Pseudo-Non-Linear Data Augmentation: A Constrained Energy Minimization Viewpoint

Pingbang Hu, Mahito Sugiyama

Keywords: data augmentation, information geometry, energy-based model

TL;DR: We propose a simple, information-geometric approach to data augmentation that is learning-free, efficient, controllable, and broadly applicable to structured data.

Abstract: We propose a simple yet novel data augmentation method for general data modalities based on energy-based modeling and principles from information geometry. Unlike most existing learning-based data augmentation methods, which rely on learning latent representations with generative models, our proposed framework enables an intuitive construction of a geometrically aware latent space that represents the structure of the data itself, supporting efficient and explicit encoding and decoding procedures. We then present and discuss how to design latent spaces that will subsequently control the augmentation with the proposed algorithm. Empirical results demonstrate that our data augmentation method achieves competitive performance in downstream tasks compared to other baselines, while offering fine-grained controllability that is lacking in the existing literature.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Submission Number: 16154