AI RESEARCH
Efficient Verification of Neural Control Barrier Functions with Smooth Nonlinear Activations
arXiv CS.LG
•
ArXi:2605.07757v1 Announce Type: new Formal verification of neural control barrier functions (NCBFs) remains challenging, especially for neural networks with nonlinear activations like \(\tanh\). Existing CROWN-based methods rely on conservative linear relaxations for Jacobian bounds, limiting scalability. We propose LightCROWN, which computes tighter Jacobian bounds by exploiting the analytical properties of activation functions.