Lecture 09 - Neuroevolution

Question 1

Neuroevolution

What is neuroevolution?

Answer 1

Artificial evolution of neural networks using genetic
algorithms (GA)

Question 2

Neuroevolution

Where has neuroevolution shown promise?

Answer 2

Reinforcement learning tasks, such as robot control and
playing a game.

Question 3

Neuroevolution

What is NEAT short for?

Answer 3

Neuroevolution of Augmenting Topologies

Question 4

Neuroevolution

What is TWEANN short for?

Answer 4

Topology and Weight Evolving Artificial Neural Network

Question 5

Neuroevolution

What’s the relationship between NEAT and TWEANN?

Answer 5

TWEANN is NEAT’s predecessor.

Question 6

Neuroevolution

What is Direct encoding?

Answer 6

Genome explicitly defines the network

Question 7

Neuroevolution

(@Rephrase) What is indirect encoding?

Answer 7

Genome has developing rules to grow the network

Question 8

Neuroevolution

What is the advantage of direct encoding?

Answer 8

Genome is smaller

Question 9

Neuroevolution

What is the disadvantage of direct encoding?

Answer 9

Reduced topology variety

Question 10

Neuroevolution

What is NEAT’s solution to the Competing Conventions Problem?

Answer 10

Historical Marking

Question 11

Neuroevolution

What is Historical Marking in NEAT?

Answer 11

Tracking of “innovations” by enumerating them with a unique ID called the “innovation number”.

Question 12

Neuroevolution

What is the innovation number?

Answer 12

A rising ID number for each new “innovation” in NEAT.

Question 13

Neuroevolution

What is an innovation in NEAT?

Answer 13

Any new gene in a genotype, i.e. addition of neurons (nodes) and connections/weights (vertices).

Question 14

Neuroevolution

How does historical marking work during crossover?

Answer 14

• Matching genes are inherited randomly

Question 15

Neuroevolution

What are disjoint genes in NEAT?

Answer 15

those that do not match in the middle

Question 16

Neuroevolution

What are genes that do not match in the middle called?

Answer 16

Disjoint genes

Question 17

Neuroevolution

What are excess genes in NEAT?

Answer 17

those that do not match in the end

Question 18

Neuroevolution

What are genes that do not match in the end called?

Answer 18

Excess genes

Question 19

Neuroevolution

Which parent are disjoint/excess genes inherited from?

Answer 19

The more fit parent

Question 20

Neuroevolution

How does NEAT protect innovation?

Answer 20

Speciation

Question 21

Neuroevolution

Why does NEAT use Speciation?

Answer 21

To protect innovation

Question 22

Neuroevolution

What is speciation?

Answer 22

Separation orf solutions into “species”, where only individuals within the same species are allowed to reproduce with each other.

Question 23

Neuroevolution

What’s the requirement for a species?

Answer 23

Having similar chromosomes

Question 24

Neuroevolution

Describe the parts of the compatibility distance.

Answer 24

• cs: importance coefficients
• E: #excess genes
• D: #disjoint genes
• N: #genes in the larger genome
• W: Average weight differences of matching genes.

Question 25

# Neuroevolution Describe the speciation process. (4)

Answer 25

- Using the distance measure "d" and a threshold "d_t". - A random genome from the previous generation represents a species. - For every genome: - If the genome is compatible with a species, add it to that species. - Otherwise, create a new species from that genome.

Question 26

# Neuroevolution What is adjusted fitness?

Answer 26

The individual's non-adjusted fitness divided by #individuals in the population.

Question 27

# Neuroevolution Why is adjusted fitness used?

Answer 27

The adjusted fitness allows the balance of members among all species in a generation

Question 28

# Neuroevolution How does speciation protect innovation?

Answer 28

The number of offspring of a species is proportional to the sum of adjusted fitnesses of its members

Question 29

# Neuroevolution How does extinction work?

Answer 29

If the maximum fitness of a species did not improve in 15 generations, the networks in the stagnant species are not allowed to reproduce

Question 30

# Neuroevolution How does elitism work?

Answer 30

the champion of each species with more than five networks are copied into the next generation unchanged

Question 31

# Neuroevolution What are the types of mutations in NEAT? (3)

Answer 31

- Add connection - Add node - Change weight connection

Question 32

# Neuroevolution What is HyperNEAT?

Answer 32

A "Hypercube-based NEAT". An indirect encoding where a CPPN defines the connectivity.

Question 33

# Neuroevolution What is CPPN short for?

Answer 33

Compositional pattern-producing network

Question 34

# Neuroevolution What does a Compositional pattern-producing network (CPPN) do?

Answer 34

It produces a geometric pattern.

Question 35

# Neuroevolution How does HyperNEAT use the CPPN?

Answer 35

HyperNEAT uses CPPN to define the weights of the ANN that executes the desired task

Question 36

# Neuroevolution How is the CPPN trained?

Answer 36

CPPN is trained using NEAT

Question 37

# Neuroevolution What are the activation functions in aa CPPN?

Answer 37

Pretty much any activation function, like sigmoid, cosine, linear, absolute value etc.