Enregistré dans:
Détails bibliographiques
Auteurs principaux: Spinaci, Marco, Polewczyk, Marek, Hoffart, Johannes, Kohler, Markus C., Thelin, Sam, Klein, Tassilo
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2410.13516
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909353421832192
author Spinaci, Marco
Polewczyk, Marek
Hoffart, Johannes
Kohler, Markus C.
Thelin, Sam
Klein, Tassilo
author_facet Spinaci, Marco
Polewczyk, Marek
Hoffart, Johannes
Kohler, Markus C.
Thelin, Sam
Klein, Tassilo
contents Self-supervised learning on tabular data seeks to apply advances from natural language and image domains to the diverse domain of tables. However, current techniques often struggle with integrating multi-domain data and require data cleaning or specific structural requirements, limiting the scalability of pre-training datasets. We introduce PORTAL (Pretraining One-Row-at-a-Time for All tabLes), a framework that handles various data modalities without the need for cleaning or preprocessing. This simple yet powerful approach can be effectively pre-trained on online-collected datasets and fine-tuned to match state-of-the-art methods on complex classification and regression tasks. This work offers a practical advancement in self-supervised learning for large-scale tabular data.
format Preprint
id arxiv_https___arxiv_org_abs_2410_13516
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle PORTAL: Scalable Tabular Foundation Models via Content-Specific Tokenization
Spinaci, Marco
Polewczyk, Marek
Hoffart, Johannes
Kohler, Markus C.
Thelin, Sam
Klein, Tassilo
Machine Learning
Self-supervised learning on tabular data seeks to apply advances from natural language and image domains to the diverse domain of tables. However, current techniques often struggle with integrating multi-domain data and require data cleaning or specific structural requirements, limiting the scalability of pre-training datasets. We introduce PORTAL (Pretraining One-Row-at-a-Time for All tabLes), a framework that handles various data modalities without the need for cleaning or preprocessing. This simple yet powerful approach can be effectively pre-trained on online-collected datasets and fine-tuned to match state-of-the-art methods on complex classification and regression tasks. This work offers a practical advancement in self-supervised learning for large-scale tabular data.
title PORTAL: Scalable Tabular Foundation Models via Content-Specific Tokenization
topic Machine Learning
url https://arxiv.org/abs/2410.13516