-
Notifications
You must be signed in to change notification settings - Fork 95
/
Copy pathquiz
63 lines (50 loc) · 2.46 KB
/
quiz
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
constraints - why we didn't encode a constraint handling mechanism in the GA we used in the exercise for the TSP?
It's already there
The encoding used for the problem restricts the search space**
It isn't there and we need to add.
We always can find good solutions for the TSP.
Main differences between ga and es?
Population size**
There is crossover in GA and not in ES.**
There is mutation in GA and not in ES.
ES performs better than GA.
What are the benefits of using a population?
Obtain useful information about the search space.**
Historical value.
We don't need to use a population.
We can use mutation more efficiently.
Why do we care about pareto dominance?
We want to select one solution to be the elite.
We want to find many solutions that perform differently to each objective.**
We don't need to worry about it, the algorithm codes it.
We don't, it's possible to compare solutions as in GA.
What are the components that distinguish NSGA-II to the other algorithms?
Crowding distance and non-dominated sorting, in this order.
Non-dominated sorting and crowding distance, in this order.**
The ability to perform well as the number of objective increases.
Crossover and mutation.
Are the problems with calculating dominance?
No, as long as the algorithm uses the non-dominance.
Yes, finding if a solution is dominant is computationally expensive.
No, it's part of the problem and we just have to acknowledge that.
Yes, finding if solutions are dominant as the number of problems increase is computationally expensive.**
How can we deal with problems with non-dominance?
We can't.
We can increase the population size.**
Consider the distribution of solutions in the search space.
Only work with more than 3 problems.
Look at the pymoo webpage, under the problems category. Can NSGA-II deal with:
Work with binary or discrete optimization only.
Work with all types of problems.**
Work with continuous problems only.
For continuous problems, it's better to use ES.
About visualizations, why shouldn't we use them?
Images are worth a thousand words.
They help us understand differences in performance.
They provide all information necessary to compare algorithms.**
We can see how well an algorithm is performing.
What it's impossible to verity when using visualizations for multi-objective problems?
Quality of solutions.
Performance convergence.
Distribution of solutions in the objective space.
Distribution of solutions in the search space.**