neural networks (NNs) are increasingly applied in safety-critical systems
such as autonomous vehicles. However, they are fragile and are often
ill-behaved. Consequently, their behaviors should undergo rigorous guarantees
before deployment in practice. In this paper we propose a set-bou