English

Effective Neural Network $L_0$ Regularization With BinMask

Machine Learning 2023-04-25 v1

Abstract

L0L_0 regularization of neural networks is a fundamental problem. In addition to regularizing models for better generalizability, L0L_0 regularization also applies to selecting input features and training sparse neural networks. There is a large body of research on related topics, some with quite complicated methods. In this paper, we show that a straightforward formulation, BinMask, which multiplies weights with deterministic binary masks and uses the identity straight-through estimator for backpropagation, is an effective L0L_0 regularizer. We evaluate BinMask on three tasks: feature selection, network sparsification, and model regularization. Despite its simplicity, BinMask achieves competitive performance on all the benchmarks without task-specific tuning compared to methods designed for each task. Our results suggest that decoupling weights from mask optimization, which has been widely adopted by previous work, is a key component for effective L0L_0 regularization.

Keywords

Cite

@article{arxiv.2304.11237,
  title  = {Effective Neural Network $L_0$ Regularization With BinMask},
  author = {Kai Jia and Martin Rinard},
  journal= {arXiv preprint arXiv:2304.11237},
  year   = {2023}
}
R2 v1 2026-06-28T10:14:13.355Z