Anda di halaman 1dari 6

CSCI 1410: Artificial Intelligence

Spring 2017 Practice Final

1 Probabilistic Reasoning and Bayes Nets (40 points)

a) Write down an expression for the joint probability distribution represented by the following Bayes Net:

A
C D E

How many entries are there in all of the conditional probability tables (CPTs) combined, if the random
variables in the net are all binary?

(10 points)

1
b) Write down Bayes Rule, and explain how it can be used to update the probability of an event A given
some evidence B. Explain what each term means.

(10 points)
c) Imagine that you have a data set with two random boolean variables, A and B. You are trying to decide
whether the relevant Bayes Net (which has only two nodes) should contain edge A B, or no edges at
all. (Your prior knowledge tells you that B A is out.) Explain how you could use the data youve got to
determine whether that edge should be in the graph or not.

(10 points)

2
d) Explain how the Markov assumption applies to the Hidden Markov Model, and how it is reflected in the
relevant Bayes Net.

(10 points)

3
2 Planning (15 points)

a) A PDDL operator description consists of an action name plus three components. List the components
and explain what each one does.

(10 points)
b) Planning can be viewed as a search problem. If you are using PDDL, what are the nodes in this search
problem, and what are the actions? What is the goal test?

(10 points)

4
3 Machine Learning (35 points)

a) Given the following decision tree:

A?
true false

B? B?
true false true false

false C? true false


true false

true false

... write down a logical expression (using (or), (and), and (not), and the three binary variables A, B,
and C), that is true exactly when the tree returns true. What does the answer tell you about the hypothesis
class of decision trees with binary variables?

(15 points)
b) Write down an expression for the decision boundary of a perceptron that is linear in inputs x and y
(include a bias (or threshold) term). Give an example data set (a picture with just a few input points is fine)
that cannot be separated using this decision boundary.

(10 points)

5
c) Explain the difference between reinforcement learning and supervised learning. Be precise about the
inputs and objective functions of each.

(10 points)

4 Philosophy of AI (10 points)

a) Is the Chinese Room an argument against the Weak AI Hypothesis, or the Strong AI Hypothesis? Explain
why.

(10 points)

Anda mungkin juga menyukai