{"id":"arxiv/2603.01842","url":"https://arxiv.org/abs/2603.01842","tags":["cs.NE","math.PR"],"score":0,"title":"Uniform-in-time concentration in two-layer neural networks via transportation inequalities","domain":"cs.NE","source":"arxiv","authors":["Arnaud Guillin","Boris Nectoux","Paul Stos"],"reviews":0,"subject":"cs.NE","summary":"We quantify, uniformly over time and with high probability, the discrepancy between the predictions of a two-layer neural network trained by stochastic gradient descent (SGD) and their mean-field limit, for quadratic loss and ridge regularization. As a key ingredient, we establish T p transportation inequalities (p $\\in$ {1, 2}) for the law of the SGD parameters, with explicit constants independent of the iteration index. We then prove uniform-in-time concentration of the empirical parameter measure around its mean-field limit in the Wasserstein distance W 1 , and we translate these bounds…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-03-02","reactions":{}}