Name:

CSCI 561 Final, Spring 2011

This test is open book and notes, but you must work alone with no help from
any other people (any than questions asked of the instructor or TA) or
resources. Use of all electronic equipment is forbidden.

1. [5] Yes or No: Euclidean distance dominates Manhattan distance as a

heuristic for the Eight Puzzle (where both are computed as sums of
distances for pieces out of place). Circle the answer that applies.

2. [10] To exhibit via resolution behavior that is similar to forward

m
chaining in a rule system, repetitions of resolving which kind of choice

er as
from List A with which kind of choice from List B comes closest? Circle

co
one choice in each list:

eH w
o.
List A List B
rs e
ou urc
horn clause unit clause
definite clause positive unit clause
o

negative unit clause

aC s
vi y re

3. [10] Compared with minimax search, iterative deepening alpha-beta

ed d

yields what advantages? Circle all that apply:

ar stu

reduced effective branching factor reduced search depth

sh is

improved node ordering improved coping with deadlines

Th

4. [10] Which of these relationships can individually (i.e., all by itself)

guarantee that P(a|b)=P(b|a) ? Circle all that apply:

P(a)=P(b) a and b are independent

a and b are mutually exclusive a and b constitute an exhaustive
decomposition

https://www.coursehero.com/file/8054322/CSCI-561-2011-Final-Answers/
https://www.coursehero.com/file/17380923/CSCI-561-2011-Final-Answers/
 Name:

5. [25] You have seen that single-layer Perceptrons can’t encode an XOR
function, A⊕B, while such a support vector machine can be built. For
each of the following formalisms, provide an instance that encodes XOR.

a. [5] Decision Trees

m
er as
co
eH w
o.
b. [10] Two-layer Perceptrons
rs e
ou urc
o
aC s
vi y re
ed d
ar stu

c. [10] Bayesian networks

sh is
Th

Page 2
https://www.coursehero.com/file/8054322/CSCI-561-2011-Final-Answers/
https://www.coursehero.com/file/17380923/CSCI-561-2011-Final-Answers/
 Name:

6. [30] Consider the following two Bayesian networks, with CPTs omitted.

m
er as
NBN: Naive Bayes Network DN: Diamond Network

co
a. [10] Write the factored expression each network defines for the

eH w
joint probability distribution P(A,B,C,D).

o.
rs e
NBN: P(A)P(B | A)P(C | A)P(D | A)
ou urc
o

DN: P(A)P(B | A)P(C | A)P(D | B, C)

aC s
vi y re

b. [10] What is the Markov blanket for node C in each network?

ed d

NBN: A
ar stu

DN: A, B, D
sh is

c. [5] Which of these networks are polytrees?

Th

NBN

d. [5] If training examples include values for all nodes that are either
parentless or childless, which of these networks require
unsupervised learning in support of Bayesian learning?

DN

Page 3
https://www.coursehero.com/file/8054322/CSCI-561-2011-Final-Answers/
https://www.coursehero.com/file/17380923/CSCI-561-2011-Final-Answers/
 Name:

7. [10] Consider the problem of preparing a birthday dinner for a friend who
is currently sleeping in the apartment. The goal is to have the meal ready
(dinner), the gift wrapped (present), and the trash gone (¬garb).
In the initial state, the meal isn’t ready, the gift isn’t wrapped, your hands
are clean (cleanH), there is trash, and the apartment is quiet (quiet).
There are four actions that can be performed:

Action Preconditions Effects

cook (cook dinner) cleanH dinner
wrap (wrap present) quiet present
carry (carry out garbage) ¬garb, ¬cleanH
dolly (roll out garbage) ¬garb, ¬quiet

m
er as
Consider an Extract-Solution algorithm (within Graphplan)

co
eH w
that is based on backwards search. Assume that as the search proceeds

o.
backwards, a maximal set of compatible actions is included at each level.
rs e
In the following planning graph, circle the actions (including persistence
ou urc
actions) for the plan (to achieve the goal mentioned above) that would be
derived by such an algorithm.
o
aC s
vi y re
ed d
ar stu
sh is
Th

Page 4
https://www.coursehero.com/file/8054322/CSCI-561-2011-Final-Answers/
https://www.coursehero.com/file/17380923/CSCI-561-2011-Final-Answers/

Powered by TCPDF (www.tcpdf.org)