FIT3233 · Optimisation and reinforcement learning

Overview

Many artificial intelligence systems must do more than recognise patterns: they must choose actions, optimise outcomes and adapt through interaction with changing environments. This unit develops advanced capability in optimisation and reinforcement learning, building on prior deep learning knowledge to explore how intelligent systems can learn to make decisions under uncertainty. You will study optimisation methods used in artificial intelligence and machine learning, including objective functions, constraints, gradient-based optimisation, stochastic optimisation, hyperparameter optimisation and trade-offs between exploration, exploitation, performance and efficiency. You will also examine reinforcement learning foundations, including agents, environments, rewards, policies, value functions, temporal-difference learning, policy gradients, deep reinforcement learning and evaluation of learned behaviour. You will formulate optimisation and sequential decision-making problems, implement and evaluate reinforcement learning approaches, and analyse the behaviour, limitations and performance of artificial intelligence systems. The unit emphasises informed judgement in selecting methods, designing reward structures, interpreting experimental results and communicating the strengths, risks and limitations of optimisation and reinforcement learning solutions.

Offerings

The Handbook publishes no offerings for this unit.

Assessment

The Handbook publishes no assessment items for this unit yet.

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

Requisites

Learning outcomes

1. Use optimisation and reinforcement learning methods, tools and workflows to develop, train, evaluate and refine artificial intelligence systems;
2. Apply artificial intelligence engineering approaches to design agents, policies, reward structures and adaptive decision-making systems for sequential and interactive environments;
3. Formulate and analyse complex optimisation and reinforcement learning problems by defining objectives, constraints, states, actions, rewards, uncertainty and performance trade-offs;
4. Develop and evaluate machine learning models and policies using optimisation, value-based learning, policy-based learning, function approximation and deep reinforcement learning techniques;
5. Communicate optimisation and reinforcement learning formulations, experimental results, model behaviour, trade-offs and limitations using appropriate technical language, evidence and visualisations.

Workload

No workload detail published.

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