Generalized variational autoencoders for learning disentangled representation

Open Webpage

Hazal Mogultay Ozcan, Sinan Kalkan, Fatos T. Yarman-Vural

Published: 01 Jan 2025, Last Modified: 04 Nov 2025Neurocomputing 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•Generalizes the loss function of vanilla VAE to cover its extensions.•Introduces mdL-VAE which learns independent weights per dimension.•Shows that learned weights can be used to discriminate disentangled dimensions.