Submodularity is a fundamental aspect of combinatorial optimization that can be used to characterize different combinatorial families of objects. Therefore, the problem of maximizing/minimizing submodular functions has been widely investigated over the years by the mathematics and computer science communities. This fourth edition of the workshop will focus on several aspects of submodularity, e.g.: applications to machine learning; submodularity in the context of approximation algorithms; approximating / learning submodular functions; combinatorial properties; submodular flows. In particular, the goal of the workshop is to highlight the "fil rouge" that runs through the different approaches and applications of submodular optimization.