ai sem1 2012-13-w2-representation
TRANSCRIPT
CHAPTER 2
KNOWLEDGE REPRESENTATION
What is AI
AI Techniques
AI Applications
- Raw facts
- Static
- E.g. Liza, S’pore
- Processed data
- Has some meaning and purposes
- E.g. Liza was born in S’pore
- Derived from information
- Stored in human brain
- What we know
1. Introduction
2. Semantic Networks
3. Decision Tables
4. Decision Trees
5. Frames
6. Production Rules
7. Logic
a. Propositional Logic
b. Predicate Calculus
Learning Outcome
Analysis Representation
Coding Representation
“The know-how to program a computer to mimic
the thought processes of an expert through an
appropriate representation scheme” is called
knowledge representation (KR)
It involves knowledge of a shell or a programming
language that will represent the expert
knowledge
A number of KR schemes shares 2 common
characteristics:
a. They contain facts that can be used in
reasoning
b. They can be programmed with existing
computer languages
There are generally 2 types of KR schemes:
a. Analysis representation
Support knowledge acquisition during scope
establishment and initial knowledge gathering
Most techniques are pictorial such as
-semantic networks - decision trees -tables
b. Coding representation
The working code of the ES either in the form of
-frames or - production rules
Knowledge
Representation
Analysis
Representation
Coding
Representation
Inference
Frames
Production rules
Semantic networks
Decision tables
Decision trees
Selected KR Schemes
Key
Idea
Analysis Coding
Semantic Networks
Decision Tables
Decision Trees
Semantic networks are the most general representation scheme.
Represent a graphical representation of knowledge that show hierarchical relationships between objects.
Made up of a network of nodes and arc.
The nodes represent objects and the arc the relationships between objects.
KR Scheme 1
Node Arc
Example:
KR Scheme 1
Example:
License
Seal
Examination
Air
rescue
Emergency
landing
procedures
Olesek
Insignia
Shirt
sleeve
Two
Male
Harding
Person
Apparel
Uniform
Black
Cap
has-a
certifies
has-a
in-a in-an
has-
an
is-on
is-a
number-
of has-a
is-a
race-is
is-a
wears
is-a
is-an
KR Scheme 1
Nodes represent the objects, concepts, or events in
the world.
Names of the arcs
correspond to names of relations
indicate which concepts or objects are linked by
the relations.
KR Scheme 1
The 2 common arcs used are:
IS-A is used to show class relationship.
HAS-A is used to identify characteristics or
attributes of the object nodes
other arcs are used for definitional purpose only
KR Scheme 1
Organized in a spreadsheet format, using columns
and rows
The table divided into two parts:
A list of attributes is developed and for each attribute,
all possible values are listed
A list of possible conclusions. The different
configurations of attributes are match against the
conclusion.
Attributes Conclusions
KR Scheme 2
Decision Table for Gift Problem
Decision Factors Result
Money Age Gift
Much Adult Car
Much Child Computer
Little Adult Toaster
Little Childs Calculator
KR Scheme 2
KR Scheme 2
A hierarchical arranged semantic network and is
closely related to a decision table
It is composed of nodes representing goals and
links representing decisions
Rules can be extracted from the decision tree,
that can be executed by computer program
A major advantage can simplify knowledge
acquisition process
KR Scheme 3
KR Scheme 3
A decision tree for an electrical system diagnosis
Terminals
Battery Voltage
Distributor Charger
Not
Loose
Loose
Tighten Terminals
<12 >12
OK OK Bad Bad
Check
Starter
Replace
Distributor
Replace
Charger
Replace
Battery
KR Scheme 3
Analysis Coding
Frames
Production Rules
Logic
A frame is a data structure for representing
common concepts and situations (stereotype
knowledge).
Like semantic nets, frames can be organized in a
hierarchy with general concepts near the top and
specific concepts placed at the lower levels.
General -top
Specific -lower
KR Scheme 4
Unlike semantic nets, each frame or node in this
hierarchy can be very rich in supplementary
information.
KR Scheme 4
Values that describe one object are grouped
together into single unit.
Knowledge partition into slots.
Each slot describes:
• declarative knowledge (colour of a car)
• procedural knowledge (activate a certain rule if
the value exceeds a certain value)
KR Scheme 4
Frames describe an object in great detail. The
detail in form of slots that describe the various and
characteristic of the object or situation.
KR Scheme 4
Basic frame design
Frame
Name:
Class:
Properties:
Frame1
**Eg: Bird
***2 ***Eg: No. of wings
***Worms ***Eg: Eat
Value2 Property2
Value1 Property1
KR Scheme 4
Class frame
Represents general characteristics of some set of
common objects. For example Cars, Boats, and Birds.
Defines those properties that are common to all the
objects within the class, and possibly default property
values.
static: describe an object
feature whose value doesn’t
change
dynamic: is a feature whose value
is likely to change during
operation of the system
KR Scheme 4
Example of Class Frame
Class Name:
Properties:
Bird
Try Flies
2 No Wings
Worms Eats
Unknown Color
Unknown Hungry
Unknown Activity
KR Scheme 4
Class Name:
Properties:
KR Scheme 4
An Instance of Frame
Describes a specific instance (sub-class or examples)
of a class frame.
The instance inherits both properties and property values
from its frame class.
The property values can be changed (recall: static/dynamic) to
tailor the object represented in the instance.
Many instances of the frame class can be created.
The instances immediately inherit the frame’s properties.
Can speed up system coding.
KR Scheme 4
Instance frame
Frame Name:
Class Name:
Properties:
Tweety
Bird
False Flies
1 No Wings
Eats
Yellow Color
Hungry
Activity
Lives Cage
KR Scheme 4
Frame Inheritance
From example, “Tweety” is an instance of Bird class.
Can allow an instance to accept the class default
values or provide values unique to the instance.
Like most bird Tweety eats worms, but has only one
wing and cannot fly.
Can also provide unique properties. e.g. if Tweety lives
in a cage.
KR Scheme 4
Frame Inheritance
Inheriting behaviour
Beside inheriting descriptive information from its
class, an instance also inherits its behaviour.
Need to include a procedure (method) within class
frame that define some actions that the frame
performs.
KR Scheme 4
A form of procedural knowledge that describe how to solve a problem.
The procedural and/or factual knowledge is represented as rules, called productions, in the form of condition-action pairs.
Are stated as follows: "IF this condition occurs, THEN do this action; or this
result (or conclusion or consequence) will occur.
KR Scheme 5
Examples
IF flammable liquid was spilled,
THEN call the fire department.
IF the pH of the spill is less than 6,
THEN the spill material is an acid.
IF the spill material is an acid,
and the spill smells like vinegar,
THEN the spill material is acetic acid.
KR Scheme 5
When the IF portion of a rule is satisfied by the
facts, the action specified by the THEN is
performed.
When this happens, the rule is said to "fire" or
"execute".
KR Scheme 5
Relationship
IF The battery is dead
THEN The car will not start
Recommendation
IF The car will not start
THEN take a cab
Directive
IF The car will not start
AND the fuel is okay
THEN check out the electrical system
KR Scheme 5
Strategy
IF The car will not start
THEN first check out the fuel system then check
out electrical system
Heuristic
IF The car will not start
AND The car is a 1957 Ford
THEN check the float
KR Scheme 5
Uncertain Rules
IF inflation is high
THEN Almost certainly interest rates are high
Can assign Certainty Factors:
IF inflation is high
THEN interest are high CF=0.8
KR Scheme 5
Meta-Rules
A rule that describe how other rules should be used.
IF the car will not start
AND the electrical system is operating normally
THEN use rules concerning the fuel system
KR Scheme 5
Rules are easy to understand
Inference and explanation are easy to derive
Modifications and maintenance are relatively easy
Uncertainty is easily combined with rules
Each rule is usually independent of all others
KR Scheme 5
The oldest form of knowledge representation in a computer is logic
Logic is concerned with the truthfulness of a chain of statements.
An argument is true if and only if, when all assumptions are true, then all conclusions are also true.
2 kinds of logic:
Propositional Logic
Predicate Calculus
Both use symbols to represent knowledge and operators applied to the symbols to produce logical reasoning
KR Scheme 6
Propositional logic represents and reasons with propositions.
PL assigns symbolic variable to a proposition such as A = The car will start
In PL, if we are concern with the truth of the statement, we will check the truth of A.
KR Scheme 6a
KR Scheme 6a
Propositions that are linked together with connectives, such as AND, OR, NOT, IMPLIES, and EQUIVALENT, are called compound statements.
Example:
IF The students work hard
AND Always come to lectures
AND Do all their homework
THEN They will get a good grade
Using logic symbols: A B C -> D
Propositional logic is concerned with the truthfulness of compound statements, depending on the connectives.
KR Scheme 6a
F F F F T T
F T F T T F
T F F T F F
T T T T F T
Truth Table
KR Scheme 6a
Implies Operator: C = A B (C is A implies B)
For an implication of C, if A is true, then B is implied to be true
(A B) ( A B)
The implies return an F when A is TRUE and B is FALSE
Otherwise it returns TRUE.
A B A B
F F T
F T T
T F F
T T T
KR Scheme 6a
Example to illustrate Implies
IF The battery is dead (A)
THEN The car won’t start (B)
A B A B
F F T
F T T
T F F
T T T
KR Scheme 6a
PL offers techniques for capturing facts or rules in
a symbolic form and then operates on them
through use of logical operators.
PL provides methods for managing statements
that are either TRUE or FALSE.
KR Scheme 6a
Some PL weakness:
1. Limited ability to express knowledge and lose much
of their meanings.
The Pacific Ocean contains water.
Florida is a state within the USA.
Only assigning a true value without making any
statement about ‘oceanhood’ or ‘statehood’.
KR Scheme 6a
Some PL weakness:
2. Not all statements can be represented.
All men are mortals.
Some dogs like cats.
Thus, need a more general form of logic that
capable of representing the details.
Therefore Predicate Calculus is introduced.
KR Scheme 6a
Enhances processing by allowing the use of
variables and functions.
Use symbols that represent
• constants
• predicates
• variables
• functions
Operate on these symbols using PL operators (
.
KR Scheme 6b
Specific objects or properties about a problem.
Begin with lower case.
Example: ahmad, elephant and temperature
KR Scheme 6b
Divide a proposition into 2 parts:
predicate: assertion about object
argument: represents the object
To represent a statement “John likes Mary.” in a
predicate calculus.
likes(john, mary)
A predicate Arguments
KR Scheme 6b
Represents general classes of objects or properties.
Written as symbols beginning with upper case.
likes(john, X)
KR Scheme 6b
Permits symbols to be used to represent functions.
A function denotes a mapping from entities of a
set to a unique element of another set.
father_of(john) mother_of(john)
Can be also used within predicates. For
example: married(father_of(john), mother_of(john))
KR Scheme 6b
PC uses the same operators as in PL.
Proposition:
David is John’s father. father(david, john)
Jane is John’s mother. mother(jane, john)
If X is John’s father and Y is John’s mother then X is Y’s
husband.
father(X, john) mother(Y, john) husband(X, Y)
KR Scheme 6b
PC introduces 2 symbols called variable
quantifiers.
1. universal quantifier: for all
2. existential quantifier: there exist
KR Scheme 6b
indicates an expression is TRUE for all values of
designated variable.
Example:
X likes (X, mary)
means for all values of X, X likes Mary.
“Everyone likes Mary.”
KR Scheme 6b
indicates an expression is TRUE for some values of
the variable; at least one value existed that makes
the statement is true:
Example:
X likes (X,mary)
means there exist X where X likes Mary.
“Someone likes Mary.”
KR Scheme 6b
Parentheses are used to indicate the scope of
quantification
X (likes(X,mary) nice(mary) nice (X))
determines all instances of X who like Mary and if
Mary is nice, then it is implied that those who like
Mary are nice too.
X (man(X) mortal(X))
All men are mortal. Man(X)
KR Scheme 6b
PC can provide reasoning capability to intelligent
systems
Reasoning requires the ability to infer conclusions
from available facts.
One simple form of inference is modus ponens:
IF A is true
AND A B is true
THEN B is true
Based on the available facts below:
X (man(X) mortal(X))
man(socrates)
We can infer a conclusion of mortal(socrates)
KR Scheme 6b
father(david, john)
mother(jane, john)
father(X, john) mother(Y, john) husband(X, Y)
Who is X?
Who is Y?
Who is Y’s husband?
From the above facts and rule, we can infer some more conclusions …..
KR Scheme 6b
The semantic network and its equivalent predicate calculus.
is_a(E1,elephant)
name(E1, clyde)
num_trunk(elephant, 1)
tail(elephant, 1)
num_legs(elephant, 4)
skin_colour(elephant, grey)
Convert the predicate calculus below into its equivalent semantic networks.
has_size(bluebird, small)
has_covering(bird, feathers)
has_colour(bluebird, blue)
has_property(bird, flies)
is_a(bluebird, bird)
is_a(bird, vetebrate)
Represent the following English sentences in predicate calculus:
1. Monkeys like bananas.
2. Dogs chase cats.
3. John doesn’t like ice-creams.
4. If weather is good, I go jogging.
Prepare a frame of an automobile that you
know, show 2 levels of hierarchy. Fill some
property and property values. (static and
dynamic)
Try to crank the starter. If it is dead or cranks slowly, turn on the headlights. If the headlights are bright (or dim only slightly), the trouble is either in the starter itself, the solenoid, or in the wiring. To find the trouble, short the two large solenoid terminals together (not to ground). If the starter cranks normally, the problem is in the wiring or in the solenoid; check them up to the ignition switch. If the starter does not work normally, check the bushings (see section 7-3 of the manual for instructions). If the bushings are good send the starter to the test station or replace it. If the headlights are out or very dim, check the battery (see section 7-4 for instructions). If the battery and connecting wires are not at fault, turn the headlights on and try to crank the starter. If the lights dim drastically, it is probably because the starter is shorted to the ground. Have the starter tested or replace it. (Based on Carrice et al. [5]).