|
Theory:
45 min:
Bayesian reasoning, Fuzzy expert systems, Frame based systems
A summary of
the lecture together with the power point presentation and associated
material can be downloaded via the link to
lecture 3.
Pause :
Following
the pause, students form 3 groups for a workshop discussion.
Workshop:
30 min. Each
group is given one of the questions from the homework assignment tied to
this lecture. The
purpose of the discussion is to lay down the guide lines and to prepare
answers to the questions which students will have to submit in their
individual homework assignment.
Homework assignment
For the homework assignment students will be asked to
answer three questions on the subject of intelligent systems and to
upload their answers in a word document to our E-learning forum,
iknow.iie.nl. The questions are:
Question 1
Bayesian probability
The
probability of A, GIVEN B, is the joint probability of A and B, divided
by the probability of A:
Casus
1000
students submit an entry in a competition to find the best computer
solution for solving a chess problem in the least possible moves.
We are
told that 20 % of the entries actually succeed in solving the problem.
80% of
the successful solutions contain less than 4 moves.
Not all
the entries were successful. 10% of the entries which did not actually
succeed in solving the problem also contained less than 4 moves.
If an
entry contains less than 4 moves, what is the probability that the entry
will also turn out to be a successful solution to the chess problem.?
Question 2
Fuzzified expert systems
In the
assignment related to lecture 2 students were required to describe a
rule-based expert system.
Most of
the students should now be aware that, amongst other things, this
description requires a set of IF THEN rules in the knowledge base.
Combining the IF THEN rules with the facts entered into the database via
the user interface is performed by the inference engine. When individual
rules ‘fire’ due to the antecedent value being set to TRUE by the
inference engine, then the consequent will be set. If the
consequent of a rule matches the antecedent of another rule, then that
rule will also fire.
This
sort of simplified expert system is a system based on crisp logic – that
is, a system in which the antecedents and consequents of the rules can
only contain absolute values or logical values TRUE and FALSE.
In
this assignment you are asked
a)
to describe three ‘crisp’ rules similar to rules which you
described in the previous assignment
b)
to transform these rules to ‘fuzzy’ rules and
c)
to demonstrate how these fuzzy rules can be used in a fuzzy
inference engine.
The
explanation of the concepts involved in fuzzy expert systems is
contained in chapter 4 of the text book. More precisely, the process of
‘fuzzy inference’ is explained in paragraph 4.6. where Mamdani
inferencing and Sugeno inferencing methods are demonstrated. Both
methods involve the steps:
1)
fuzzification
2)
rule evaluation
3)
aggregation
4)
defuzzification
Specifically, in this exercise you are asked to apply the Mamdani
inferencing method to the fuzzified rules which you obtained by
transforming your choice of ‘crisp’ rules.
Question 3
Frame based systems
• What
is the equivalence of linguistic variables in frame theory? (frame,
slot, facet). Why?
• What
is the equivalence of a fuzzy set in frame theory? (frame, slot, facet).
Why?
• Why
is the “frame” from frame theory missing in the fuzzy theory?
Think to different programming techniques.
•Additional information
–
The assignment is individual
–
The assignment tends to bring together knowledge from different
areas:
o
information management
o
object oriented paradigm
o
artificial intelligence
o
programming techniques and algorithms
|
