KAEM adapts the Kolmogorov-Arnold theorem to energy models with univariate latent priors, enabling fast single-pass sampling and interpretable 1D-density priors that outperform other latent models on SVHN and CIFAR10 FID.
34 35 Pre-publication work
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Kolmogorov-Arnold Energy Models: Fast, Interpretable Generative Modeling
KAEM adapts the Kolmogorov-Arnold theorem to energy models with univariate latent priors, enabling fast single-pass sampling and interpretable 1D-density priors that outperform other latent models on SVHN and CIFAR10 FID.