def test_parameterized_hamiltonian_frechet_derivative(self):
"""
Test gradient-based optimization on parameterized Hamiltonian blocks
"""
vqa = VQA(num_qubits=1)
vqa.cost_observable = qutip.sigmaz()
block = VQABlock(ParameterizedHamiltonian([qutip.sigmax()]))
vqa.add_block(block)
res = vqa.optimize_parameters(
initial=[np.pi / 2 + 0.2], method="BFGS", use_jac=True
)
assert res.get_top_bitstring() == "|1>"
def test_bfgs_optimization(self, use_jac):
"""
test bfgs optimizer, starting from a close initial guess
"""
block = VQABlock(qutip.sigmax())
vqa = VQA(num_qubits=1, cost_method="OBSERVABLE")
vqa.add_block(block)
# try to reach the |1> state from the |0> state
vqa.cost_observable = qutip.sigmaz()
res = vqa.optimize_parameters(
initial=[np.pi / 2 + 0.2], method="BFGS", use_jac=use_jac
)
assert res.get_top_bitstring() == "|1>"
# check we actually found the function minimum
assert round(res.res.x[0], 2) == round(np.pi / 2, 2)
def test_optimization_errors(self):
"""
Tests for value errors relating to optimization procedure
"""
vqa = VQA(num_qubits=1)
vqa.add_block(VQABlock(qutip.sigmax()))
vqa.cost_observable = qutip.sigmax()
with pytest.raises(ValueError):
# Invalid initialization string
res = vqa.optimize_parameters(initial="something else")
with pytest.raises(ValueError):
# Incorrect number of parameters
res = vqa.optimize_parameters(initial=[1, 1])
# Valid initialization string
res = vqa.optimize_parameters(initial="ones")
assert res is not None