My research covers both machine learning and music information retrieval (MIR). Broadly speaking, I am interested in
- Unsupervised learning: How can we leverage unlabelled data to improve generalization in MIR tasks?
- Multi-task learning: How can we solve multiple related MIR tasks in one model, to increase parameter efficiency and generalization?
- Bayesian machine learning, which helps us build model priors and transfer knowledge between tasks