Understanding behaviour and predicting events is a core machine learning task, and has many applications in areas including computer vision (to detect or prediction actions in video) and signal processing (to detect events in medical signals).
While a large body of research exists exploring these tasks, a number of common challenges persist including:
- capturing variations in how behaviours or events appear across different subjects, such that predictions can be accurately made for previously unseen subjects
- modelling and incorporating long-term relationships, such as previously observed behaviours and temporal dependencies between behaviours
- learning from limited (or partially labelled) data
- effectively incorporating data from multiple sources, where sources may not have a common sampling rate, or always be available.
This project will develop new methods to predict behaviours/events in signal data. This will involve the development of new machine learning methods, and evaluating these on public datasets. Research activities will include:
- research and development of novel machine learning methods
- experimental design
- writing up, publishing and presenting research outcomes.
This project will build on an existing body of research conducted by the supervisory team.
This project will develop novel machine learning methods for classification and prediction of actions, behaviours and events from signal data (i.e. video, medial signals, audio).
Skills and experience
Strong programming experience (preferably Python) and some machine learning experience.
Contact the supervisor for more information.