arXiv paper proposes iteratively reweighted least squares for fixed-charge network flow

A new arXiv study, cross-listed in AI and machine learning, tackles the fixed-charge network flow problem, where continuous flow decisions are intertwined with binary choices about which arcs to activate. The authors introduce a method based on iteratively reweighted least squares to discover the support of active arcs, addressing a computationally hard model that underpins many network design and resource allocation tasks. The work offers an alternative optimization perspective on a classic combinatorial challenge.