This work has two goals, foremost the design and implementation of methods to compress a Neural Network. Two experimental compression methods were designed and implemented for this purpose, the first method named Least Significant Path Pruner, calculates the shortest paths from each output node of the Neural Network to any input node. And the pruning of edges common in these paths, by setting the weight to zero until a target sparsity for the neural network is reached. The second method is named Least Significant Node Merger, which merges multiple nodes from a layer into one, until the targeted size is reached. Removing edges and nodes of the network in the process. The second goal of the work is to implement a tool, which can visualize Neural Networks as graphs and features the ability to compare two different Networks against each other. The tool should also be able to display the differences emerging through the compression of the Neural Network. The result is the application “ShowDiff”, which combines these two goals into an application, making it possible to compress a model and then compare with the Network it originated from.