Introduction to RuleBased Systems Expert Systems Fuzzy Systems

Introduction to Rule-Based Systems, Expert Systems, Fuzzy Systems (sections 2. 7, 2. 8, pp. 118 -132; and partially 3. 1 and 3. 4) N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Sub-topics: n n n n Production rules and production systems How to program in rules? Advantages and limitations of the production systems Expert systems Fuzzy sets Fuzzy rules and fuzzy inference Fuzzy information retrieval and fuzzy databases Fuzzy expert systems N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Production Rules and Production Systems n n n A production rule consists of two parts: condition (antecedent) part and conclusion (action, consequent) part, i. e: IF (conditions) THEN (actions) Example IF Gauge is OK AND [TEMPERATURE] > 120 THEN Cooling system is in the state of overheating N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Production Rules and Production Systems. . . n This rule consists of 2 propositions given on separate lines (2 condition elements) and a conclusion. The secondition element contains a variable. Condition elements in a rule can be connected by different connectives, the most used being AND, OR, NOT. N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Production Rules and Production Systems. . . n A production system consists of: • Working memory (facts memory) • Production rules memory • Inference engine, it cycles through three steps: – match facts against rules – select a rule – execute the rule N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Production Rules and Production Systems. . . n Figure 2. 25: A production system cycle N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

How to Program in Production Rules? n Figure 2. 27: A program written in a production language for the family relationship problem N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Advantages and Limitations of the Production Systems (PS): n n n PS are universal computational mechanism PS are universal function approximators readability explanation expressiveness modularity N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Expert Systems n n They are information systems for solving a specific problem which provides an expertise similar to those of experts in the problem area. An ES contains expert knowledge. N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Expert Systems. . . n A typical ES architecture consists of: • knowledge base module • working memory module (for the current data) • inference engine • forward chaining (inductive, data driven) • backward chaining (deductive, goal driven) • user interface (possibly a NLI, menu, windows, etc) • explanation module N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Expert Systems. . . n Figure 2. 29: An expert system architecture N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Expert Systems. . . n `How' and `Why' explanations in ES n Figure 2. 30 HOW and WHY explanation for The Car Monitoring Production System N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Data Analysis. . . n n Expert systems design • identification • conceptualization • formalization • realization • validation The knowledge acquisition problem: • interview experts • learning from data • literature • agents on the Web N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Sets n Figure 3. 1 Membership functions representing three fuzzy sets for the variable "height". N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Sets. . . n Figure 3. 2 Representin g crisp and fuzzy sets as subsets of a domain (universe) U N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Sets. . . n n Figure 3. 3 Support of a fuzzy set A see also fig 3. 21 for an example of fuzzy sets definitions for the The Bank Loan Decision problem. N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Rules and Fuzzy Inference n Rule 1: IF (CScore is high) and (CRatio is good) and (CCredit is good) then (Decision is approve) n Rule 2: IF (CScore is low) and (CRatio is bad) or (CCredit is bad) then (Decision is disapprove) N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Rules and Fuzzy Inference. . . n Inputs to a fuzzy system can be: • fuzzy, e. g. (Score = Moderate), defined by membership functions • exact, e. g. : (Score = 190); (Theta = 35), defined by crisp values. n Outputs from a fuzzy system can be: • fuzzy, i. e. a whole membership function, or • exact, i. e. a single value is produced on the output. N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Rules and Fuzzy Inference. . . n n n Fuzzy inference methods: `Fuzzification- rule evaluation- defuzzification' inference see Figure 3. 27 for an illustration of "crisp input data rules evaluation defuzzification" inference for a particular crisp input data for the Bank Loan Decision system. N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Rules and Fuzzy Inference. . . n n Methods for defuzzification: • center of gravity • mean of maxima see Figure 3. 26 Methods of defuzzification: the centre of gravity method (COG), and the mean of maxima method (MOM) applied over the same membership function for a fuzzy output variable y. They calculate different crisp output values. N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Information Retrieval and Fuzzy Databases n n n Fuzzy interfaces to standard databases (see fig 3. 32) Fuzzy databases (see fig. 3. 33) Fuzzy expert system shells (see fig. 3. 36, 3. 37) N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Information Retrieval and Fuzzy Databases. . . n Fuzzy expert systems n Figure 3. 35: A block diagram of a fuzzy expert system. N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996

Fuzzy Expert Systems n Fuzzy systems are: • easy to develop and debug • easy to understand • easy and cheap to maintain N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, MIT Press, 1996