Research map / Time series / Causal and generative time series
Spatiotemporal Observer Design for Predictive Learning of High-Dimensional Data
Tongyi Liang, Han‐Xiong Li
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 ST-Observer in place of SimVP | ST-Observer · TaxiBJ | avg MSE: 0.414 -> 0.406 | Improved |
| use ST-Observer in place of SimVP | ST-Observer · Moving MNIST | MSE: 23.8 -> 21.2 | Improved |
| use ST-Observer on CIKM | ST-Observer · CIKM | MAE: 5.90 -> 5.66 | Improved |
| ST-Observer vs ConvLSTM at high threshold | ST-Observer · CIKM HSS at 40 dBZ | HSS: 0.1470 -> 0.1588 | Mixed |
| add sigmoid constraint on A (ablation) | ST-Observer · TaxiBJ | MSE: 0.428 -> 0.423 | Improved |
| use Kaiming uniform initialization for A (ablation) | ST-Observer · TaxiBJ | MSE: 0.421 -> 0.406 | Improved |
| add B projection (Inception) instead of none (ablation) | ST-Observer · TaxiBJ | MSE: 0.419 -> 0.410 | Improved |
| use Inception instead of 3*3 conv for B (ablation) | ST-Observer · TaxiBJ | MSE: 0.423 -> 0.410 | Improved |
| add L1 term to loss (weight 1 vs 0) (ablation) | ST-Observer · TaxiBJ | MSE: 0.4297 -> 0.4085 | Improved |
| add dynamic regularization on z (lambda 1 vs 0) (ablation) | ST-Observer · TaxiBJ | MSE: 0.4085 -> 0.4231 | Hurt |
| add dynamic regularization on xi (lambda 0.1) (ablation) | ST-Observer · TaxiBJ | MSE: 0.4085 -> 0.4048 | Improved |
| combine dynamic regularizations 0.1/1/1 (ablation) | ST-Observer · TaxiBJ | MSE: 0.4085 -> 0.4055 | Improved |
Closest papers by meaning
- Learning Spatiotemporal Dynamical Systems from Point Process Observations (ICLR 2025 · similarity 0.86)
- FlowMixer: A Depth-Agnostic Neural Architecture for Interpretable Spatiotemporal Forecasting (NeurIPS 2025 · similarity 0.86)
- Graph-based Forecasting with Missing Data through Spatiotemporal Downsampling (ICML 2024 · similarity 0.85)
- Wavelet-Driven Spatiotemporal Predictive Learning: Bridging Frequency and Time Variations (AAAI 2024 · similarity 0.85)
- UniSTD: Towards Unified Spatio-Temporal Learning across Diverse Disciplines (CVPR 2025 · similarity 0.84)
- Earthfarsser: Versatile Spatio-Temporal Dynamical Systems Modeling in One Model (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.
