Interpretable Spatio-Temporal Representation Learning

The growth of IoT and spatio-temporal sensor data provides both challenges and opportunities to designing more efficient transport systems and mobility services. Despite the success that deep learning models have gained in spatio-temporal data modelling, neural networks are always considered as a black box. In fact, interpretability is critical for decision-making. A trustworthy model requires interpretability as well as accurate forecasting / classification. In this project, we will investigate interpretable representations for transport and mobility. Our aim is to learn representations that provide semantic-rich information and thus makes the model more transparent and trustworthy. The project is funded by the ARC Centre of Excellence for Automated Decision-Making and Society (ADM+S). There are several possible directions:

1. Spatio-temporal disentanglement and fairness
2. Causal inference on spatiotemporal graphs
3. Spatiotemporal interpretability with knowledge distillation

Scholarship

• $37,684 per annum (2024 rate, indexed)

Eligibility

• Domestic applicants only
• PhD only

How to apply

Email a copy of your CV, transcripts, and publications (if any) to flora.salim@unsw.edu.au.