ReCIPH: Relational Coefficients for Input Partitioning Heuristic

With the rapidly advancing improvements to thealready successful Deep Learning artifacts, Neu-ral Networks (NN) are poised to permeate agrowing number of everyday applications, includ-ing ones where safety is paramount and, there-fore, formal guarantees are a precious commodity.To this end, Formal Methods, a long-standing,mathematically-inspired field of research saw aneffervescent outgrowth targeting NN and advanc-ing almost as rapidly as AI itself. Without a doubt,the most challenging problem facing this newresearch direction is the scalability to the ever-growing NN models. This paper stems from thisneed and introduces Relational Coefficients forInput partitioning Heuristic (ReCIPH), accelerat-ing NN analysis. Extensive experimentation issupplied to assert the added value to two differentsolvers handling several models and properties(coming, in part, from two industrial use-cases).