OmniST paper proposes topology-invariant spatiotemporal forecasting with coverage-guaranteed uncertainty at unseen sensors
An arXiv paper introduces OmniST, a spatiotemporal forecasting approach that removes the fixed adjacency-matrix constraint tying graph convolutional networks to a specific sensor count, enabling predictions with guaranteed-coverage uncertainty at unseen sensor locations.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-04 05:00 · arXiv stat.ML
OmniST: Topology-Invariant Spatiotemporal Forecasting with Guaranteed-Coverage Uncertainty at Unseen Sensor Locations