COMP529 Research Topics in AI

Computing & Information Systems · Research capstone

COMP529 Research Topics in Artificial Intelligence

The final step of the sequence. Read and critique the frontier literature, reproduce published results, formulate your own research question, and carry out an original investigation to the standard of a conference submission — then defend it at a research symposium.

Explore the curriculum View syllabus
8 modules 16 units 130 tutorial pages 5 research tracks

Course overview

From AI practitioner to AI researcher

COMP529 is not another tools course. You already know how to build, deploy and operate AI systems. This course teaches you how to contribute new knowledge to the field.

There is a substantial difference between applying an existing method and producing a new one. Applying a method requires understanding how to use it. Producing one requires understanding why it works, where it fails, what has already been tried, and how to design an experiment that would convince a sceptical reader.

COMP529 develops that second capability. You begin by learning to read research critically — mapping a field, identifying assumptions, spotting weak evaluation. You then reproduce a published result, which is often the most instructive experience in the course: papers are frequently harder to replicate than they appear.

From there you formulate your own research question, design and run experiments, analyse results honestly, and write up your work in conference format. The semester closes with a research symposium in which you present, defend and peer-review.

Prerequisites

COMP329 and COMP429, or equivalent. You should be able to train and evaluate deep learning models, engineer AI systems, and read technical material independently. Mathematical maturity matters more here than in any previous course.

Fewer units, deeper work

COMP529 has 130 tutorials against COMP429's 197. This is deliberate: the remaining time is spent conducting research, not consuming content.

Course details

Level
Undergraduate · Honours / research stream
Credits
15 credits, one semester
Delivery
1 seminar and 1 research supervision session per week, plus independent work
Language & tools
Python, PyTorch, experiment tracking, LaTeX, reference management
Assessment
5 research assignments, 16 quizzes, proposal defence, major research project and symposium presentation
Literature review Reproducibility Experiment design Scholarly writing
Course structure at a glance
Module Research focus Units Tutorials
1 · Foundations of AI Research Research paradigms, scientific method and literature analysis 2 16
2 · Foundation Models & Emerging Architectures Scaling laws, emergence and reasoning architectures 2 16
3 · Neuro-Symbolic and Causal AI Hybrid intelligence, causal inference and counterfactual reasoning 2 17
4 · Scientific Machine Learning Physics-informed models and AI-driven knowledge discovery 2 16
5 · Distributed Intelligence Multi-agent systems and privacy-preserving collaborative learning 2 16
6 · Embodied and Autonomous Intelligence Embodied cognition, robotics and autonomous systems 2 16
7 · Trustworthy Artificial General Intelligence Alignment, safety engineering and governance of advanced AI 2 17
8 · Future Directions in Artificial Intelligence AGI concepts and the open research frontier 2 16
Total Eight modules · sixteen units 16 130

Curriculum position

Where COMP529 fits

COMP529 is the final course in the sequence. It assumes the full ladder beneath it and prepares you for postgraduate research.

  1. COMP326 · Data Science Foundations
  2. COMP327 · Generative AI
  3. COMP328 · Machine Learning
  4. COMP329 · Advanced ML & Deep Learning
  5. COMP429 · AI Systems Engineering
  6. COMP529 · Research Topics in AI

The full pathway

Six levels from data literacy to AI research

A student completing the full sequence progresses through six distinct levels of capability, each building on the last.

Capability ladder across the AI curriculum
Level Capability Course
1 Data literacy COMP326 · Data Science Foundations
2 AI literacy COMP327 · Responsible and Effective Use of Generative AI
3 Machine learning practitioner COMP328 · Fundamentals of Machine Learning
4 Advanced AI developer COMP329 · Advanced Machine Learning and Deep Learning
5 AI systems engineer COMP429 · AI Systems Engineering and Intelligent Applications
6 AI researcher COMP529 · Research Topics in Artificial Intelligence

Learning outcomes

What you will be able to do

On successful completion of COMP529 you will be able to operate as an independent researcher in artificial intelligence.

Research literacy

  • Read and critique AI research literature.
  • Conduct systematic literature reviews.
  • Identify genuine research gaps.

Experimental design

  • Formulate precise, testable research questions.
  • Design rigorous, well-controlled experiments.
  • Evaluate AI methods scientifically and fairly.

Advanced AI knowledge

  • Explain frontier AI topics and open problems.
  • Compare competing research approaches.
  • Analyse the limitations of current methods.

Scholarly communication

  • Write research reports in conference format.
  • Present scientific findings to a technical audience.
  • Defend research conclusions under questioning.

Innovation

  • Reproduce published research results.
  • Extend existing methods in a principled way.
  • Propose and justify novel solutions.

Research practice

  • Work reproducibly with versioned code and tracked experiments.
  • Peer-review the work of fellow researchers constructively.
  • Apply research ethics and open-science principles.

Curriculum

Eight modules, sixteen units

Each unit surveys a research area: its foundational theory, its seminal papers, its recent developments, and the open challenges that remain.

Research tracks

Choose your research stream

You select one stream for your major research project and pursue it across the second half of the semester, supported by supervision sessions.

Stream A

Foundation Models

Reasoning abilities, prompt optimisation, agentic architectures, long-context reasoning.

 Stream B

Scientific AI

Bioinformatics, computational biology, digital twins, scientific discovery.

 Stream C

Causal AI

Causal discovery, counterfactual analysis, decision intelligence.

 Stream D

Multi-Agent Systems

Coordination, emergent behaviour, collaborative reasoning.

 Stream E

AI Safety

Alignment, interpretability, robustness, governance.

How it works

A research rhythm, not a lecture rhythm

Each unit follows the same four-part structure, designed to move you from consumer of research to producer of research.

  1. Read the literature

    Survey the foundational theory, the seminal papers and the most recent developments in the area.

  2. Critique the method

    Interrogate assumptions, evaluation choices and claims. Ask what the paper does not show.

  3. Reproduce or extend

    Re-implement a result, or design a small experiment that probes an open question.

  4. Discuss and debate

    Bring your findings to the seminar, defend your reasoning and review your peers' work.

Assessment

How your work is assessed

Assessment is overwhelmingly research-based. The five assignments build the skills you need for the major project, which carries the largest weighting.

40%

Major research project

An original investigation culminating in a journal-style paper and symposium presentation.

30%

Five research assignments

Paper analysis, literature review, reproducibility study, proposal and pilot investigation.

10%

Proposal defence

Oral defence of your research question, method and evaluation plan.

10%

Symposium presentation

Conference-format talk with questions from faculty and peer review.

10%

Unit quizzes

Sixteen short online quizzes, one per unit, with unlimited attempts.

Weightings are indicative and confirmed in the course syllabus.

  • Assignment 1 Research Paper Analysis — critique a recent AI paper Brief
  • Assignment 2 Literature Review — structured review with gap analysis Brief
  • Assignment 3 Reproducibility Study — replicate a published experiment Brief
  • Assignment 4 Research Proposal — a publishable-quality proposal Brief
  • Assignment 5 Pilot Investigation — preliminary experiments and analysis Brief

Major research project

An original contribution to the field

The project runs across the whole semester and carries 40% of your grade. You work through sixteen phases, from topic selection to final presentation, under regular supervision.

What you submit

  • Research question and gap analysis
  • Literature review and positioning
  • Method design and experimental setup
  • Reproducible code and collected data
  • Baseline comparison and proposed method
  • Evaluation, statistical analysis and discussion
  • Journal-style paper (6,000–10,000 words)
  • Research poster and symposium presentation

Publication-oriented

The final paper is written in IEEE or ACM format and reviewed by faculty acting as a program committee. Strong submissions are encouraged to submit to undergraduate research venues.

Reproducibility is assessed

Your code, environment, seeds and data handling must allow an independent reader to reproduce your headline result. A finding that cannot be reproduced does not count as a finding.

  1. Topic SelectionChoose a research stream and a tractable area within it.
  2. Research QuestionFormulate a precise, answerable and worthwhile question.
  3. Literature ReviewMap what is known and how it has been established.
  4. Gap IdentificationArticulate exactly what is missing or unresolved.
  5. Hypothesis FormulationState a testable claim that your experiments can falsify.
  6. Method DesignDesign an approach that could genuinely answer the question.
  7. Experimental SetupSpecify datasets, metrics, controls and compute budget.
  8. Data CollectionGather, document and prepare the data you need.
  9. Baseline ModelsEstablish what existing methods already achieve.
  10. Proposed MethodImplement and train your approach.
  11. EvaluationMeasure performance rigorously against the baselines.
  12. Statistical AnalysisTest whether the difference is meaningful, not just present.
  13. DiscussionInterpret the results honestly, including negative findings.
  14. Future WorkIdentify what you would do next and what remains open.
  15. Research PaperWrite up the work to conference standard.
  16. PresentationPresent and defend your research at the symposium.

Scholarly communication

The capstone research symposium

The semester concludes with a mini-conference. You present your work, defend your methodology, answer questions and review the work of your peers.

What you produce

  • Research poster — conference format, presented in a poster session.
  • Oral presentation — 15–20 minutes, followed by questions.
  • Journal-style paper — 6,000–10,000 words in IEEE or ACM format.

How the symposium runs

  • Students present papers, defend methodology, answer questions and review peers.
  • Faculty act as program committee, reviewers and session chairs.
  • Outcome a reviewed paper, a presented talk and structured peer feedback on your research.

Questions

Frequently asked questions

Do I need to have a research topic before the course starts?

No. The first weeks are dedicated to research literacy and literature mapping, and you choose your stream after exploring the options. By the time you commit to a topic you will have the tools to judge whether it is tractable.

How much of the course is writing?

Substantially more than in any previous course. Between the literature review, the proposal and the final paper you will produce well over 10,000 words of formal academic writing. Clear, precise prose is a core assessed skill.

What if my experiments do not produce the expected results?

That is a normal and acceptable outcome in research. A well-designed experiment that falsifies its hypothesis, reported honestly and analysed carefully, can be a stronger project than a marginal positive result that was not properly controlled.

Will I have compute resources for my experiments?

Yes. Research infrastructure is provided, including shared GPU access and curated datasets. You are expected to design experiments that fit within realistic resource limits — which is itself a research skill.

Can my project be published?

Strong projects are encouraged to submit to undergraduate research venues, workshops or student tracks at conferences. Faculty supervision continues to support publication after the course ends where the work merits it.

How is this different from COMP429?

COMP429 teaches you to engineer systems from existing methods — the emphasis is on integration, deployment and governance. COMP529 teaches you to create new methods — the emphasis is on literature, experimental design, rigorous evaluation and scholarly writing.

What support do I get during the project?

Weekly supervision sessions, seminar discussion, templates for literature reviews, proposals and papers, and peer review from fellow students. You are expected to drive the work; the support is there to keep it rigorous.

Ready to contribute to the field?

Begin with Module 1, Unit 1 — what distinguishes AI research from AI practice, and why reproducibility is the discipline's central problem.

Start Unit 1 Download syllabus
👤