This course is required for the Data-centric Computing Stream and for the Smart Systems Stream.

Prerequisites:  COMP 3200; or  COMP 2001 or the former COMP 2710,  COMP 2002 or the former COMP 2711, and Statistics 2550 or Statistics 2550; and Mathematics 2050

Availability: This course is usually offered once per year, in Fall or Winter.

Course Objectives

This course introduces concepts and algorithms in machine learning for regression and classification tasks. The course gives the student the basic ideas and intuition behind model selection and evaluation, and selected machine learning methods such as random forests, support vector machines, and hidden Markov models.

Representative Workload

• Assignments (5) 30%
• In-class Exam 25%
• In-class Participation 10%
• Final Exam 35%

Representative Course Outline

• Introduction to Machine Learning (3 hours)
• Definition and examples of machine learning tasks, such as: Classification
• Types of learning: supervised, unsupervised and reinforcement
• Linear methods for regression and classification (5 hours)
• Model Assessment and Selection (3 hours)
• Bias, variance, overfitting, and model complexity
• Measuring classifier performance (3 hours)
• Cross-validation
• Precision / Recall
• Area under ROC curve
• Supervised learning (6 hours)
• Nearest-neighbour
• Decision Trees
• Combining classifiers (6 hours)
• Boosting
• Random Forests
• Other approaches such as support vector machines, hidden Markov models, and so on. (4 hours)

Page last updated May 24th 2021