Improving Conditional VAE with approximation using Normalizing Flows

Tuhin Subhra De · 2025

Read Improving Conditional VAE with approximation using Normalizing Flows by Tuhin Subhra De (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-08946.txt.

Description

Tuhin Subhra De revisits an unfashionable model, the conditional variational autoencoder, and fixes a premise baked into it. Variational autoencoders fell behind diffusion models years ago and are known for blurry, low-diversity images, but they remain simple and steerable when a user wants pictures with chosen traits.

Earlier conditional versions assumed the latent distribution given the labels equals the prior. Real data violates that. De estimates that conditional spread with non-volume-preserving maps, a flexible family of invertible changes, and keeps a trick that treats the decoder's variance as a learnable parameter. The result cuts the FID quality score by about four percent and raises log likelihood by about seven point six percent over previous approaches. The paper is a small, old-school argument that classical generative models still have room to grow.

Catalog record

CREATORTuhin Subhra De
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE24 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-08946

Browse the full library · About the Tower of Babel