Belief Propagation is a simple message-passing algorithm working on a graph representing a probability distribution. The goal of the algorithm is to perform the task of inference, calculating the marginal distributions of any unobserved nodes given a set of observed nodes. It is well known that Belief Propagation struggles when applied to cyclic graphs, albeit it can still provide surprisingly good approximation. To combat the issues arising from having a loopy graph, many variations and adjustments to original Belief Propagation have been published, shedding more light on and providing proofs for how exactly the algorithm handles and is impacted by cycles. This work attempts to provide an entry point and overview of the landscape and the families of Belief Propagation variations, while highlighting the strengths weaknesses of the different approaches.