Inference Techniques

1
Chapter 13

• Inference Techniques

2
Reasoning in Artificial Intelligence

• Knowledge must be processed (reasoned with)
• Computer program accesses knowledge for
  inferencing
• Inference engine or control program
• Rule interpreter (in rule-based systems)
• The inference engine directs search through the
  knowledge base
• The application of inference rules is called
  pattern matching

3
Reasoning Methods
Deductive Reasoning

• Move from a general principle to a specific
  inference. A general principle is composed of two
  or more premises
• Deduce new information from logically related
  known information.
• Uses problem facts or axioms and related general
  knowledge in the form of rules or implications

4
Reasoning Methods
Deductive Reasoning

• The process begins with comparing the axioms with
  a set of implications to conclude new axioms.
• Ex.,
• Implication I will get wet if I am standing
• in the rain
• Axiom I am standing in the rain
• Conclusion I will get wet

5
Reasoning Methods
Inductive Reasoning

• Move from some established facts to draw general
  conclusions
• Arrive at a general conclusion from a limited set
  of facts by the process of generalization.
• We form a generalization which we believe applies
  to all cases of a certain type, on the basis of a
  limited number of cases.

6
Reasoning Methods
Inductive Reasoning

• Ex.,
• Premise Monkeys in the Pittsburgh zoo
• eat bananas
• Premise Monkeys in the Cleveland zoo
• eat bananas
• Conclusion In general, all monkeys eat
• bananas

7
Reasoning Methods
Analogical Reasoning

• Derive an answer to a question by known analogy.
  It is a verbalization of internalized learning
  process. Use of similar experiences
• They draw analogies between the two, looking for
  similarities and differences to guide their
  reasoning.

8
Reasoning Methods
Analogical Reasoning

• Tiger Frame
• Specialization-of ANIMALS
• Number-of-legs 4
• Eats meat
• Lives India and Southeast Asia
• Color tawny with stripe

9
Reasoning with Logic
Modus Ponens

• If A, then B and the fact that A is true, then B
  is also true
• A AND (A ? B) ? B
• A and (A ? B) are propositions in a knowledge
  base
• A It is sunny.
• B We will go to the beach
• A-gtB If it is sunny, then we will go to the
  beach
• Conclusion We will go to the beach

10
Reasoning with Logic

• Modus Tollens
• If (A-gtB) is known to be true, and B is false,
  then A is flase
• Resolution combines substitution, modus ponens,
  and other logical syllogisms
• If (AVB) is true and (BVC) is true
• THEN (AVC) is true

11
Inferencing with Rules
Forward and Backward Chaining

• Apply modus ponens
• Firing a rule When all of the rule's hypotheses
  (the if parts) are satisfied
• Can check every rule in the knowledge base in a
  forward or backward direction
• Continues until no more rules can fire, or until
  a goal is achieved

12
Forward and Backward Chaining

• Chaining Linking a set of pertinent rules
• Search process directed by a rule interpreter
  approach
• Forward chaining if the premise clauses match
  the situation, then the process attempts to
  assert the conclusion
• Backward chaining if the current goal is to
  determine the correct conclusion, then the
  process attempts to determine whether the premise
  clauses (facts) match the situation

13
Backward Chaining

• Goal-driven - Start from a potential conclusion
  (hypothesis), then seek evidence that supports
  (or contradicts) it
• Often involves formulating and testing
  intermediate hypotheses (or subhypotheses)

14
Backward Chaining
Procedure (goal Pay the bill)

• The inference engine starts by assuming the
  conclusion Action needed pay the bill as the
  possible solution to the goal, rule 5. It will
  try to answer the question, Under what
  conditions should the bill be paid?
• The next step is to prove the premise of rule 5.
  It looks at each rule in the KB to find the
  action part of a rule that matches the premise of
  rule 5. It finds it in rule 3.

15
Backward Chaining
Procedure (goal Pay the bill)

• The inference engine goes to rule 3 and looks for
  the action part og a rule that matches the
  premise of rule 3. It find it in rule 1.
• The systems tries to prove the premise of rule 1.
  Since no other rules action matches the premise
  of rule 1, in this case, the inference engine
  asks the user, What is todays date? Whatever
  it is, the default message would be Todays date
  is due date.

16
Backward Chaining

• Step Rule Rule
• 4 R1 IF due date on or before to day
• THEN payment is due
• 3 R2 IF due date is after today
• THEN payment is NOT due
• R3 IF payment is due
• THEN paying is recommended
• 2
• R5 IF paying is recommended
• 1 THEN action needed. Pay the bill

GOAL
17
Forward Chaining

• Data-driven - Start from available information as
  it becomes available, then try to draw
  conclusions
• What to Use?
• If all facts available up front (as in auditing)
  - forward chaining
• Diagnostic problems - backward chaining

18
Forward Chaining
Procedure

• The system obtains problem information from the
  user and places it in the working memory.
• The inference engine scans the rules in some
  predefined sequence looking for one whose premise
  match the contents in the working memory.
• If it finds a rule, it adds the rules conclusion
  to the working memory (called firing the rule).
• Then cycles and checks the rules again looking
  for new matches.
• On the new cycle, rules that previously fired are
  ignored.
• This process continues until no matches are found

19
Forward Chaining
Step Rule Rule R1 IF due date
on or before to day THEN payment is due 1
R2 IF due date is after today THEN
payment is NOT due R3 IF payment is
due THEN paying is recommended 2
R5 IF paying is recommended THEN action
needed. Pay the bill
20
The Functions of The Inference Engine

• Fire the rules
• Present the user with questions
• Add the answer to the ES blackboard (assertion
  base)
• Infer a new fact from a rule
• Add the inference fact to the blackboard
• Match the blackboard to the rules
• It there are any matches, fire the rules
• If there are two further matches, check to see
  whether a goal is reached
• Fire the lowest-numbered unfired rule

21
The Inference Tree

• (Goal Tree or Logical Tree)
• Schematic view of the inference process
• Similar to a decision tree (Figure 13.3)
• Inferencing tree traversal
• Advantage Guide for the Why and How Explanations

22
The Inference Tree
G
Root
R5
F
OR
R2
R3
BE
CD
Branches
AND
AND
E
R4
C
R1
AND
Fact
Fact
Fact
R4
Dead end
Fact
D
B
B
A
C
B
23
Explanation
Explanation

• Human experts justify and explain their actions
• ES should also do so
• Explanation attempt by an ES to clarify
  reasoning, recommendations, other actions (asking
  a question)
• Explanation facility (justifier)

24
Explanation
Explanation Purposes

• Make the system more intelligible
• Uncover shortcomings of the rules and knowledge
  base (debugging)
• Explain unanticipated situations
• Satisfy users psychological and/or social needs
• Clarify the assumptions underlying the system's
  operations
• Conduct sensitivity analyses

25
Rule Tracing Technique

• Why Provides a Chain of Reasoning
• Good Explanation Facility is critical in large ES
• Understanding depends on explanation
• Explanation is essential in ES
• Used for training

26
Two Basic Explanations
Why Explanations - Why is a fact requested?

• Computer What is your annual income?
• Client Why? (Why do you need to know?)
• Computer In checking R2, I need to know whether
  your income is above 40,000. If this is true, I
  will conclude that b/c you have a college degree,
  you should invest in growth stocks

27
Two Basic Explanations
How Explanations - To determine how a certain
conclusion or recommendation was reached.

• Computer Invest in IBM stocks
• Client How? (How was the conclusion reached?)
• Computer (Displays the rules R5-R3-R7-R4 in the
  sequence that they were fired)

28
Generating Explanations

• Static Explanation preinsert pieces of English
  text (scripts) in the system
• Dynamic Explanation reconstruct explanation
  according to the execution pattern of the rules