Meta-learning approach for variational autoencoder hyperparameter tuning

Synthetic data generation is a promising alternative to traditional data anonymization, with Variational Autoencoders (VAEs) excelling at generating high-quality synthetic tabular datasets. However, VAE hyperparameter selection is often computationally expensive or suboptimal. We propose a meta-lear...

Full description

Saved in:
Bibliographic Details
Main Authors: Michele Berti, Matheus Camilo da Silva, Sebastiano Saccani, Sylvio Barbon Junior
Format: Article
Language:English
Published: Graz University of Technology 2025-06-01
Series:Journal of Universal Computer Science
Subjects:
Online Access:https://lib.jucs.org/article/124087/download/pdf/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849431591507460096
author Michele Berti
Matheus Camilo da Silva
Sebastiano Saccani
Sylvio Barbon Junior
author_facet Michele Berti
Matheus Camilo da Silva
Sebastiano Saccani
Sylvio Barbon Junior
author_sort Michele Berti
collection DOAJ
description Synthetic data generation is a promising alternative to traditional data anonymization, with Variational Autoencoders (VAEs) excelling at generating high-quality synthetic tabular datasets. However, VAE hyperparameter selection is often computationally expensive or suboptimal. We propose a meta-learning (MtL) method for hyperparameter recommendation, which achieves competitive performance to state-of-the-art Bayesian Optimization (BO) with median AUC values of 0.660 ± 0.038 (MtL) and 0.650 ± 0.041 (BO), showing no statistically significant difference. Notably, our approach reduces configuration time to under three minutes, compared to BO’s multi-hour requirement, while also enabling incremental improvements through new data integration. This combination of efficiency, adaptability, and performance establishes MtL as a practical solution for hyperparameter tuning in synthetic data generation.
format Article
id doaj-art-f646b0c98dba4398b1d9d571fb1acc61
institution Kabale University
issn 0948-6968
language English
publishDate 2025-06-01
publisher Graz University of Technology
record_format Article
series Journal of Universal Computer Science
spelling doaj-art-f646b0c98dba4398b1d9d571fb1acc612025-08-20T03:27:36ZengGraz University of TechnologyJournal of Universal Computer Science0948-69682025-06-0131766868210.3897/jucs.124087124087Meta-learning approach for variational autoencoder hyperparameter tuningMichele Berti0Matheus Camilo da Silva1Sebastiano Saccani2Sylvio Barbon Junior3University of TriesteUniversity of TriesteAindo SrlUniversity of TriesteSynthetic data generation is a promising alternative to traditional data anonymization, with Variational Autoencoders (VAEs) excelling at generating high-quality synthetic tabular datasets. However, VAE hyperparameter selection is often computationally expensive or suboptimal. We propose a meta-learning (MtL) method for hyperparameter recommendation, which achieves competitive performance to state-of-the-art Bayesian Optimization (BO) with median AUC values of 0.660 ± 0.038 (MtL) and 0.650 ± 0.041 (BO), showing no statistically significant difference. Notably, our approach reduces configuration time to under three minutes, compared to BO’s multi-hour requirement, while also enabling incremental improvements through new data integration. This combination of efficiency, adaptability, and performance establishes MtL as a practical solution for hyperparameter tuning in synthetic data generation.https://lib.jucs.org/article/124087/download/pdf/HPOMeta-learningVariational AutoencodersVAE
spellingShingle Michele Berti
Matheus Camilo da Silva
Sebastiano Saccani
Sylvio Barbon Junior
Meta-learning approach for variational autoencoder hyperparameter tuning
Journal of Universal Computer Science
HPO
Meta-learning
Variational Autoencoders
VAE
title Meta-learning approach for variational autoencoder hyperparameter tuning
title_full Meta-learning approach for variational autoencoder hyperparameter tuning
title_fullStr Meta-learning approach for variational autoencoder hyperparameter tuning
title_full_unstemmed Meta-learning approach for variational autoencoder hyperparameter tuning
title_short Meta-learning approach for variational autoencoder hyperparameter tuning
title_sort meta learning approach for variational autoencoder hyperparameter tuning
topic HPO
Meta-learning
Variational Autoencoders
VAE
url https://lib.jucs.org/article/124087/download/pdf/
work_keys_str_mv AT micheleberti metalearningapproachforvariationalautoencoderhyperparametertuning
AT matheuscamilodasilva metalearningapproachforvariationalautoencoderhyperparametertuning
AT sebastianosaccani metalearningapproachforvariationalautoencoderhyperparametertuning
AT sylviobarbonjunior metalearningapproachforvariationalautoencoderhyperparametertuning