T-JEPA: Augmentation-Free Self-Supervised Learning for Tabular Data

TL;DR

T-JEPA introduces an augmentation-free self-supervised learning method for tabular data, leveraging latent space predictions to learn meaningful representations without data augmentation, leading to improved downstream task performance.

Contribution

The paper proposes T-JEPA, a novel augmentation-free SSL approach for tabular data using latent space predictions, and introduces regularization tokens for effective training.

Findings

Outperforms traditional models on classification and regression tasks.

Enables some methods to match or surpass Gradient Boosted Decision Trees.

Effectively identifies relevant features without label access.

Abstract

Self-supervision is often used for pre-training to foster performance on a downstream task by constructing meaningful representations of samples. Self-supervised learning (SSL) generally involves generating different views of the same sample and thus requires data augmentations that are challenging to construct for tabular data. This constitutes one of the main challenges of self-supervision for structured data. In the present work, we propose a novel augmentation-free SSL method for tabular data. Our approach, T-JEPA, relies on a Joint Embedding Predictive Architecture (JEPA) and is akin to mask reconstruction in the latent space. It involves predicting the latent representation of one subset of features from the latent representation of a different subset within the same sample, thereby learning rich representations without augmentations. We use our method as a pre-training technique and train several deep classifiers on the obtained representation. Our experimental results demonstrate a substantial improvement in both classification and regression tasks, outperforming models trained directly on samples in their original data space. Moreover, T-JEPA enables some methods to consistently outperform or match the performance of traditional methods likes Gradient Boosted Decision Trees. To understand why, we extensively characterize the obtained representations and show that T-JEPA effectively identifies relevant features for downstream tasks without access to the labels. Additionally, we introduce regularization tokens, a novel regularization method critical for training of JEPA-based models on structured data.