The scholarship includes AUD $10,000 living allowance per year and tuition fee waiver. The student will be based in Vietnam but does not need to be a Vietnam national.
The candidate will pick one of the two projects to write their proposal on:
- Project title: Federated learning using unlabelled sensor and user-generated data
As a machine learning framework mostly suited for large-scale labelled data, federated learning could be a key technology to bring fragmented sensitive data, such as live sensor monitoring data, thus enables training a shared global model to improve the accuracy of monitoring services while keeping the sensitive data where they originate. In this project, we investigate how to apply the Federated Learning model to enhance the accuracy and efficiency of the constant monitoring and diagnosis process in different domains, such as within smart cities domain, or in medical domain.
Melbourne contact (senior supervisor): Professor Flora Salim
VN contact (associate supervisor): Dr. Minh Ngoc Dinh - Project title: Explainable AI for personalised models
This project aims to investigate techniques to generate actionable explanations, for a range of problems and data types and modality, from large-scale unstructured data, to highly varied sensor data and multimodal data. Existing surrogate models and techniques for explaining machine learning predictions (output) or models have recently been proposed, such as LIME, but they are only limited to local explanations, based on the prediction outputs, and are not robust to the variations in sample and hyperparameter settings. Recent models such as SHAP provides global explanations, however the computation is expensive and fully factorised explanations can be fully intractable. This project seeks to advance effective and efficient surrogate models for local and global model explanations, applicable to personalised prediction and recommender systems.
Melbourne contact (senior supervisor): Professor Flora Salim
VN contact (associate supervisor): Dr Nhi Vo Ngoc Yen