Using an EM loop with Langevin sampling, a generative state-space model learns ocean dynamics from incomplete satellite observations and improves SST reconstruction and one-step forecasts over its pretrained baseline.
Title resolution pending
1 Pith paper cite this work, alongside 482 external citations. Polarity classification is still indexing.
1
Pith paper citing it
482
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Incomplete Observations Boost Evolutionary Performance in Ocean Modeling
Using an EM loop with Langevin sampling, a generative state-space model learns ocean dynamics from incomplete satellite observations and improves SST reconstruction and one-step forecasts over its pretrained baseline.