def fel_measmax(x, fun, identifier, fun_evals):
fun_none = []
vqe = vqe_nelder_mead(fatol=1e-3)
ansatz_name = identifier[1]
size = identifier[0]
for fun_evals_ in fun_evals:
fun_evals_ = int(fun_evals_)
if len(fun) < fun_evals_:
print(len(fun))
raise ValueError(f'fun_evals={fun_evals_} is to big')
fun_ = fun[:fun_evals_]
idx = np.argmin(fun_)
x_min = x[idx]
mat_idx = identifier[4]
h = hamiltonians_of_size(size)[0][mat_idx]
H, qc, ansatz_, _ = create_and_convert(ansatz_name, h)
fun_none_ = vqe.expectation(ansatz_(x_min), H, samples=None, qc=qc)[0]
fun_none.append(fun_none_)
return np.asarray(fun_none)
def simulate(size, ansatz_name, minimizer, repeats, hamiltonian_idx, max_meas,
samples, h, eig):
H, qc, ansatz_, initial_params = \
core.interface.create_and_convert(ansatz_name, h)
if minimizer == 'nelder-mead':
vqe = vqe_nelder_mead(samples=samples, H=H, fatol=0, xatol=0)
tol_para = 1e-3
max_same_para = 3
callback = cb.restart(max_same_para, tol_para)
elif minimizer == 'bayes':
n_calls = int(round(max_meas/samples))
vqe = vqe_bayes(n_calls=n_calls)
def callback(*args, **kwargs):
pass
if ansatz_name == 'multi_particle':
initial_params = [(-1.0, 1.0)]*(size - 1)
elif ansatz_name == 'one_particle_ucc':
initial_params = [(-3.0, 3.0)] * (size - 1)
else:
raise RuntimeError("Don't know that ansatz.")
else:
raise RuntimeError('Bad minimizer')
result = vqe_eig.smallest(H, qc, initial_params, vqe,
ansatz_, samples,
callback=callback, max_meas=max_meas)
result.correct = eig
return result
def simulate(size, hamiltonian_idx, V, max_meas, samples, h, eig):
H, qc, ansatz_, initial_params = \
core.interface.create_and_convert('multi_particle', h)
vqe = vqe_nelder_mead(samples=samples, H=H, fatol=0, xatol=0)
tol_para = 1e-3
max_same_para = 3
callback = cb.restart(max_same_para, tol_para)
result = vqe_eig.smallest(H, qc, initial_params, vqe,
ansatz_, samples,
callback=callback, max_meas=max_meas)
result.correct = eig
return result
def simulate(ansatz_name, size, hamiltonian_idx, samples, repeats, h, eig):
# TODO: create VQE-object here! (not multiprocess safe)
# TODO: run e.g. smallest here and return result.
H, qc, ansatz_, initial_params = core.interface.create_and_convert(
ansatz_name, h)
vqe = vqe_nelder_mead(samples=samples, H=H)
max_fun_evals = 200
result = vqe_eig.smallest(H,
qc,
initial_params,
vqe,
ansatz_,
samples,
max_meas=samples * max_fun_evals)
result.correct = eig
return result