Saved in:
Bibliographic Details
Main Author: De, Tuhin Subhra
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.08946
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911497615048704
author De, Tuhin Subhra
author_facet De, Tuhin Subhra
contents Variational Autoencoders and Generative Adversarial Networks remained the state-of-the-art (SOTA) generative models until 2022. Now they are superseded by diffusion-based models. Efforts to improve traditional models have stagnated as a result. In old-school fashion, we explore image generation with conditional Variational Autoencoders (CVAE) to incorporate desired attributes within the images. VAEs are known to produce blurry images with less diversity; we refer to a method that solves this issue by leveraging the variance of the gaussian decoder as a learnable parameter during training. Previous works on CVAEs assumed that the conditional distribution of the latent space given the labels is equal to the prior distribution, which is not the case in reality. We show that estimating it using Non-Volume Preserving (NVP) transformations results in better image generation than existing methods by reducing the FID by 4% and increasing log likelihood by 7.6% compared to the previous cases.
format Preprint
id arxiv_https___arxiv_org_abs_2511_08946
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Improving Conditional VAE with Non-Volume Preserving transformations
De, Tuhin Subhra
Machine Learning
Variational Autoencoders and Generative Adversarial Networks remained the state-of-the-art (SOTA) generative models until 2022. Now they are superseded by diffusion-based models. Efforts to improve traditional models have stagnated as a result. In old-school fashion, we explore image generation with conditional Variational Autoencoders (CVAE) to incorporate desired attributes within the images. VAEs are known to produce blurry images with less diversity; we refer to a method that solves this issue by leveraging the variance of the gaussian decoder as a learnable parameter during training. Previous works on CVAEs assumed that the conditional distribution of the latent space given the labels is equal to the prior distribution, which is not the case in reality. We show that estimating it using Non-Volume Preserving (NVP) transformations results in better image generation than existing methods by reducing the FID by 4% and increasing log likelihood by 7.6% compared to the previous cases.
title Improving Conditional VAE with Non-Volume Preserving transformations
topic Machine Learning
url https://arxiv.org/abs/2511.08946