Under the Hood of Tabular Data Generation Models: the Strong Impact of Hyperparameter Tuning - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail (Preprint/Prepublication) Année : 2024

Under the Hood of Tabular Data Generation Models: the Strong Impact of Hyperparameter Tuning

Résumé

We investigate the impact of dataset-specific hyperparameter, feature encoding, and architecture tuning on five recent model families for tabular data generation through an extensive benchmark on 16 datasets. This study addresses the practical need for a unified evaluation of models that fully considers hyperparameter optimization. Additionally, we propose a reduced search space for each model that allows for quick optimization, achieving nearly equivalent performance at a significantly lower cost. Our benchmark demonstrates that, for most models, large-scale dataset-specific tuning substantially improves performance compared to the original configurations. Furthermore, we confirm that diffusion-based models generally outperform other models on tabular data. However, this advantage is not significant when the entire tuning and training process is restricted to the same GPU budget for all models.
Fichier principal
Vignette du fichier
main_all.pdf (1.25 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04612244 , version 1 (17-06-2024)

Licence

Identifiants

  • HAL Id : hal-04612244 , version 1

Citer

G. Charbel N. Kindji, Lina Maria Rojas-Barahona, Elisa Fromont, Tanguy Urvoy. Under the Hood of Tabular Data Generation Models: the Strong Impact of Hyperparameter Tuning. 2024. ⟨hal-04612244⟩
16 Consultations
0 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More