ST-ReP pre-trains a compact spatial-temporal encoder by jointly reconstructing current series and predicting future values with multi-scale losses, and it reports better accuracy and memory footprint than self-supervised baselines on six forecasting datasets.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
ST-ReP: Learning Predictive Representations Efficiently for Spatial-Temporal Forecasting
ST-ReP pre-trains a compact spatial-temporal encoder by jointly reconstructing current series and predicting future values with multi-scale losses, and it reports better accuracy and memory footprint than self-supervised baselines on six forecasting datasets.