SA-SGLD adapts SGLD stepsizes via gradient-norm-based time rescaling to sample BNN posteriors more accurately than standard SGLD on toy examples and image classification tasks without introducing bias.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
verdicts
UNVERDICTED 2representative citing papers
Deterministic envelopes decouple stochastic-gradient noise from taming in SGLD, splitting stationary error into oracle-dependent bias and deterministic stabilization error, with a hybrid soft-hard design for far tails.
citing papers explorer
-
Adaptive Stepsizing for Stochastic Gradient Langevin Dynamics in Bayesian Neural Networks
SA-SGLD adapts SGLD stepsizes via gradient-norm-based time rescaling to sample BNN posteriors more accurately than standard SGLD on toy examples and image classification tasks without introducing bias.
-
Deterministic Envelopes for Tamed SGLD: Decoupling Stochastic-Gradient Noise and Localizing Taming
Deterministic envelopes decouple stochastic-gradient noise from taming in SGLD, splitting stationary error into oracle-dependent bias and deterministic stabilization error, with a hybrid soft-hard design for far tails.