Exploiting symmetries has shown great results in increasing the performance of tensor networks. This thesis explores approaches to implement tensor networks with non-Abelian symmetries, specifically the SU(2) symmetry. The main idea is to reduce the number of free parameters in simulations by restricting the analysis to symmetric tensors. This involves splitting the symmetric tensor into a degeneracy part that holds the degrees of freedom and a structural part that is completely identified by the symmetry. Consequently, it is only necessary to save the degeneracy tensors explicitly. For the structural tensors, we choose a graphical representation of the tensors based on fusion trees, thus eliminating the need to save these tensors.