Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Learning Temporal Evolution of Spatial Dependence with Generalized Spatiotemporal Gaussian Process Models
Shiwei Lan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use qKron-sum (model II) instead of stationary or nonstationary parametric covariances | Spatiotemporal GP · Simulated process fit, K=100 | TESD MSE: 0.111 -> 5.14e-3 | Improved |
| Use qKron-sum (model II) instead of the best baseline | Spatiotemporal GP · Simulated process fit, K=1000 | TESD MSE: 9.52e-2 -> 3.85e-4 | Improved |
| Use quasi Kronecker sum structure instead of quasi Kronecker product structure (ablation) | Generalized STGP · Simulated process fit, K=1000 | TESD MSE: 0.189 -> 3.85e-4 | Improved |
| Use qKron-sum instead of qKron-prod for predicting TESD to future (ablation) | Generalized STGP · Simulated prediction, K=100 | MSPE TESD to future: 0.109 -> 1.04e-2 | Improved |
| Use qKron-sum instead of qKron-prod for predicting TESD to neighbor (ablation) | Generalized STGP · Simulated prediction, K=1000 | MSPE TESD to neighbor: 0.187 -> 1.48e-3 | Improved |
| Use qKron-sum instead of qKron-prod for mean prediction (ablation) | Generalized STGP · Simulated prediction, K=1000 | MSPE mean: 4.18e-4 -> 1.6e-3 | Hurt |
| Increase trials from K=100 to K=1000 | qKron-sum (II) · Simulated process fit | TESD MSE: 5.14e-3 -> 3.85e-4 | Improved |
| Use qKron-sum instead of dense parametric models on real brain images | Spatiotemporal GP · ADNI PET AD group (40x40) | weighted MSE: 57731.13 -> 4545.54 | Improved |
| Use qKron-sum instead of stat-nonsep for compute time | Spatiotemporal GP · ADNI PET AD group (40x40) | time (seconds): 270202.49 -> 5087.06 | Improved |
| Use qKron-sum instead of stat-sep on mean fit | Spatiotemporal GP · Simulated process fit, K=100 | mean function MSE: 2.17e-2 -> 2.06e-2 | No clear effect |
| Use smoother, more informative prior for spatial kernel (kappa>0 instead of 0) (ablation) | qKron-sum (II) · ADNI PET images | TESD quality | Hurt |
Closest papers by meaning
- Modeling Latent Neural Dynamics with Gaussian Process Switching Linear Dynamical Systems (NeurIPS 2024 · similarity 0.86)
- Causal Deciphering and Inpainting in Spatio-Temporal Dynamics via Diffusion Model (NeurIPS 2024 · similarity 0.85)
- Implicit Gaussian process representation of vector fields over arbitrary latent manifolds (ICLR 2024 · similarity 0.84)
- Discovering Latent Causal Graphs from Spatiotemporal Data (ICML 2025 · similarity 0.84)
- Wavelet Canonical Coherence for Nonstationary Signals (NeurIPS 2025 · similarity 0.84)
- Seahorse: A Unified Benchmarking Framework for Spatiotemporal Event Modeling (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
