Scribd Logo
Upload

Welcome to Scribd!

  • 15 views

    ICS 3202 - Artificial Intelligence - December 2022

    Uploaded by

    nyamosiyego
    This document contains instructions and questions for an end of semester examination on Artificial Intelligence. It includes 5 questions, with Question 1 being compulsory and students required to answer any 2 of the remaining 4 questions. Question 1 covers topics like handling missing data, classification algorithms, neural networks and their activation functions, reducing overfitting, supervised vs unsupervised vs reinforcement learning. Other questions cover topics such as components of an AI agent, natural language processing, problem solving agents, heuristics, fairness in AI and ethical concerns regarding AI adoption.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    ICS 3202 - Artificial Intelligence - December 2022

    Uploaded by

    nyamosiyego
    15 views5 pages
    This document contains instructions and questions for an end of semester examination on Artificial Intelligence. It includes 5 questions, with Question 1 being compulsory and students required to answer any 2 of the remaining 4 questions. Question 1 covers topics like handling missing data, classification algorithms, neural networks and their activation functions, reducing overfitting, supervised vs unsupervised vs reinforcement learning. Other questions cover topics such as components of an AI agent, natural language processing, problem solving agents, heuristics, fairness in AI and ethical concerns regarding AI adoption.

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document contains instructions and questions for an end of semester examination on Artificial Intelligence. It includes 5 questions, with Question 1 being compulsory and students required to answer any 2 of the remaining 4 questions. Question 1 covers topics like handling missing data, classification algorithms, neural networks and their activation functions, reducing overfitting, supervised vs unsupervised vs reinforcement learning. Other questions cover topics such as components of an AI agent, natural language processing, problem solving agents, heuristics, fairness in AI and ethical concerns regarding AI adoption.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    15 views5 pages

    ICS 3202 - Artificial Intelligence - December 2022

    Uploaded by

    nyamosiyego
    This document contains instructions and questions for an end of semester examination on Artificial Intelligence. It includes 5 questions, with Question 1 being compulsory and students required to answer any 2 of the remaining 4 questions. Question 1 covers topics like handling missing data, classification algorithms, neural networks and their activation functions, reducing overfitting, supervised vs unsupervised vs reinforcement learning. Other questions cover topics such as components of an AI agent, natural language processing, problem solving agents, heuristics, fairness in AI and ethical concerns regarding AI adoption.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 5

    SCHOOL OF COMPUTING AND ENGENEERING SCIENCES

    BACHELOR OF INFORMATICS AND COMPUTER SCIENCE


    ICS 3202: ARTIFICIAL INTELLIGENCE
    END OF SEMESTER EXAMINATION

    DATE: 6th December 2022 Time: 2 Hours


    INSTRUCTIONS
    1. Answer Question 1 (Compulsory) and any other Two Questions.
    2. In the calculations, show your workings for full marks.

    QUESTION One: - COMPULSORY 30 Marks

    a. An ICS student is in the process of developing a machine learning solution for a Classification
    task. As such they need to prepare their data and then use it for training alongside a relevant
    algorithm.
    i. During data preparation they have to handle missing values. Discuss any two ways for
    handling missing values. (2 Marks)
    ii. Assuming they opt to use kNN algorithm, what are some of the metrics to adhere to
    when picking the value of k? (2 Marks)
    iii. Apart from kNN, what other four classical algorithms would be applicable is the task
    described above? (2 Marks)
    b. Using a relevant diagram, describe a perceptron. (3 Marks)
    c. List any four-activation functions that can be used in an artificial neural network e.g. the one
    from question “d” above (2 Marks)
    d. Suppose you are training a neural network model and notice it is overfitting on the training
    data, what three ways can be used to reduce chances of overfitting. (3 Marks)

    1
    e. By providing relevant supporting examples, discuss each of the following three terms and
    demonstrate how they differ from each other. (6 Marks)
    i. Supervised Learning
    ii. Reinforcement Learning
    iii. Unsupervised Learning

    f. Draw a truth table for the following propositional formula (4 Marks)


    ~𝑝 𝑜𝑟 ~𝑞 ~𝑞 ^ ~𝑟

    g. Translate the sentences below into first order predicate logic. (6 Marks)
    i. All students are smart.
    ii. There exists a student.
    iii. There exists a smart student.
    iv. Every student likes some student.
    v. Every student likes some other student.
    vi. There is a student who is liked by every other student.

    2
    QUESTION TWO 15 Marks
    a. What are the building blocks of Artificial Intelligence (2 Marks)
    b. Using a relevant diagram, describe they key components of an AI agent. (4 Marks)
    c. You’ve been appointed to assist in developing an AI used by a self-sailing public boat
    sponsored by R4M Auto Inc. Given that a Task environment can be described using four main
    areas, use relevant examples to describe the four areas for this task environment (4 Marks)
    d. Two students working on an IS Project plan to create AI agents with a goal of predicting share
    prices; these agents will later be set free in the same environment. Describe the two ways in
    which these agents can be deployed/set up (2 Marks)
    e. You are provided with the following confusion matrix for a Binary Classification task.

    Positive Negative
    Positive 130 17
    Negative 10 125

    Using the values in the confusion matrix, calculate the following performance metrics. Show
    your working / formulas for each. (3 Marks)
    i. Accuracy:
    ii. Precision:
    iii. Recall:

    QUESTION THREE 15 Marks


    a. NLP applications range from sentiment classification, Machine Translation and Social Media
    Monitoring. However, the same is not 100% true when dealing with Low resource Languages
    such as Swahili, Zulu, Kinyarwanda etc. What are some of the emerging techniques being used
    to resolve challenges for low resource NLP applications? (3 Marks)
    b. Suppose you are doing a multiclass classification task where the raw input is a single string
    containing the text of the entire document. Using a diagram describe the pipeline to go from
    the raw input to a predicted class with possibility of improving the model in future. (6 Marks)
    c. In an NLP pipeline, Bag of Words (BoW) can be used for feature extraction. Given the
    following corpus of tweets by customers, generate a vocabulary and the corresponding vectors
    when using BoW. Do not perform any stop word removal on the given corpus. (4 Marks)
    Review 1: This movie is very scary and long

    3
    Review 2: This movie is not scary and is slow
    Review 3: This movie is spooky and good

    d. Referring to the examples in question ‘c’ above, what are the drawbacks of using a Bag-of-
    Words (BoW) Model. (2 Marks)

    QUESTION FOUR 15 Marks


    a. Describe the three steps followed by Problem Solving Agents when attempting to reach a Goal
    State? (3 Marks).
    b. Company XYZ deploys sales-agents across Kenya every 2 to 3 months. To ensure efficient
    deployment and travel of sales-agents from Nairobi to other destinations, the company plans
    to develop some sort of intelligent agent that can return efficient pathways that sales-agents
    can use while traveling.
    i. During development, the company discovers lots of loops returned by initial prototypes of
    the intelligent agent. Advice on ways of how this huddle in development can be overcome.
    (2 Marks)
    ii. Describe the Goal state and the Initial State for the problem above. (2 Mark)
    c. The graph below presents one Starting Point ‘S’ and 2 Possible Destination points G1, G2. Use
    the provided Heuristics (Indicated within the node e.g for node S the value is 5, and for Z the
    value is 7).

    4
    i. By showing your expansion approaches from the source ‘S’ to the other nodes, write in
    the final A-star scores for all nodes. (5 Marks)
    Node Score
    S
    X
    Z
    U
    V
    W
    Q
    G1
    G2
    S

    ii. State the pathways to the specified goals when using the following algorithms (3 Marks).
    i. Greedy Search (To G2)
    ii. A-Star Search (To G1)
    iii. Iterative Deepening Depth First Search (Consider shortest path to any goal with
    a Limit of 3)

    QUESTION FIVE 15 Marks

    a. There are many ways of defining what we might look for in a fair artificial intelligence system.
    Describe any two criteria which can be used to describe fairness in AI models. (1 Mark)
    b. In their research paper titled A Framework for Understanding Sources of Harm throughout the
    Machine Learning Life Cycle, HARINI SURESH and JOHN GUTTAG discussed various
    types of bias that may arise during development and deployment of AI systems. By
    using/providing relevant examples, discuss any 5 types of bias in AI systems. (10 Marks).
    c. What are some of the ethical concerns with the recent adoption of AI in business sectors such
    as prediction of cryptocurrency prices (4 Marks)

    You might also like