Pruning encompasses a range of techniques aimed at increasing the sparsity of neural networks (NNs). These techniques can generally be framed as minimizing a loss function subject to an L0 norm constraint. This paper introduces CoNNect, a novel differentiable regularizer for sparse NN training that ensures connectivity between input and output layers. We prove that CoNNect approximates L0 regularization, guaranteeing maximally connected network structures while avoiding issues like layer collapse. Moreover, CoNNect is easily integrated with established structural pruning strategies. Numerical experiments demonstrate that CoNNect can improve classical pruning strategies and enhance state-of-the-art one-shot pruners, such as DepGraph and LLM-pruner.
@article{arxiv.2502.00744,
title = {CoNNect: Connectivity-Based Regularization for Structural Pruning},
author = {Christian Franssen and Jinyang Jiang and Yijie Peng and Bernd Heidergott},
journal= {arXiv preprint arXiv:2502.00744},
year = {2025}
}