IN4050-Introduction to Artificial Intelligence and Machine Learning - Spring 2025 Exam

Aug 30, 2026

Upload by Paura

IN4050 Introduction to Artificial Intelligence and Machine Learning Spring 2025 exam contains 13+ multiple-choice questions covering exploration versus exploitation, optimization algorithms, and core evolutionary algorithm concepts. Topics include evolutionary operators (mutation/crossover/recombination), properties preserved in permutation-based genotypes, parent selection with fitness proportional selection, and fitness sharing effects using gene-position similarity. It also tests diversity mechanisms, multiobjective versus single-objective differences, Pareto optimality, and reinforcement learning updates via Q-learning and SARSA.

LanguageEnglish

CategoryExam

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Exam Instructions

Question Formats and Rules

Permitted Materials and Language

Core Concepts and Multiple-Choice Questions

Exploration in Optimization

Optimization Algorithms

Evolutionary Algorithms (EA) Components

Evolutionary Operators and Mutation/Crossover

Fitness Proportional Selection and Probabilities

Fitness Sharing and Diversity Encouragement

Multiobjective vs Single-Objective EA

Pareto Optimality

Reinforcement Learning: Q-learning Update

SARSA Properties

Cross-validation Principles

i IN4050-Introduction to Artificial Intelligence and Machine Learning Spring 2025 May 27 03:00 PM-07:00 PM(4 hours) All questions should be answered! Each question is assigned a weight which is indicated.The maximum number of points for the whole set is 100 points.The course teachers will visit the exam room once around the middle of the exam time. Multiple Choice questions make up 70%of the points on the exam.For these,exactly one alternative is correct,unlessotherwise specified. Permitted materials:NoneAn on-screen calculator is available. You may answer in English,Norwegian,Danish or Swedish. # 1 In Search/Optimization,what is true about exploration?Select one alternative: OExploration is only used for continuous optimization OHillclimbing is a purely exploratory technique OExhaustive search is a purely exploratory technique O Gradient Descent does more exploration than exploitation Maximum marks:2 # 2 Which is true about optimization algorithms?Select one alternative: O Simulated annealing balances exploitation and exploration with the temperature parameter O Simulated annealing only selects neighbor solutions that are better than the current solution OHillclimbing has more randomness than simulated annealing O Gradient descent can be used for discrete optimization # 3 Which is typically not a part of an evolutionary algorithm(EA)?Select one alternative: OA termination criterion OA temperature control OA fitness function OA population Maximum marks:2 # 4 Which is not a variation operator in Evolutionary Algorithms?Select one alternative: O Fitness Sharing O Recombination ○ Crossover O Mutation Maximum marks:2 # 5 What property of the elements of a solution do we often try to conserve when doing mutation and crossover on a permutation-based genotype? Select one alternative: ○ Gradients O Concurrency O Sum OAdjacency Maximum marks:2 | Individual Genotype Fitness
no.
1
2
3 | | --- | | 1
01111 | | --- | | 00001
2 | | 3
10011 | | Individual Genotype
Fitness
no.
1
2
3 | | --- | 7 Consider the individuals listed here in an evolutionary algorithm.Assume this is the full population,and you are about to doparent selection from it.You will calculate the selection probability of each individual.Note that the probability you calculate isfor a single selection,not the probability that the individual gets selected into the pool of parents after N different selections.What is the selection probability of each individual given Fitness Proportional selection?Select one alternative: O Individual 1:1/3.Individual 2:1/3.Individual 3:1/3 O Individual 1:0.Individual 2:0.Individual 3:1 O Individual 1:1/6.Individual 2:2/6.Individual 3:3/6 Od)Individual 1:0.Individual 2:1/3.Individual 3:2/3 Maximum marks:2 | 1
01111 | | --- | | 00001
2 | | 3
10011 | Assume we now wanted to do fitness sharing during selections,with the same individuals as in the previous question.Howwould that affect their fitness scores?Assume the similarity measure between individuals is based on how many genes are inthe same location in their genotype.How would the probability of selecting each individual change?Select one alternative: O Individual 3 would become more likely to be selected,the other two less likely to be selected O Individual 1 would become more likely to be selected,the other two lesslikely to be selected OIndividual 2 and 3 would become more likely to be selected,individual 1 less likely to be selected O Individual 2 would become more likely to be selected,the other two less likely to be selected # 8 Which is not a method to encourage diversity in evolutionary algorithms?Select one alternative: O Fitness Sharing O Elitism O Crowding O Speciation # 9 Which two differ most between multiobjective and single-objective evolutionary algorithms?Select exactly two alt

FAQs

What is the role of exploration in optimization according to the exam questions?

Exploration is contrasted with exploitation across optimization methods, and the exam asks which statement about exploration is true.

Which items are typically part of an evolutionary algorithm and which are not?

The exam distinguishes core EA components like population and fitness function from elements such as temperature control that are asked as the exception.

How do fitness proportional selection and fitness sharing differ in selecting individuals?

Fitness proportional selection assigns probabilities based on fitness, while fitness sharing modifies effective fitness using a similarity measure, changing which individuals become more or less likely to be selected.

Aug 30, 2026

Upload by Paura