Research map / Time series / Long-term forecasting
AZ-whiteness test: a test for signal uncorrelation on spatio-temporal graphs
Daniele Zambon, Cesare Alippi
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 FCRNN baseline instead of graph-aware forecasters | FCRNN · GPVAR | MAE: 0.321 -> 0.385 | Hurt |
| Use FCRNN baseline instead of graph-aware forecasters | FCRNN · GPVAR | AZ-test spatio-temporal statistic: 0.3 -> 5.0 | Hurt |
| Use only temporal edges (set spatial edge weights to zero) in AZ-test (ablation) | AZ-test · GPVAR FCRNN | test statistic: 8.9 -> -1.8 | Mixed |
| Use only spatial edges versus only temporal edges in AZ-test (ablation) | AZ-test · PemsBay GWNET | test statistic: 7.1 -> 590.7 | Mixed |
| Use only spatial edges versus only temporal edges in AZ-test (ablation) | AZ-test · MetrLA GWNET | test statistic: -6.5 -> 236.5 | Mixed |
| Use spatio-temporal graph rather than temporal-only graph (ablation) | AZ-test · PemsBay GWNET | test statistic: 7.1 -> 422.7 | Improved |
| Subtract the empirical median from residuals before testing | AZ-test · PemsBay and MetrLA | test outcome | No clear effect |
| Increase correlation parameter c in synthetic signals | AZ-test · Synthetic graph signals | rejection rate | Improved |
| Increase number of time steps T in synthetic signals | AZ-test · Synthetic graph signals | power | Improved |
| Change data distribution P (symmetric, asymmetric, bimodal) | AZ-test · Synthetic graph signals at c=0 | false positive rate | No clear effect |
| Replace GWNET with DCRNN | forecasting model · GPVAR | MAE: 0.324 -> 0.328 | No clear effect |
| Use graph-aware model GWNET instead of FCRNN | forecasting model · PemsBay | MAE: 2.016 -> 0.841 | Improved |
Closest papers by meaning
- Designing Specialized Two-Dimensional Graph Spectral Filters for Spatial-Temporal Graph Modeling (AAAI 2025 · similarity 0.84)
- Graph Neural Networks Are More Than Filters: Revisiting and Benchmarking from A Spectral Perspective (ICLR 2025 · similarity 0.84)
- Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective (AAAI 2024 · similarity 0.83)
- Online GNN Evaluation Under Test-time Graph Distribution Shifts (ICLR 2024 · similarity 0.83)
- Rayleigh Quotient Graph Neural Networks for Graph-level Anomaly Detection (ICLR 2024 · similarity 0.83)
- GOODAT: Towards Test-Time Graph Out-of-Distribution Detection (AAAI 2024 · similarity 0.83)
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.
