Under a random temporal shift model, the asymptotically optimal ERM weights solve a bias-variance trade-off, and pooling, most-recent, and exponential weighting emerge as special cases.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ME 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Optimal Empirical Risk Minimization under Temporal Distribution Shifts
Under a random temporal shift model, the asymptotically optimal ERM weights solve a bias-variance trade-off, and pooling, most-recent, and exponential weighting emerge as special cases.