MinBLoG: Minimization of Boolean Logic Functions using Graph Attention Network

The initial steps of logic synthesis of digital designs involve finding minimized representations of Boolean logic functions. Existing optimization methods rely on iterative minimization operations that can result in a rapid increase in the runtime when the number of variables and terms of the Boolean functions increase. We propose a graph attention network (GAT) based logic minimization approach called MinBLoG, to narrow down the solution search space for Boolean functions. Our approach achieves more than 96% accuracy in identifying implicants that are a part of the minimized solution and ensures functional equivalency through correctness checking procedures. Experiments show that MinBLoG delivers minimization results for a wide range of Boolean functions significantly faster than well-known existing methods.