Spatiotemporal Causal Disentanglement
Spatiotemporal Data Analysis; Unsupervised Learning; VAE; Gaussian Process; Graph Factorization
Reseach Project:
Study unsupervised Multimodal Spatiotemporal (MST) data analysis, especially focusing on two tasks:
- Disentangle the underlying dynamic systems of partially observed spatiotemporal interactions between multiple subjects
- Predict missing data or future data based on learned disentangled representations.
Proposed Method:
A causal disentanglement framework based on graph-factorized Gaussian Process VAE.
