Given an undirected graph G and an error parameter > 0, the graph sparsification problem requires sampling edges in G and giving the sampled edges appropriate weights to obtain a sparse graph G with the following property: the weight of every cut in G is within a factor of (1