Influence propagation has been the subject of extensive study due to its important role in social networks, epidemiology, and many other areas. Understanding the propagation mechanism is critical, e.g., to control the spread of fake news or to control an epidemic. In this work, we follow an optimization perspective, and attempt to identify the smallest group of users that needs to be converted to achieve an certain influence level over the entire network. We therefore formally study the generalized least cost influence problem, proposing mathematical programming algorithms to solve the challenging problem. We introduce new cutting plane and separation algorithms and embed them into a branch-and-cut algorithm. Our experimental results on classical benchmark instances demonstrates the method ability to solve small-to medium-scale benchmark instances, also finding optimal solutions for some open problems.