Projects

Relational Deep Learning for Reliable Time Series Forecasting at Scale

Graph deep learning methods forecast collections of related time series by accounting for the relationships among them, but they operate at a single spatiotemporal scale, which makes it hard to capture long-range dependencies and multi-scale dynamics. This project designs hierarchical relational forecasting models that process clusters of time series at adaptive spatial and temporal resolutions, together with scalable neural operators and uncertainty quantification methods, with applications in energy analytics and healthcare.