def test_edge_driver_errors(self):
"""Tests that the edge driver Hamiltonian throws the correct errors"""
with pytest.raises(ValueError, match=r"Encountered invalid entry in 'reward', expected 2-bit bitstrings."):
qaoa.edge_driver(Graph([(0, 1), (1, 2)]), ['10', '11', 21, 'g'])
with pytest.raises(
ValueError,
match=r"'reward' cannot contain either '10' or '01', must contain neither or both."
):
qaoa.edge_driver(Graph([(0, 1), (1, 2)]), ['11', '00', '01'])
with pytest.raises(ValueError, match=r"Input graph must be a nx.Graph"):
qaoa.edge_driver([(0, 1), (1, 2)], ['00', '11'])
def test_edge_driver_output(self, graph, reward, hamiltonian):
"""Tests that the edge driver Hamiltonian throws the correct errors"""
H = qaoa.edge_driver(graph, reward)
assert decompose_hamiltonian(H) == decompose_hamiltonian(hamiltonian)
# one of the core principles behind the entire PennyLane library is customizability, and this principle hold true for # QAOA submodule as well! # # The QAOA workflow above gave us two optimal solutions: :math:`|6\rangle = |0110\rangle` # and :math:`|10\rangle = |1010\rangle`. What if we add a constraint # that made one of these solutions "better" than the other? Let's imagine that we are interested in # solutions that minimize the original cost function, # *but also colour the first and third vertices* :math:`1`. A constraint of this form will # favour :math:`|10\rangle`, making it the only true ground state. # # It is easy to introduce constraints of this form in PennyLane. # We can use the :func:`~.pennylane.qaoa.edge_driver` cost # Hamiltonian to "reward" cases in which the first and last vertices of the graph # are :math:`0`: reward_h = qaoa.edge_driver(nx.Graph([(0, 2)]), ['11']) ###################################################################### # # We then weigh and add the constraining term # to the original minimum vertex cover Hamiltonian: new_cost_h = cost_h + 2 * reward_h ###################################################################### # # Notice that PennyLane allows for simple addition and multiplication of # Hamiltonian objects using inline arithmetic operations ➕ ➖ ✖️➗! Finally, we can # use this new cost Hamiltonian to define a new QAOA workflow: