Research projects in Information Technology

Markov Decision Processes (MDPs) are frameworks used to model decision-making in situations where outcomes are partly random and partly under the control of a decision maker. While small MDPs are inherently interpretable for people, MDPs with thousands of states are difficult to understand by humans. The K-MDP problem is the problem of finding the best MDP with, at most, K states by leveraging state abstraction approaches to aggregate states into sub-groups.

Turning Point is a renowned addiction treatment and research centre specialising in the prevention, treatment, and support services for individuals affected by substance use disorders, gambling addiction, and mental health issues. Turning Point operates a network of 26 helplines across the country, ensuring accessible and immediate support for individuals in need. These helplines serve as a vital resource for individuals seeking assistance, information, and guidance related to addiction and mental health concerns.

Classical propositional logic (CPL) captures our basic understanding of the linguistic connectives “and”, “or” and “not”. It also provides a very good basis for digital circuits. But it does not account for more sophisticated linguistic notions such as “always”, “possibly”, “believed” or “knows”. Philosophers therefore invented many different non-classical logics which extend CPL with further operators for these notions.

Classical propositional logic (CPL) captures our basic understanding of the linguistic connectives “and”, “or” and “not”. It also provides a very good basis for digital circuits. But it does not account for more sophisticated linguistic notions such as “always”, “possibly”, “believed” or “knows”. Philosophers therefore invented many different non-classical logics which extend CPL with further operators for these notions.

Neuro-symbolic AI combines the strengths of neural and symbolic methods to efficiently learn and reason over models of the world. Typically, many of the assurances that can be provided by such systems are limited to the learning errors due to the neural component. In this Ph.D. project, you will be exploring the use of Lipschitz Continuous Neural Networks to learn Lipschitz-bounded neural models of the world to create a Neuro-symbolic AI that can provide assurance guarantees.

Neuro-symbolic AI combines the strengths of neural and symbolic methods to efficiently learn and reason over models of the world. Typically, many of the assurances that can be provided by such systems are limited to the learning errors due to the neural component. In this Ph.D. project, you will be exploring the use of Physics-Informed Neural Networks to encode the symbolic knowledge into the learned neural models of the world to create a Neuro-symbolic AI that can provide assurance guarantees.

Planning is the reasoning side of acting in Artificial Intelligence. Planning automates the selection and the organisation of actions to reach desired states of the world as best as possible. For many real-world planning problems however, it is difficult to obtain the full model of the world that captures its complex dynamics. Fortunately, the unknown parts model can be accurately approximated as neuro-symbolic (deep) neural networks which then can be used in planning.

This project will result in a much fuller understanding of the state of the Birrarung than is currently possible, as well as qualitative and quantitative results to model different interventions and their effect on swimmability.

The project will build tools and techniques to understand and decide on effective interventions to improve the Birrarung’s swimmability.

In many branches of science (e.g., Artificial Intelligence, Engineering etc.), the modelling of the problem is done through the use of functions (e.g., f(x) = y). On a very high-level, we can think of Machine Learning as the problem of approximating function f from the pair of measurements (x,y), and Optimization as the problem of finding the value of input x that maximizes the output y given function f.

Automating code generation, SQL query formulation, and data preprocessing pipelines is a crucial step toward intelligent and efficient software development. This project aims to leverage large language models (LLMs) to address these challenges by developing a comprehensive framework that seamlessly integrates LLM capabilities for generating accurate and optimised code, constructing complex SQL queries, and automating data transformations.