Tabu search, simulated annealing or genetic algorithme

Hi everyone !
I'am studing myself some meta-heuristics methods. I would like to solve this problem using one of these methods: Given a directed graph G = (V,E), U is a subset of V, G' = (U,E(U)) is a induced acyclic subgraph of G if E(U) is a subset of E that contain only arcs with head and tail in U. I would like to find a maximal induced acyclic subgraph of G, that's a induced acyclic subgraph of G that contain a maximal number of vertex.
The first question coming in my head is: which method (tabu search, simulated annealing or genetic algorithm) i should use to solve this problem ? After a very short time of studying 3 methods, i think of using tabu search(I haven't got any idea for the solution, that's just because i like tabu search ).
Could you give me some ideas about this choice ?
Thank in advance !

http://en.wikipedia.org/wiki/Tabu_search