COMP5328: Advanced Machine Learning

Learning Outcomes

• Algorithm design & evaluation: Present the design of an ML algorithm, including experimental evaluation and reporting.
• Bias–variance trade-off: Explain and reason about variance, bias, under/overfitting, and regularisation.
• Algorithm analysis & improvement: Analyse representative ML algorithms and identify avenues for performance/robustness improvements.
• Problem framing & model adaptation: Formulate learning problems and adapt existing models to new objectives or constraints.
• Paper-to-code implementation: Implement algorithms described in peer-reviewed research papers.
• Statistical foundations: Understand the statistical principles behind learning algorithm design/adaptation.
• Model suitability: Compare introduced models—their strengths, weaknesses, and appropriate use cases.

Takeaways

This course strengthened my ability to use statistical modelling as a practical tool for understanding data-generating processes and supporting business decision-making, rather than treating models as purely technical artifacts. By learning how to specify, estimate, and interpret statistical models, I developed a clearer understanding of how relationships between variables can be quantified, tested, and translated into meaningful insights. Working with Python to manage data, build models, and visualize results reinforced the end-to-end nature of applied data analysis, from raw data preparation to communicating conclusions. The emphasis on interpreting coefficients, uncertainty, and model assumptions improved my ability to reason about causality, limitations, and robustness, while forecasting and relationship analysis highlighted how statistical models inform planning under uncertainty. Equally important, the course focused on communicating results clearly to non-technical stakeholders, ensuring that statistical evidence can be used effectively in real decision contexts. Overall, this course provided a strong statistical foundation that complements optimisation and decision analysis, forming a core pillar for advanced work in data science, analytics, and AI-driven decision support systems.