Honours and Minor Thesis projects

This project aims to develop techniques that enable users to find relevant audio content by inputting textual queries. This process leverages machine learning models, particularly natural language processing and audio signal processing, to bridge the gap between text and audio. When a user submits a query, the system analyses the text to understand its intent and context.

This project involves the automated generation of textual descriptions for audio content, such as spoken language, sound events, or music. This process typically employs deep learning techniques, such as recurrent neural networks, transformer models, and so on, to analyse audio signals and generate coherent captions. By training on large datasets that include both audio recordings and corresponding textual descriptions, these models learn to recognize patterns and contextual meanings within the audio.

Adaptively smoothing one-dimensional signals remains an important problem, with applications in time series analysis, additive modelling and forecasting. The trend filter provides an novel class of adaptive smoothers; however, it is usually implemented in a frequentist framework using tools like the lasso and cross-validation. Bayesian implementations tend to rely on posterior sampling and as such do not provide simple, sparse point-estimates of the underlying curve.

Learning appropriate prior distributions from replications of experiments is a important problem in the space of hierarchical and empirical Bayes. In this problem, we exploit the fact that we have multiple repeats of similar experiments and pool these to learn an appropriate prior distribution for the unknown parameters of this set of problems. Standard solutions to this type of problem tend to be of mixed Bayesian and non-Bayesian form, and are somewhat ad-hoc in nature.

NOTE THIS IS A SUMMER INTERNSHIP PROJECT FOR SUMMER 2024 (Nov 2024-Feb 2025).

There may be an opportunity to work on this project as an honours/minor thesis project in S1 2025. Plans for this will evolve over the summer. 

The spectral density of a time series (a series of time ordered data points -- for example, daily rainfall in the Amazon or the monthly stocks of fish in the Pacific) gives substantial information about the periodic patterns hidden in the data. Learning a good model of the spectral density is usually done through parametric methods like autoregressive moving average processes [1] because non-parametric methods struggle to deal with the interesting “non-smooth” nature of spectral densities. This project aims to apply a powerful and new non-parametric smoothing technique to this problem.

Feedback is crucial to learning success; yet, higher education continues to struggle with effective feedback processes. It is important to recognise that feedback as a process requires both teachers and students to take active roles and work as partners. However, one challenge to facilitate a two-way process of feedback is the difficulty to track feedback impact on learning, particularly how students interact with feedback.

State-of- the-art Large Language Models hallucinate between 69-88% of responses to legal queries; hallucinate in at least 75% of responses regarding a court's main ruling; and reinforce incorrect legal assumptions and beliefs (Source: first reference under references section below). While there is excitement about LLMs facilitating access to justice by offering a low-cost method for the public to obtain legal advice, their limitations could exacerbate rather than mitigate the problem of access to justice.

Verifiable Dynamic Searchable Symmetric Encryption (VDSSE) enables users to securely outsource databases (document sets) to cloud servers and perform searches and updates. The verifiability property prevents users from accepting incorrect search results returned by a malicious server. However, the community currently only focuses on preventing malicious behavior from the server but ignores incorrect updates from the client, which are very likely to happen in multi-user settings. Indeed most existing VDSSE schemes are not sufficient to tolerate incorrect updates from users. For instance,…

LLMs have enabled us to break the performance ceiling in several tasks. Language Agents took this further by connecting language models with tools, environments and enriched them with memory and feedback.