Pith. sign in

REVIEW 4 cited by

Long-Range Transformers for Dynamic Spatiotemporal Forecasting

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2109.12218 v3 pith:Q6JNYTXJ submitted 2021-09-24 cs.LG stat.ML

classification cs.LGstat.ML
keywords forecastingtimerelationshipsspatiotemporalvariablelearninglong-rangemethods
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Multivariate time series forecasting focuses on predicting future values based on historical context. State-of-the-art sequence-to-sequence models rely on neural attention between timesteps, which allows for temporal learning but fails to consider distinct spatial relationships between variables. In contrast, methods based on graph neural networks explicitly model variable relationships. However, these methods often rely on predefined graphs that cannot change over time and perform separate spatial and temporal updates without establishing direct connections between each variable at every timestep. Our work addresses these problems by translating multivariate forecasting into a "spatiotemporal sequence" formulation where each Transformer input token represents the value of a single variable at a given time. Long-Range Transformers can then learn interactions between space, time, and value information jointly along this extended sequence. Our method, which we call Spacetimeformer, achieves competitive results on benchmarks from traffic forecasting to electricity demand and weather prediction while learning spatiotemporal relationships purely from data.

Discussion (0). Sign in to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. VMDNet: Temporal Leakage-Free Variational Mode Decomposition for Electricity Demand Forecasting

    cs.LG 2025-09 conditional novelty 6.0 of 10

    VMDNet applies sample-wise variational mode decomposition with frequency embeddings, per-mode TCN decoders, and a bilevel Stackelberg search for mode count K and bandwidth penalty alpha, achieving best results on stro...

  2. Position: Current Benchmarking Hinders Real Progress in Deep Learning for Time Series Forecasting

    cs.LG 2025-12 conditional novelty 5.0 of 10

    Current benchmarks fail to credit performance to the right design choices; simple, well-configured models match state-of-the-art time series forecasters.

  3. LSDM: LLM-Enhanced Spatio-temporal Diffusion Model for Service-Level Mobile Traffic Prediction

    cs.LG 2025-07 conditional novelty 5.0 of 10

    LSDM predicts next-hour mobile traffic per app category by feeding a diffusion model with satellite imagery, POI counts, and LLM-generated text descriptions, outperforming eight baselines on a single real-world dataset.

  4. Dynamic Modes as Time Representation for Spatiotemporal Forecasting

    cs.LG 2025-06 conditional novelty 4.0 of 10

    A DMD-based time embedding, built from sine and cosine functions at data-derived frequencies, improves long-horizon spatiotemporal forecasting accuracy in most tested settings.

Pith tools