Id: arxiv/2611.08775 Url: https://arxiv.org/abs/2611.08775 Tags: cs.LG, cs.AI, math.OC, stat.ML Score: 0 Title: On Lipschitz Explosion in Deep Neural Networks with Normalization: Consequences for Optimization and Adversarial Robustness Domain: cs.LG Source: arxiv Authors: Ashkan Soleymani, Reyhaneh Hosseinpourkhoshkbari, Hadi Daneshmand, Patrick Jaillet Reviews: 0 Subject: cs.LG Category: research Comments: 0 Disputed: false Division: cs Mentions: 0 Discussed: 0 Published: 2026-11-05 Page: https://delphi.science/r/arxiv/2611.08775 Can a network look stable at every layer and still be exponentially unstable as a whole? The Lipschitz constant of a neural network is a basic stability quantity appearing in optimization guarantees, generalization bounds, and robustness certificates. Deep networks naturally admit two such notions, input Lipschitzness, measuring sensitivity to data perturbations, and parameter Lipschitzness, measuring sensitivity to weight perturbations. We prove that, for deep networks with normalization layers, with batch normalization as the canonical example, both Lipschitz constants can grow…