Rolling-origin evaluation windows misrepresent the atom mass of intermittent time series, a permutation control isolates coupling contributions at fixed CRPS, and an autoregressive hurdle beats a conditional flow on five of six datasets under run-length statistics.
Non-stationary diffusion for probabilistic time series forecasting,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Evaluating Generative Time-Series Models on Data with Point Masses
Rolling-origin evaluation windows misrepresent the atom mass of intermittent time series, a permutation control isolates coupling contributions at fixed CRPS, and an autoregressive hurdle beats a conditional flow on five of six datasets under run-length statistics.