Lecture 3 Cheat Sheet – Fuzzy Logic & Mamdani
Inference
Extension Principle
• Definition: Extends mathematical functions from crisp to fuzzy domains.
• Analogy: Like upgrading recipes from exact (1 cup sugar) to approximate (‘about 1 cup’).
Linguistic Variables
• Definition: Variables described in words (e.g., small, medium, large) instead of numbers.
• Analogy: Saying ‘tall’ instead of ‘180 cm’.
Membership Function (MF)
• Definition: Maps crisp inputs to fuzzy values between 0 and 1.
• Analogy: At 45 km/h, a car may be 0.6 ‘medium speed’ and 0.4 ‘slow’.
Combining Logic and Sets
• Definition: Logical tautologies (like p ⇒ q ≡ ¬p ∨ q) link logic with fuzzy set operations.
• Analogy: Rule ‘If it rains, then take umbrella’ ≡ ‘Either it doesn’t rain OR you take an umbrella’.
Inference Rule – Modus Ponens
• Definition: If p ⇒ q and p is true, then q is true.
• Analogy: If ‘If it rains, ground is wet’ and it rains, then the ground is wet.
Generalised Modus Ponens
• Definition: Extends Modus Ponens to fuzzy rules where antecedents and consequents are
fuzzy sets.
• Analogy: ‘If it is somewhat raining, then the ground is somewhat wet.’
Fuzzy If–Then Rules
• Definition: Rules connecting fuzzy conditions to fuzzy outputs.
• Analogy: Car heater: IF cold THEN heating high.
Fuzzification
• Definition: Converting crisp inputs into fuzzy values.
• Analogy: Room temp 15°C → 0.7 ‘cold,’ 0.3 ‘cool’.
Firing Strength of a Rule
• Definition: The degree of truth of the antecedent (calculated via AND/OR).
• Analogy: If Tall = 0.5 and Young = 0.5, rule fires at 0.25 (weakest link effect).
Consequent Evaluation
• Definition: Firing strength scales the consequent fuzzy set.
• Analogy: Rule says ‘Good potential,’ firing strength = 0.25 → ‘25% good’.
Multiple Rules
• Definition: Outputs from different rules are combined (union/maximum).
• Analogy: Two doctors give diagnoses (OK, Good) → final opinion merges both.
Rule Aggregation
• Definition: Combining all rule outputs into a single fuzzy set.
• Analogy: Merging multiple weather forecasts into one final prediction.
Defuzzification
• Definition: Convert aggregated fuzzy set back into a single crisp value.
• Analogy: After fuzzy advice (‘0.6 OK, 0.25 Good’), final decision = score 6.5/10.
Applications
• Definition: Examples: 1975 Cement Kiln (Denmark), 1985 Japanese Railway braking.
• Analogy: Like having a human operator’s judgment embedded in a machine.
