A variational autoencoder objective based on coupled free energy with escort sampling claims robust heavy-tailed training, but the headline 3% FID gain is absent from the paper's own table.
Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
REJECT 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Variational Inference Optimized Using the Curved Geometry of Coupled Free Energy
A variational autoencoder objective based on coupled free energy with escort sampling claims robust heavy-tailed training, but the headline 3% FID gain is absent from the paper's own table.