FIT5201 · Machine learning

Overview

This unit introduces machine learning and the major kinds of statistical learning models and algorithms used in data analysis. Learning and the different kinds of learning will be covered and their usage will be discussed. The unit presents foundational concepts in machine learning and statistical learning theory, e.g. bias-variance, model selection, and how model complexity interplays with model's performance on unobserved data. A series of different models and algorithms will be presented and interpreted based on the foundational concepts: linear models for regression and classification (e.g. linear basis function models, logistic regression, Bayesian classifiers, generalised linear models), discriminative, probabilistic, and generative models, non-parametric models (e.g., k-nearest neighbour, Gaussian process regression), k-means and latent variable models (e.g. Gaussian mixture model), expectation-maximisation, and neural networks and deep learning. Moreover, implementation techniques will be introduced and practiced that allow to practically implement the introduced algorithms in a scalable manner with robust and standardised interfaces.

Areas of study: Advanced data analytics

Offerings

Assessment

The Handbook lists an examination for this unit.

For on-campus offerings: This unit has threshold mark hurdles. You must achieve at least 45% of the available marks in the final scheduled assessment, at least 45% in total for in-semester assessments, and an overall unit mark of 50% or more to be able to pass the unit. If you do not achieve the threshold mark, you will receive a fail grade (NH) and a maximum mark of 45 for the unit.

Assessment details may change. Please refer to the assessment information in Moodle closer to the start of the teaching period.

Requisites

Learning outcomes

1. Describe the components and theoretical concepts of statistical machine learning;
2. Assess and explain theoretically the performance of machine learning approaches and derive recommendations for algorithm and model selection;
3. Derive and implement the most widely used machine learning models and algorithms and apply them to real-world and synthetic datasets;
4. Develop scalable and standardised implementations of typical machine learning algorithms using suitable programming techniques and libraries;
5. Describe and discuss ethical challenges when deploying machine learning systems in practice.

Workload

Minimum total expected workload to achieve the learning outcomes for this unit is 144 hours per semester typically comprising a mixture of scheduled online and face to face learning activities and independent study. Independent study may include associated reading and preparation for scheduled teaching activities.

Student experience, not official rules. Nothing here changes what the Handbook says.