CS540 Spring 2006
Midterm I Study Guide
The exam will be held in class on March 10. It will cover all material discussed in class or presented in the readings through Genetic Algorithms. The basic topics are: Scheduling, Search and Genetic Algorithms. The readings are listed here. The point distribution will roughly match the amount of time spent in class or on assignments on the area.
To help you prepare, I am including some questions taken from previous exams. The points are just to give an idea of how muh a question might be worth; don't expect the set to sum to 100. I encourage you to ask (via email or in person) questions or seek help if you want to confirm your own answers. I will be checking my email during Spring Break.
Genetic Algorithms and
Local Search [39 points]
1) [12
points] Genitor:
For the initial population, determine the next population that results after one iteration of Genitor. Assume
v Maximizing fitness,
v Reduced surrogate recombination,
v Mutation rate of 0.1.
If you need random numbers, choose in order from the following:
.86, .59, .67, .14, .34, .08, .11, .29, .85, .76, .43, .47, .89, .80, .98, .58, .03, .57, .49, .92
repeat from beginning if you need more.
|
Initial Population |
Fitness |
New Population |
|
100010111 |
4 |
|
|
100000001 |
-10 |
|
|
010101010 |
-6 |
|
|
010100110 |
1 |
|
|
001100111 |
-5 |
|
|
110110110 |
21 |
|
2) [8 points] Hyperplanes: Explain why hyperplane deception poses problems for genetic algorithms.
3) [5 points] Selection: Why does selection usually include a stochastic
component?
4) [8 points] Convergence: State two aspects of a GA’s design that would cause a population to converge quickly? Refer to the two basic steps of GAs (selection and reproduction) in your answer.
5) Local Search [6 points]
Why do we try to avoid large neighborhoods when applying local search to a problem? Be as specific as possible about the effect of neighborhood size on local search algorithms.
1.
Scheduling
[25 points]
a. [20
points] A job shop problem has
the following routing table:
|
|
Oi,1 |
|
Oi,2 |
|
O1,3 |
|
|
Job |
Dur |
Machine |
Dur |
Machine |
Dur |
Machine |
|
A |
1 |
2 |
3 |
0 |
6 |
1 |
|
B |
8 |
1 |
5 |
2 |
10 |
0 |
|
C |
5 |
2 |
9 |
0 |
1 |
1 |
|
D |
5 |
1 |
5 |
0 |
5 |
2 |
Pick
one of the following two questions to answer about this problem. Circle the
number of the one that should be graded. Each tick on the timeline is 2 time
units.
i.
Compute the texture
for this problem on the timeline below. Circle the highest point of contention.
Which task should be fixed first in the schedule and what is its est?
ii. Construct the schedule given a job ordering of <A, B, C, D> and show it below. What is a critical path for it? What swap should be done under the N1 operator?
Timeline
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
b. [5 points] Why are applications with set-up times harder, in general, to
schedule?
2.
Search
[20 points]
a. [10 points] Based on your answer to the last question, describe how the
computation (texture or critical path) is used to bias search. State what
search algorithm you are assuming.
b. [10 points] For a search technique of your choosing, describe how stochasticity
is used and why.