Reasoning (Inference Engine)

Transcription

1 Reasoning (Inference Engine)

2 Materials Forward chaining Backward chaining Practice 2

3 Inference Engine In simple rule-based system there are two kinds of inference: 1. Forward Chaining (FC) 2. Backward Chaining (BC) 3

4 Forward Chaining A chain that is searched/traversed from a problem to its solution: Data-directed inference Antecedent determine search Reasoning from facts to the conclusion Bottom-up reasoning The process is broad and not deep 4

5 Forward Chaining Data-directed inference Data gets put into working memory This triggers rules whose conditions match the new data These rules then perform their actions The actions may add new data to memory, thus triggering more rules And so on. inference is triggered by the arrival of new data in working memory 5

6 Forward Chaining Antecedent determine search (1) Examples: Rules: 1. If X is a penguin, than X is a bird 2. If X is an elephant, then X is a mammal 3. If X is a mammal, then X is an animal Input: Clyde is an elephant Output: Clyde is an animal 6

7 Forward Chaining Antecedent determine search (2) Elephant(Clyde) Animal(Clyde) Elephant(X) Mammal(X) Mammal(X) Animal(X) Forward Chaining 7

8 Forward Chaining Reasoning from facts to the conclusion (1) Examples: Rules: R1, R2, R3, R4, R5, R6, R7, R8, R9 Input: C, D, E, F 8

9 Forward Chaining Reasoning from facts to the conclusion (2) R8 R9 Conclusion M N O P R5 R6 R7 J K L Inferred Facts R1 R2 R3 R4 A B C D E F G Facts 9

10 Forward Chaining Bottom-Up Reasoning R8 R9 Up M N O P R5 R6 R7 J K L R1 R2 R3 R4 A B C D E F G Bottom 10

11 Forward Chaining Broad and not deep A R1 L R7 P B R2 J R5 M C D R3 K R6 N E R4 O Facts Conclusion 11

12 Backward Chaining A chain that is searched/traversed from a hypothesis back to the facts that support the hypothesis : Goal-directed inference or hypothesis driven Consequent determine search Reasoning from a goal that can be accomplished by satisfying subgoals/facts Top-down reasoning The process is narrow and deep 12

13 Backward Chaining Goal-directed inference or hypothesis driven The system needs to know the value of a piece of data It searches for rules whose conclusions mention this data Before it can use the rules, it must test their conditions This may entail discovering the value of more pieces of data, and so on inference are not performed until the system is made to prove a particular goal 13

14 Backward Chaining Consequent determine search (1) Examples: Rules: 1. If X is a penguin, than X is a bird 2. If X is an elephant, then X is a mammal 3. If X is a mammal, then X is an animal Input: Clyde is an elephant Output: Clyde is an animal 14

15 Backward Chaining Consequent determine search (2) Elephant(Clyde) Animal(Clyde) Elephant(X) Mammal(X) Mammal(X) Animal(X) Backward Chaining 15

16 Backward Chaining Reasoning from goal to facts (1) Examples: Rules: R1, R2, R3, R4, R5, R6, R7, R8 Input: B, D, E 16

17 Backward Chaining Reasoning from goal to facts (2) G R1 R2 R3 Initial hypothesis (goal) N O P R4 R5 Q R S Intermediate hypothesis (subgoals) R6 R7 R8 A B C D E Facts 17

18 Backward Chaining Bottom-Up Reasoning G Top R1 R2 R3 N O P R4 R5 Q R S R6 R7 R8 A B C D E Down 18

19 Backward Chaining Narrow and deep A R1 J B R2 K R6 M R8 C R3 L R7 N R9 O R10 P R11 Q D R4 E R5 Evidence (facts) Hypothesis (conclusion) 19

20 Example Rules: 1. If rating is beginner and purpose is fun, then competition is local festival 2. If rating is beginner and purpose is serious, then competition is national festival 3. If rating is advance and purpose is fun, then competition is national annual festival 4. If rating is advance and purpose is serious, then competition is international festival 20

21 Rules (2): Example 5. If age < 2 years, then rating is beginner 6. If age >= 2 years and perfomance is poor, then rating is beginner 7. If age >= 2 years and perfomance is good, then rating is advance 8. If job < 15 times, then perfomance is poor 9. If job >= 15 times, then perfomance is good 21

22 Example Try to use: forward chaining and backward chaining! 22

23 Notes 1. BC can be written in a FC and vice versa But, consider the efficiency & conceptual : Change the BC to FC (or vice versa) need a lot of work 2. Strategy The type of inference engine used The order in which questions are asked The order of rules in knowledge base 23

24 24