CO3519 Artificial Intelligence
CO3519 Lecture 2 - AI Types & Areas in AI

Lecture Documents¶

CO3519 Lecture 3.pdf

Written Notes¶

Learning Objectives¶

• Machine Learning (ML)
• Types of Machine Learning
• Relation between ML & AI
• Key steps in developing ML model

Machine Learning¶

What is Machine Learning¶

Types of Machine Learning¶

• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning

Supervised Learning¶

Supervised learning the training of a ML model on labelled datasets where both input data and the correct output (label) are known.

Key Details¶

The algorithm learns by mapping inputs to their corresponding outputs.
- The algorithm learns by mapping inputs to their corresponding outputs.
- Goal is to predict labels for new unseen data.
- Performance is measured using the metrics:
- Accuracy
- Precision
- Recall
- F1-Score

Common Algorithms¶

• Linear Regression - Predicts continuous values.
• Logistic Regression - Used for binary classification.
• Support Vector Machines (SVM) - Finds the optimal separating hyperplane.
• Decision Trees - Tree-like models for classification or regression.

Lecture Example¶

A database containing both cat and dog images are separated and labelled before being processed into a trained model which identify which are cat or dog by their assigned label.

Unsupervised Learning¶

Unsupervised learning trains a model on unlabelled data, meaning the correct outputs are unknown.

Key Details¶

• The algorithm discovers hidden patterns or structures in data.
• Goal is to group or reduce data dimensions without prior labels.
• Performance is often evaluated using internal metrics (e.g. Silhouette Score, Davies–Bouldin Index).

Common Algorithms¶

• K-Means Clustering – Groups data into k clusters based on similarity.
• Hierarchical Clustering – Builds a tree of clusters.
• Principal Component Analysis (PCA) – Reduces dimensionality while preserving variance.
• Association Rule Learning (Apriori) – Finds relationships between variables in large datasets.

Reinforcement Learning¶

Reinforcement learning (RL) trains an agent to make successive decisions by interacting with an environment to maximise rewards and minimise punishment.

Key Details¶

• No labelled data — learning via trial and error.
• The agent observes a state, performs an action, and receives a reward.
• Goal is to learn a policy that maximizes total reward over time.
• Evaluated by convergence speed and long-term reward.

Common Algorithms¶

• Q-Learning – Learns value of actions using Q-values.
• Deep Q-Network (DQN) – Combines Q-learning with neural networks.
• Policy Gradient Methods – Directly optimize the policy function.
• SARSA – On-policy version of Q-learning.

Lecture Example¶

Positive reinforcement on a chicken.

Agent - Chicken
Action - Touching PINK circle.
Reward - corn grain

Key Steps in Developing an ML Model¶

flowchart TD
    A[Problem Definition • Clearly define the task you are solving] --> 
    B[Data Collection • Collect data for training/testing] --> 
    C[Data Pre-processing • Clean and transform data] --> 
    D[Features Extraction • Transform data into measurable properties] --> 
    E[Model Selection • Choose algorithm based on problem type] --> 
    F[Model Training / Evaluation • Use training data to optimize parameters]

    classDef default fill:#007ACC,color:#ffffff,stroke:#333,stroke-width:1px;

Lab Activity¶

Solve house price prediction problem using LR.

Next Week¶

• Key Challenges in Machine Learning
• Supervised Learning
• Linear Regression
• Logistic Regression
• Support Vector Machines (SVM)
• Decision Trees

CO3519 Lecture 4 - Machine Learning II