English

Certifying Robustness via Topological Representations

Machine Learning 2025-01-22 v1 Computational Geometry Machine Learning

Abstract

We propose a neural network architecture that can learn discriminative geometric representations of data from persistence diagrams, common descriptors of Topological Data Analysis. The learned representations enjoy Lipschitz stability with a controllable Lipschitz constant. In adversarial learning, this stability can be used to certify ϵ\epsilon-robustness for samples in a dataset, which we demonstrate on the ORBIT5K dataset representing the orbits of a discrete dynamical system.

Keywords

Cite

@article{arxiv.2501.10876,
  title  = {Certifying Robustness via Topological Representations},
  author = {Jens Agerberg and Andrea Guidolin and Andrea Martinelli and Pepijn Roos Hoefgeest and David Eklund and Martina Scolamiero},
  journal= {arXiv preprint arXiv:2501.10876},
  year   = {2025}
}

Comments

Workshop on Symmetry and Geometry in Neural Representations (NeurReps) at NeurIPS 2024, Extended Abstract Track

R2 v1 2026-06-28T21:10:22.933Z