English

Demystifying Batch Normalization in ReLU Networks: Equivalent Convex Optimization Models and Implicit Regularization

Machine Learning 2022-03-22 v3 Optimization and Control Machine Learning

Abstract

Batch Normalization (BN) is a commonly used technique to accelerate and stabilize training of deep neural networks. Despite its empirical success, a full theoretical understanding of BN is yet to be developed. In this work, we analyze BN through the lens of convex optimization. We introduce an analytic framework based on convex duality to obtain exact convex representations of weight-decay regularized ReLU networks with BN, which can be trained in polynomial-time. Our analyses also show that optimal layer weights can be obtained as simple closed-form formulas in the high-dimensional and/or overparameterized regimes. Furthermore, we find that Gradient Descent provides an algorithmic bias effect on the standard non-convex BN network, and we design an approach to explicitly encode this implicit regularization into the convex objective. Experiments with CIFAR image classification highlight the effectiveness of this explicit regularization for mimicking and substantially improving the performance of standard BN networks.

Keywords

Cite

@article{arxiv.2103.01499,
  title  = {Demystifying Batch Normalization in ReLU Networks: Equivalent Convex Optimization Models and Implicit Regularization},
  author = {Tolga Ergen and Arda Sahiner and Batu Ozturkler and John Pauly and Morteza Mardani and Mert Pilanci},
  journal= {arXiv preprint arXiv:2103.01499},
  year   = {2022}
}

Comments

Accepted to ICLR 2022. First two authors contributed equally to this work; 36 pages, 13 figures

R2 v1 2026-06-23T23:38:53.350Z