6. Genetic Algorithms

Question 1

How is the Darwinian theory (the principle of the survival of the fittest) of evolution applied as a model for technical optimization procedures?

Answer 1

The take on optimization inspired by the Darwinian theory of evolution has been used as a model for technical optimization procedures. Here the environment is static though.

Question 2

How does the evolution algorithm work?

Answer 2

An evolutionary algorithm starts by randomly generating a set of potential solutions, which are evaluated using a fitness function. The best solutions are selected and evolved over successive generations. This process repeats until a near-optimal solution is found and the fitness function stabilizes.

Question 3

What is a population in the evolutionary algorithm?

Answer 3

A set of feasible solutions that may contain the best solution for a specific problem.

Question 4

How does the technique that is used for Genetic A work?

Answer 4

This technique evolves a population by combining and modifying traits, similar to how offspring inherit genes from parents. The fittest individuals are chosen to create the next generation, gradually improving toward the goal.

Question 5

What evolution algorithms do we know of?

Answer 5

Genetic Algorithms, Swarm Algorithms, Ant Colony algorithms

Question 6

Hoes does the Swarm A work?

Answer 6

Simulate the movement of a group of animals (e.g., fish or birds) towards a specific target.

Question 7

How does the Ant Colony A work?

Answer 7

These algorithms mimic how ants find the shortest path to food. As ants travel, they leave a chemical trail called a pheromone. Other ants follow stronger pheromone trails, reinforcing the path. This indirect communication, called “stigmergy,” helps them work together efficiently.

Question 8

What is a common use care in feature selection?

Answer 8

The algorithm picks the best combination of features to improve a machine learning model’s performance.

Question 9

GAs are commonly used in hyperparameter tuning. What is it?

Answer 9

A variable that influences a model’s behavior and output but is not adjusted by the training process. Examples of hyperparameters are the number of clusters in k-means, or limits on the depth of trees in decision tree learning. In particular, GA gained notable popularity in the optimization of structures and learning rules for artificial neural networks.

Question 10

How are the individual properties represented?

Answer 10

By a piece of code that can be modified to yield new examples. (1000011, 1000101)

Question 11

Is global optimalit achieved?

Answer 11

No. It’s a heuristic technique which can´t guarantee it. But the results are suficient for many purposes.

Question 12

How do genetic changes happen?

Answer 12

Crossover and mutation of the genetic structure.

Question 13

What are the five phases in GA?

Answer 13

1. Creation of population 2. Evaluation of fittest function 3. Selection of fittest 4. Crossover 5. Mutation

Question 14

What is the first phase about?

Answer 14

The individuals of the population are generated randomly in such a way that the corresponding chromosomes have a high coverage of the solution domain. The number of genes in individuals’ chromosomes is dependent on the problem.

Question 15

What is the input and output in the fittest function?

Answer 15

Input –> the chromosone of the individual, Output –> fittest score of the individual

Question 16

What is the selection phase about?

Answer 16

The individuals with the highest fitness scores are selected for generating the new population. Individuals with a fitness score that is higher than a specific threshold are selected, where the threshold constitutes a parameter that is defined according to the problem.

Question 17

What is cross-over about?

Answer 17

Mixing the genes from the parents’ pair of chromosomes. There are two types of crossover methods: single- and multi-point

Question 18

What is locus?

Answer 18

A specific location, or locus, is a point in the chromosome.

Question 19

Explain single and multi-point cross-over.

Answer 19

In single-point crossover, the procedure randomly chooses a locus and exchanges the subsequences before and after that locus between two chromosomes to create two offspring. Multi-point crossover selects a set of points at random. Chromosomes are cut at the crossover points. Corresponding sections are swapped.

Question 20

What is mutation about?

Answer 20

The flipping of random genes at arbitraty locations in the individual chromosome.

Question 21

How does the final phase play out?

Answer 21

The genetic algorithm goes through multiple iterations, where each iteration is formed by the above mentioned five phases. The iteration stops when an acceptable level of fitness is achieved by at least one individual, when no further improvement in fitness can be achieved, or when a pre-defined maximum number of iterations has been reached.

Question 22

The application of a genetic algorithm to solve the knapsack problem.

Answer 22

To solve the knapsack problem with a genetic algorithm, potential solutions are represented as chromosomes, where each gene indicates if an item is included in the knapsack. The fitness function evaluates each solution based on the total value of selected items, penalizing those that exceed the knapsack’s weight limit. The algorithm selects the fittest solutions, applies crossover and mutation to create a new generation of solutions, and repeats this process until it finds an optimal or near-optimal solution that maximizes value without exceeding the weight limit.