While a decision expert may be able to manually specify the capacity values for a small input set, most real-world applications will require software and data-driven approaches. In this chapter, we focus on optimisation-based methods for learning capacities from data. The capacity data-fitting problem can be approached with respect to different metrics, resulting in either convex or non-convex optimisation problems. We will show that the sparse capacities of the previous chapter result in simplified problems that can help us address tractability issues when working with a large number of variables. Specialised software packages used to implement these methods are also presented.