Scribd
Upload
54 views2 pages

COMP3308/3608 Artificial Intelligence Tutorial Week 10: May 11th, 2015

1. The document summarizes the topics that will be covered in Tutorial Week 10 of the COMP3308/3608 Artificial Intelligence course, including: - Showing that a is conditionally independent of b and c given d. - Analyzing a DNA evidence case against Lucky Jim and computing probabilities. - Showing that random variables X and Z are conditionally independent given Y in a joint distribution. - Modeling a Bayesian network for an intruder alarm system and computing probabilities.

Uploaded by

m
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
54 views2 pages

COMP3308/3608 Artificial Intelligence Tutorial Week 10: May 11th, 2015

1. The document summarizes the topics that will be covered in Tutorial Week 10 of the COMP3308/3608 Artificial Intelligence course, including: - Showing that a is conditionally independent of b and c given d. - Analyzing a DNA evidence case against Lucky Jim and computing probabilities. - Showing that random variables X and Z are conditionally independent given Y in a joint distribution. - Modeling a Bayesian network for an intruder alarm system and computing probabilities.

Uploaded by

m
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

COMP3308/3608 Artificial Intelligence

Tutorial Week 10

May 11th, 2015

1. Homework: Show that a independent of b and c given d (denoted as a b, c|d) implies

|=
a independent of b given d, i.e. a b|d.
|=
2. Lucky Jim is in court accused of murder. The prosecutor in his attempt to convince
the jury brings up the incredible accuracy of the DNA lab test that has been carried
out. He says: “The probability of a DNA match for an innocent person is 1 in 100 000
Even more, the probability of a match for a guilty person is 1”. Assume that Lucky Jim
lives in a town of 105 people who could have committed the crime, and that there is a
DNA match to him. The prosecutor goes on and states: “given the DNA evidence, it is
very improbable that the defendant is innocent and therefore he should be found guilty”.
Should Lucky Jim really be declared guilty? Compute the probabilities of interest and
show how these change as a function of the town population (use 103 , 104 , . . . , 108 ).
3. Consider the random variables X, Y, Z which have the following joint distribution:
p(X, Y, Z) = p(X)p(Y |X)p(Z|Y )
(a) Show that X and Z are conditionally independent given Y .
(b) If X, Y and Z are binary variables, how many parameters are needed to specify a
distribution of this form?
4. Consider the following scenario. You have an intruder alarm at your house. This alarm
may go off because of an intruder at your house or because of an earthquake near your
place. Your neighbour Mr Wilson, who is overly paranoid, may hear your alarm. Similarly,
your other neighbour Mrs Jones, who lives further from your house, may hear your alarm.
Let the variables alarm (A), earthquake (E), intruder (I), Mr Wilson (W ) and Mrs Jones
(J) denote whether the alarm went off; an earthquake happened; there is an intruder at
your house; Mr Wilson heard the alarm; and Mrs Jones heard the alarm, respectively,
with A, E, I, W, J ∈ {0, 1}. It is sensible to model the joint distribution of these variables
as follows:
P (E, I, A, W, J) = P (E)P (I)P (A|E, I)P (W |A)P (J|A).
The conditional probability tables (CPTs) that fully determine the above scenario are:
P (E = 1) = 10−6 P (I = 1) = 0.01
P (A = 1|I = 1, E = 0) = 0.99 P (A = 1|I = 0, E = 1) = 0.99
P (A = 1|I = 1, E = 1) = 0.9999 P (A = 1|I = 0, E = 0) = 10−4
P (W = 1|A = 1) = 0.95 P (W = 1|A = 0) = 0.2
P (J = 1|A = 1) = 0.8 P (J = 1|A = 0) = 0.05

1
(a) Draw the Bayesian network corresponding to the above joint distribution.
(b) Show that E and I are statistically independent.
(c) Compute P (I, W ), P (I, J), P (I|W ) and P (I|J).

You might also like