Interpretable Generative Modeling Using a Hierarchical Topological VAE - CentraleSupélec Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Interpretable Generative Modeling Using a Hierarchical Topological VAE

Résumé

Generating realistic datasets with fine-grained control over their properties can help overcome challenges linked to the scarcity of data in many domains, such as medical applications. To that end, we extend Variational Autoencoders by using a hierarchical and topological prior consisting of a sequence of Self-Organizing Maps (SOM), which are stacked in the latent space and learned without supervision, jointly with the parameters of the variational autoencoder. We induce a hierarchy between the codes of the SOM sequence, each SOM corresponding to a different hierarchical level and learning increasingly finer-grained representations of the data. Our model combines the power of deep learning with the interpretability of hierarchical and topological clustering and produces competitive results when evaluated on three well-known computer vision benchmarks and a custom medical dataset.
Fichier principal
Vignette du fichier
GHSOM_Paper-1.pdf (746.24 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04573531 , version 1 (17-05-2024)

Identifiants

Citer

Etienne Desticourt, Véronique Letort, Florence d'Alché-Buc. Interpretable Generative Modeling Using a Hierarchical Topological VAE. 2022 International Conference on Computational Science and Computational Intelligence (CSCI), Dec 2022, Las Vegas, United States. pp.1415-1421, ⟨10.1109/CSCI58124.2022.00253⟩. ⟨hal-04573531⟩
0 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Mastodon Facebook X LinkedIn More