Subsampling in SGLD does not asymptotically reduce the total computation needed for accurate posterior sampling in typical exponential-family models.
Minibatch G ibbs sampling on large graphical models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.CO 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
No Free Lunch for Stochastic Gradient Langevin Dynamics
Subsampling in SGLD does not asymptotically reduce the total computation needed for accurate posterior sampling in typical exponential-family models.