Academic Year 2024/25

M33615 Foundation of Artificial Intelligence

Coursework

Deadline For Submission: 15th January 2025 by 12:00 pm (UK time)

Hand-in Instructions: The submission process requires two main documents to be uploaded
to Moodle:
1. A report detailing your findings and methodologies.
2. Python notebooks or Matlab code used for Task3. This is to show
the steps taken to generate the sections of the report.
You have the option to submit your code in the form of a Google
Colab or Jupyter Notebook file; however, a simple plain text file is also
fine. This ensures that the work is easily accessible and assessable.

Instructions for completing the For your submission, please adhere to the following guidelines:
assessment:
1. The report must be submitted in PDF format, not as a
Microsoft Word document.
2. Both the report and the code file - should be named using
your student ID (for example, UP1234567).
3. Do not submit any handwritten work that has been scanned;
this format is not acceptable.
4. The report should not exceed 20 pages of A4 size. Any pages
beyond this limit will not be considered for marking.
5. There is a word limit of 4000 words for the report, which does
not include code or references.
6. Ensure that your student ID number is included on each page
of your report; placing it in the header or footer is
recommended for consistency.
Additionally, be mindful of the university's stance on plagiarism. The
submission of work that is not your own or heavily based on someone
else's work without appropriate credit is considered plagiarism and is
subject to university sanctions. It's important to ensure that your
work is original and properly cites any sources used.

Examiners: Dr Hongjie Ma
Task1: Energy Consumption Prediction for Small Businesses (20 marks)
Background:
As an energy efficiency consultant, you're helping small businesses predict and manage their monthly energy
consumption. You've been asked to develop a simple model to estimate energy usage based on two factors:
the average daily temperature and the number of employees.

Data Description:
You are provided with the following dataset of 5 small businesses:

Table T1: Monthly Energy Usage Data for 5 Small Businesses

Business ID Avg. Daily Temp (°C) Number of Employees Monthly Energy Usage (kWh)
1 18 10 3000
2 22 15 3800
3 20 12 3400
4 25 20 4500
5 16 8 2700

Task 1a) Data Analysis (5 marks)

1. Calculate the mean average daily temperature and mean number of employees. (3 marks)
2. What is the range (difference between maximum and minimum values) for monthly energy usage?
(2 marks)
Task 1b) Model Understanding (5 marks)
1. Write the general form of the hypothesis function h(x) for this multiple linear regression model. (2
marks)
2. Explain in simple terms what the cost function J(θ) represents in a linear regression model. (3 marks)
Task 1c) Model Application (5 marks)
Given the following trained model: h(x) = 1000 + 80x₁ + 100x₂ Where x₁ is the average daily temperature
(in °C) and x₂ is the number of employees.
1. Interpret the meaning of each parameter (1000, 80, and 100). (3 marks)
2. For a business with an average daily temperature of 21°C and 18 employees, predict its monthly
energy usage. (2 marks)
Task 1d) Reflection (5 marks)
1. Explain one limitation of this linear regression model for predicting energy consumption. (2 marks)
2. Suggest one additional feature that could improve the model and explain why. (3 marks)

2024-25 Page 1 of 3
Task2: Critical Analysis and Application of Machine Learning Concepts (40 marks,
suggested word count: 1500 words)

This task evaluates your understanding of fundamental machine learning principles and your ability to apply
them to real-world scenarios. For each subtask, provide clear explanations, relevant examples, and where
appropriate, simple diagrams or mathematical expressions to support your answers.

Task 2a) Supervised vs Unsupervised Learning (10 marks, suggested word count: 400 words)

1. Compare and contrast supervised and unsupervised learning, discussing their key differences and
similarities. (4 marks)
2. Identify an emerging technology or industry trend. Propose and explain how both supervised and
unsupervised learning could be applied in this context, highlighting the potential benefits and
limitations of each approach. (6 marks)

Task 2b) Model Complexity and Performance (8 marks, suggested word count: 300 words)

1. Analyse the trade-offs between model complexity, performance, and generalisation in machine
learning. Discuss how these factors relate to overfitting and underfitting. (4 marks)
2. Propose and explain two strategies, besides regularisation, that could help improve model
generalisation. Provide examples of when each strategy might be most effective. (4 marks)

Task 2c) Gradient Descent and Model Training (8 marks, suggested word count: 300 words)

1. Explain the working principle of the gradient descent algorithm in the context of model training.
Discuss how the learning rate affects the training process and the challenges in selecting an
appropriate learning rate. (4 marks)
2. Compare the application of gradient descent in linear regression and logistic regression. Discuss the
key differences in the cost function and update rule between these two scenarios. Provide a simple
example to illustrate your points. (4 marks)

Task 2d) Model Evaluation and Overfitting (8 marks, suggested word count: 300 words)

1. Discuss the importance of choosing appropriate evaluation metrics in machine learning. Compare
and contrast at least two common metrics (e.g., accuracy, precision, recall, F1-score), and explain
scenarios where each might be most appropriate. (4 marks)
2. Describe the concept of cross-validation. Explain how it can help in assessing a model's potential for
overfitting and in selecting optimal hyperparameters. (4 marks)

Task 2e) Basic Neural Networks and Convolutional Neural Networks (6 marks, suggested word count: 200
words)

1. Describe the basic structure and computation process of a simple feedforward neural network.
Explain the role of activation functions in this process. (3 marks)
2. Explain the fundamental concept of convolution in the context of Convolutional Neural Networks
(CNNs). Discuss why this operation is particularly effective for processing image data. (3 marks)

Note: In your answers, prioritise demonstrating a clear understanding of the fundamental concepts. Use
specific examples to illustrate your points, and don't hesitate to draw connections between different areas
of machine learning. While you should aim to be comprehensive, focus on depth of understanding rather
than breadth of knowledge beyond the course material.

2024-25 Page 2 of 3
Task3: Innovative Machine Learning Application Project (40 marks, suggested word
count: 1600 words)

In this task, you will have the opportunity to demonstrate your understanding of machine learning concepts
and apply them to a real-world problem of your interest. You are required to select a dataset, design and
implement a machine-learning solution, and evaluate its performance.

Task 3a) Problem Definition and Dataset Selection (10 marks, suggested word count: 400 words)
1. Choose a real-world problem that interests you and explain why it is suitable for a machine-learning
approach.
2. Find or create a dataset appropriate for addressing this problem. If using an existing dataset, provide
the source; if collecting your own data, describe the data collection process.
3. Describe your dataset, including key information such as features, target variables, and data volume.
4. Discuss potential challenges or ethical considerations in using this dataset.

Task 3b) Data Preparation and Exploratory Analysis (8 marks, suggested word count: 300 words)
1. Describe your methods for data cleaning and preprocessing.
2. Conduct exploratory data analysis, using appropriate visualisation techniques to showcase key
features or patterns in the data.
3. Based on your analysis, propose at least two insights or hypotheses about the data.

Task 3c) Model Design and Implementation (10 marks, suggested word count: 400 words)
1. Select at least two different types of machine learning models to address your problem. Explain your
choices.
2. Implement your chosen models. You may use any suitable libraries or frameworks, or write from
scratch.
3. Describe how you tuned model parameters and why you made these adjustments.

Task 3d) Model Evaluation and Comparison (8 marks, suggested word count: 300 words)
1. Design and implement an evaluation strategy to test your models' performance.
2. Compare the performance of different models, discussing the strengths and weaknesses of each.
3. Based on your evaluation results, propose suggestions for improving model performance.

Task 3e) Innovation and Practical Application (4 marks, suggested word count: 200 words)
1. Discuss any innovative elements or unique approaches in your solution.
2. Explore how your model could be applied in a real-world scenario and potential challenges in doing
so.

Note:
1. Your report should include code snippets, visualisation results, and necessary mathematical formulae to
support your arguments.
2. You are encouraged to use techniques learned in the course, but also welcome to explore and apply
methods beyond the course scope.
3. Clearly explain your thought process and decision-making, especially when facing challenges or making
trade-offs.
4. Code should be clear, well-commented, and follow good programming practices.

2024-25 Page 3 of 3