control strategy selects the rules to be applied.
Conflict resolution
principles:
• ordering of the rules
• recency
• specificity
SUMMARY
Phase 4: Testing
As the PDAMum will need to be periodically tested and evaluated to assure that its performance is
converging toward established goals. It is important that these decisions be made early, at a time
when the original project goals are established.
These are the evolution of testing or evaluation of testing for PDAMum:
Stage 1: Preliminary Testing
• Study the complete knowledge base
• Uncover deficiencies in the knowledge and reasoning strategies
• Validate knowledge representation and inference approach
Stage 2: Demonstration Testing
• Choose a problem of limited scope within the capabilities of PDAMum
• Use demonstration to validate the expert system approach
• Show off major PDAMum features
• Design interface to accommodate of the user
Stage 3: Informal Validation Testing
• Select typical past test cases
• Evaluate PDAMum’s ability in solving typical cases and
• Identify PDAMum deficiencies and obtain comments from user on the interface
Stage 4: Refinement Testing
• Select unusual past test cases
• Evaluate PDAMum’s ability in solving unusual cases
• Uncover deficiencies in PDAMum’s knowledge and control
• Identify PDAMum deficiencies
Stage 5: Formal Testing
• Select past test cases and define test criteria
• Run the PDAMum for each test case and ask evaluators to judge system’s performance for
each test case
• Obtain comments on the interface
• Identify PDAMum strength and deficiencies
Stage 6: Field Testing
• Define test criteria for the field test
• Determine if PDAMum meets its original goals when applied to real problems
Phase 5: Documentation
This documentation serves as a personal diary of the project. It contains all the material collected
during the project that needs to be reference for developing the system. If properly designed, it will
also serve the later tasks of maintaining PDAMum and writing the PDAMum’s final report.
Phase 6: Maintenance
After finished all of the designing, implementing, testing and documentation, PDAMum may need
to be refined or updated to meet current needs. It is extremely important to keep good records on
any changes made to PDAMum. If this isn’t done, it is very easy to lose track of the PDAMum’s
knowledge. And each time PDAMum is modified; the following critical pieces of information
should be documented:
• What was modified and who performed the modification
• When the modification was made
• Why the modification was made
References
Aamodt, A., & Plaza, E., Case-Based Reasoning: Foundational issues, methodological variations,
and system approaches, AICommunications, IOS Press,
http://www.iiia.csic.es/People/enric/AICom.pdf
Cawsey A., “The Essence of Artificial Intelligence”, Prentice Hall Europe, 1998.
“Computerize Prescription Reduce Errors”, Electronic Medical Records, Expert System
Application, Inc.
[online: www.emrs.com/PrescriptionWritingReducedErrors.htm ]
Durkin, J., “Expert System Design and Development”, New Jersey: Prentice-Hall. 1994.
George, F. Luger & William, A. Stubblefield, “Artificial Intelligence: Structures and Strategies for
Complex Problem Solving”, Addison Wesley Longman, Inc., 1998.
H. Basri, “An Expert System for Planning Landfill Restoration”, Water Science and Technology,
1998.
Hevner, et al., “Design Research in Information System”, 2004, http:// www.isworld.org
Richard E. Plant, Nicholas D. Stave, “Knowledge based systems in Agriculture”, McGraw-
Hill, 1991.
R. L. Hoskinson, J. R. Hess, R. K. Fink, “A Decision Support System for Optimum Use of
Fertilizers”, 1992.
Rule-Based ES's in Medicine, http://alpha.cbmi.upmc.edu/courses/fall97/Sep25/ index.htm
ARTIFICIAL INTELLIGENCE