This paper introduces a temporal correlation volatility metric, shows that graph and transformer forecasters fail when it is high, and proposes a GNN layer with path-based and static/dynamic separated propagation that reports large gains.
In: International Conference on Artificial Intelligence and Statistics, AISTATS
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
When GNNs Fail: Quantifying and Overcoming Temporal Correlation Volatility in Time Series
This paper introduces a temporal correlation volatility metric, shows that graph and transformer forecasters fail when it is high, and proposes a GNN layer with path-based and static/dynamic separated propagation that reports large gains.