Quizes Flashcards

1
Q

A mathematical function always has a global minimum

A

No

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

local optima only occurs when f’(x) = 0

A

No. can also occur when f’(x) not defined.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is a feasable set?

A

A set that satisfies the constraints on the objective function.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What properties does the Hessian(f) Matrix have?
A. Symmetric
B. Diag(H) != 0
C. No elements are zero

A

A. Symmetric.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Consider a 2x2 matrix if 1s. Is semi-definite?

A

Yes. Since all eigenvals are >= 0

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Once the search direction has been find in an iterative opt. method, the problem of finding the step size is one-dimensional.

A

Yes. Step size is a scalar.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Newton-Rapson method always finds a minima.

A

No. It finds a stationary point s.t f’(x) = 0. Second order derivative test can settle (if it exists) if the point is max/min/saddle

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

An optimization problem is convex if f(x) is convex

A

No. It also requires the set of feasable solutions to be convex.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

At a local minima of f(x) s.t h(x) = 0 what of the following conditions hold?
A. grad(fx) and grad(hx) are paralell
B. grad(fx) and grad(hx) are orthogonal
C. grad(hx) = 0

A

A. grad(fx) and grad(hx) are paralell. (see, lagrange multipliers)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

If two individuals are to have offspring they must be of the same species

A

No. Counterexample: Mules and Ligers.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Humans have more chromosomes than other species

A

No. Human 48~ and carp 100+.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Why must codons consist of at least three letters?

A

Because there are 20 amino acids and 4^x where x < 3 doesn’t have enough combinations to encode all 20 different amino acids.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Because of random initialization stochastic algorithms converges at a slower rate.

A

False

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

In tournament selection, the competing number of individuals is always two

A

No. Sometimes more.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

After selection of two individuals, crossover is always performed.

A

No. This step depends on crossover probability.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

How many possible mutations can occur given m genes and p_m mut prob?

A

m number of mutations.

17
Q

In connection to the schema theorem and GAs, what is a building block?

A

A building block is fixed number of genes that:
A. Short defining length, i.e distance between first and last non-wild card symbol
B. “Low order”, nr of non-wildcard symbols.
C. High fitness

18
Q

f(j) = j(m-j) is a function of unitation.

A

Unitation is a function whose value only depends on the number of ones in the chromosome. This depends on only on the number of ones in the chromosome since m is a constant.

19
Q

Good option to avoid premature convergence is to rerun the GA a couple of times.

A

Yes. This is equivalent to island modelling, where crossover only happens between invdividuals on the same island.

20
Q

Why use two point crossover in LGP?

A

Dont know.

21
Q

In LGP, you have a limited number of constants that cannot be changed (TRUE/FALSE)

A

TRUE. Constants are set once and only once. New constants can be formed by using the others as building blocks.

22
Q

In Interactive Evoluationary Computation, as many pictures as possible should be presented to the evaluator

A

FALSE. This will cause user fatigue

23
Q

Biological neural networks are able to carry out complex tasks because neurons are fast?

A

FALSE. Neurons have a firing capacity of 1khz, which is not that fast compared to a modern computers 3 ghz. Its the combination of many computational units that makes the animal brain so good at complex tasks.

24
Q

Gene regulation plays a role in long-term memory learning tasks.

A

TRUE. Gene regulation causes neurons to permanently change state whereas

25
feed forward neural network requires backpropagation to train.
TRUE. It just does.
26
When training a neural network, one should do the training for as long as possible.
FALSE. A good rule-of-thumb is to train until your testing set error starts increasing after decreasing.
27
Ants have skilled leaders and are capable of launching nuclear missiles.
FALSE. Ants appearent smart behaviour is an emergent property of communication using pheromones.
28
What is stigmergy? A. A mechanism for long-range communication using sound. B. A mechanism for indirect communication by means of (local) modification of the environment. C. A mechanism for long-range communication using vision.
B. A mechanism for indirect communication by means of local modification of the environment.