Download DEVELOPMENT OF ADVANCED CONTROL METHODOLOGIES

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Attribute-Value Pair
Aim of developing this representation is to reduce the
development time especially the transition time between
the rule based to the representation employed by the
software package utilisation.
Attributes are the antecedents/conclusions of the rule
Values are the limits imposed on attributes
Example: If shahid is of age 5 Then he should go
to the school
If X < 10 and Y>2 Then Z=3
Lahore University of Management Sciences, Lahore,
Pakistan
2
Dr. M.M. Awais- Computer
Science Department
Attribute-Value Pair
Example:
Rule
R1
R3
R2
R3
R4
R4
R1
R3
R2
R4
R1
R2
R3
R4
if x=5 then z=2
if y=6 then s=1
if x=2 and y=2 then z=2
if s=1 and e=2 then d=7
Clause
1
1
1
2
1
2
3
3
3
3
Lahore University of Management Sciences, Lahore,
Pakistan
Attribute
x
x
y
y
s
e
z
z
s
d
3
Value
5
2
6
2
1
2
2
2
1
7
Dr. M.M. Awais- Computer
Science Department
Example 1: Waste Disposal
R1:
R2:
R3:
R4:
If waste origin = source A
and waste amount = large
Then waste destination = site X
If waste origin = source A
and waste amount = modest
Then waste destination = site Y
If waste origin = source B
and waste contents = toxic
Then waste destination = site Z
If waste origin = source B
and waste contents = nontoxic
Then waste destination = site Y
Lahore University of Management Sciences, Lahore,
Pakistan
4
Dr. M.M. Awais- Computer
Science Department
Example 1: Solution of Waste
Disposal
Lahore University of Management Sciences, Lahore,
Pakistan
5
Dr. M.M. Awais- Computer
Science Department
Example 2:
R1:
R2:
R3:
If A = a
and B = b
Then R = r
If A = NOT a
and C = c
Then S = s
If S = s
and B = b
Then X = x
Lahore University of Management Sciences, Lahore,
Pakistan
6
Dr. M.M. Awais- Computer
Science Department
Example 2: Solution
Lahore University of Management Sciences, Lahore,
Pakistan
7
Dr. M.M. Awais- Computer
Science Department
Inference Networks
Another Graphical Approach
Symbols are used to present rules
Box: represents assertion/antecedents
Circles: represents conclusions
Connectives are represented by special symbols
Assertion
conclusion
intermediate
conclusion
Lahore University of Management Sciences, Lahore,
Pakistan
8
and
or
Dr. M.M. Awais- Computer
Science Department
Example:
R1:
R2:
If assertion 1 is true
and assertion 2 is true
Then conclusion 1 is established
If assertion 3 is true
or assertion 4 is true
Then conclusion 2 is established
Lahore University of Management Sciences, Lahore,
Pakistan
9
Dr. M.M. Awais- Computer
Science Department
Example : inference Net
Lahore University of Management Sciences, Lahore,
Pakistan
10
Dr. M.M. Awais- Computer
Science Department
Example 1:
R1:
R2:
R3:
R4:
R5:
If B and C
Then G
If A and G
Then I
If D and G
Then J
If E or F
then H
If D and H
Then K
Lahore University of Management Sciences, Lahore,
Pakistan
11
Dr. M.M. Awais- Computer
Science Department
Example 1: inference Net
Lahore University of Management Sciences, Lahore,
Pakistan
12
Dr. M.M. Awais- Computer
Science Department
Neural Networks
Neural Networks can be used to represent Knowledge but has
a unique way of storing the knowledge
Works on the principles of biological neural system
Human nervous system contains more than 10 billion neurons
These neurons are massively interconnected
1. Dendrites are the parts which receive impulses
2. Axons transmit these impulses
3. Between two neurons is the synaptic junction
4. If the impulse energy is high enough it will jump the synaptic
junction and impulse is passed to the next neurons
5. Different Impulses activate the synaptic junction differently
6. Thus knowledge can be stored as a pool of activations that
responds to its particular input impulse.
Lahore University of Management Sciences, Lahore,
Pakistan
13
Dr. M.M. Awais- Computer
Science Department
Neural Networks
Artificial Neural networks have
Inputs
Outputs
Weights
that corresponds to dendrites
synaptic jump
store synaptic activity of different inputs
numeric form
Lahore University of Management Sciences, Lahore,
Pakistan
14
Dr. M.M. Awais- Computer
Science Department