A discounted convex combination of log-likelihood and expected log-likelihood yields exact linear recursions for filtering, prediction, and smoothing in canonical exponential family time series.
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
UNVERDICTED 1representative citing papers
citing papers explorer
-
Exponentially weighted estimands and the exponential family: Filtering, prediction and smoothing
A discounted convex combination of log-likelihood and expected log-likelihood yields exact linear recursions for filtering, prediction, and smoothing in canonical exponential family time series.