def test_qaoa(self, w, prob, m, solutions):
""" QAOA test """
seed = 0
aqua_globals.random_seed = seed
os.environ.pop('QISKIT_AQUA_CIRCUIT_CACHE', None)
self.log.debug('Testing %s-step QAOA with MaxCut on graph\n%s', prob,
w)
backend = BasicAer.get_backend('statevector_simulator')
optimizer = COBYLA()
qubit_op, offset = max_cut.get_qubit_op(w)
qaoa = QAOA(qubit_op, optimizer, prob, mixer=m)
# TODO: cache fails for QAOA since we construct the evolution circuit via instruction
quantum_instance = QuantumInstance(backend,
circuit_caching=False,
seed_simulator=seed,
seed_transpiler=seed)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result['eigvecs'][0])
graph_solution = max_cut.get_graph_solution(x)
self.log.debug('energy: %s', result['energy'])
self.log.debug('time: %s', result['eval_time'])
self.log.debug('maxcut objective: %s', result['energy'] + offset)
self.log.debug('solution: %s', graph_solution)
self.log.debug('solution objective: %s', max_cut.max_cut_value(x, w))
self.assertIn(''.join([str(int(i)) for i in graph_solution]),
solutions)
if quantum_instance.has_circuit_caching:
self.assertLess(quantum_instance._circuit_cache.misses, 3)
def test_qaoa(self, w, p, m, solutions):
os.environ.pop('QISKIT_AQUA_CIRCUIT_CACHE', None)
self.log.debug('Testing {}-step QAOA with MaxCut on graph\n{}'.format(
p, w))
np.random.seed(0)
backend = BasicAer.get_backend('statevector_simulator')
optimizer = COBYLA()
qubit_op, offset = max_cut.get_max_cut_qubitops(w)
qaoa = QAOA(qubit_op, optimizer, p, mixer=m)
# TODO: cache fails for QAOA since we construct the evolution circuit via instruction
quantum_instance = QuantumInstance(backend, circuit_caching=False)
result = qaoa.run(quantum_instance)
x = max_cut.sample_most_likely(result['eigvecs'][0])
graph_solution = max_cut.get_graph_solution(x)
self.log.debug('energy: {}'.format(result['energy']))
self.log.debug('time: {}'.format(result['eval_time']))
self.log.debug('maxcut objective: {}'.format(result['energy'] +
offset))
self.log.debug('solution: {}'.format(graph_solution))
self.log.debug('solution objective: {}'.format(
max_cut.max_cut_value(x, w)))
self.assertIn(''.join([str(int(i)) for i in graph_solution]),
solutions)
if quantum_instance.has_circuit_caching:
self.assertLess(quantum_instance._circuit_cache.misses, 3)
def test_change_operator_size(self):
""" QAOA test """
backend = BasicAer.get_backend('statevector_simulator')
optimizer = COBYLA(maxiter=2)
qubit_op, _ = max_cut.get_operator(W1)
qubit_op = qubit_op.to_opflow().to_matrix_op()
seed = 0
aqua_globals.random_seed = seed
qaoa = QAOA(qubit_op, optimizer, P1)
quantum_instance = QuantumInstance(backend,
seed_simulator=seed,
seed_transpiler=seed)
qaoa.run(quantum_instance)
qaoa.operator = (X ^ qubit_op ^ Z)
qaoa.run()
def test_qaoa_qc_mixer_many_parameters(self):
""" QAOA test with a mixer as a parameterized circuit with the num of parameters > 1. """
seed = 0
aqua_globals.random_seed = seed
optimizer = COBYLA()
qubit_op, _ = max_cut.get_operator(W1)
qubit_op = qubit_op.to_opflow()
num_qubits = qubit_op.num_qubits
mixer = QuantumCircuit(num_qubits)
for i in range(num_qubits):
theta = Parameter('θ' + str(i))
mixer.rx(theta, range(num_qubits))
qaoa = QAOA(qubit_op, optimizer=optimizer, p=2, mixer=mixer)
backend = BasicAer.get_backend('statevector_simulator')
quantum_instance = QuantumInstance(backend,
seed_simulator=seed,
seed_transpiler=seed)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result.eigenstate)
print(x)
graph_solution = max_cut.get_graph_solution(x)
self.assertIn(''.join([str(int(i)) for i in graph_solution]), S1)
def test_qaoa_qc_mixer(self, w, prob, solutions, convert_to_matrix_op):
""" QAOA test with a mixer as a parameterized circuit"""
seed = 0
aqua_globals.random_seed = seed
self.log.debug(
'Testing %s-step QAOA with MaxCut on graph with '
'a mixer as a parameterized circuit\n%s', prob, w)
backend = BasicAer.get_backend('statevector_simulator')
optimizer = COBYLA()
qubit_op, _ = max_cut.get_operator(w)
qubit_op = qubit_op.to_opflow()
if convert_to_matrix_op:
qubit_op = qubit_op.to_matrix_op()
num_qubits = qubit_op.num_qubits
mixer = QuantumCircuit(num_qubits)
theta = Parameter('θ')
mixer.rx(theta, range(num_qubits))
qaoa = QAOA(qubit_op, optimizer, prob, mixer=mixer)
quantum_instance = QuantumInstance(backend,
seed_simulator=seed,
seed_transpiler=seed)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result.eigenstate)
graph_solution = max_cut.get_graph_solution(x)
self.assertIn(''.join([str(int(i)) for i in graph_solution]),
solutions)
def run_IBM(H=None,
backend=None,
num_samples=100,
qaoa_steps=3,
variables=None):
num_vars = len(list(H.linear))
qubit_op, offset = get_ising_opt_qubitops(H, variables)
if backend == None:
backend = BasicAer.get_backend('qasm_simulator')
quantum_instance = QuantumInstance(backend)
spsa = SPSA(max_trials=10)
qaoa = QAOA(qubit_op, spsa, qaoa_steps)
result = qaoa.run(quantum_instance)
circ = qaoa.get_optimal_circuit()
q = circ.qregs[0]
c = ClassicalRegister(num_vars, 'c')
circ.cregs.append(c)
circ.measure(q, c)
job = execute(circ, backend, shots=num_samples, memory=True)
return job.result().get_counts(circ)
def test_qaoa(self, w, prob, m, solutions, convert_to_matrix_op):
""" QAOA test """
seed = 0
aqua_globals.random_seed = seed
self.log.debug('Testing %s-step QAOA with MaxCut on graph\n%s', prob,
w)
backend = BasicAer.get_backend('statevector_simulator')
optimizer = COBYLA()
qubit_op, offset = max_cut.get_operator(w)
qubit_op = qubit_op.to_opflow()
if convert_to_matrix_op:
qubit_op = qubit_op.to_matrix_op()
qaoa = QAOA(qubit_op, optimizer, prob, mixer=m)
quantum_instance = QuantumInstance(backend,
seed_simulator=seed,
seed_transpiler=seed)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result.eigenstate)
graph_solution = max_cut.get_graph_solution(x)
self.log.debug('energy: %s', result.eigenvalue.real)
self.log.debug('time: %s', result.optimizer_time)
self.log.debug('maxcut objective: %s',
result.eigenvalue.real + offset)
self.log.debug('solution: %s', graph_solution)
self.log.debug('solution objective: %s', max_cut.max_cut_value(x, w))
self.assertIn(''.join([str(int(i)) for i in graph_solution]),
solutions)
def test_qaoa(self, w, p, m, solutions):
self.log.debug('Testing {}-step QAOA with MaxCut on graph\n{}'.format(
p, w))
np.random.seed(0)
backend = BasicAer.get_backend('statevector_simulator')
optimizer = COBYLA()
qubitOp, offset = max_cut.get_max_cut_qubitops(w)
qaoa = QAOA(qubitOp, optimizer, p, operator_mode='matrix', mixer=m)
quantum_instance = QuantumInstance(backend)
result = qaoa.run(quantum_instance)
x = max_cut.sample_most_likely(result['eigvecs'][0])
graph_solution = max_cut.get_graph_solution(x)
self.log.debug('energy: {}'.format(result['energy']))
self.log.debug('time: {}'.format(result['eval_time']))
self.log.debug('maxcut objective: {}'.format(result['energy'] +
offset))
self.log.debug('solution: {}'.format(graph_solution))
self.log.debug('solution objective: {}'.format(
max_cut.max_cut_value(x, w)))
self.assertIn(''.join([str(int(i)) for i in graph_solution]),
solutions)
if quantum_instance.has_circuit_caching:
self.assertLess(quantum_instance._circuit_cache.misses, 3)
def run_simulation(self, backend):
seed = int(os.environ.get("SEED", "40598"))
n = int(os.environ.get("N", "4"))
#
# Random 3-regular graph with 12 nodes
#
graph = nx.random_regular_graph(3, n, seed=seed)
for e in graph.edges():
graph[e[0]][e[1]]["weight"] = 1.0
# Compute the weight matrix from the graph
w = np.zeros([n, n])
for i in range(n):
for j in range(n):
temp = graph.get_edge_data(i, j, default=0)
if temp != 0:
w[i, j] = temp["weight"]
# Create an Ising Hamiltonian with docplex.
mdl = Model(name="max_cut")
mdl.node_vars = mdl.binary_var_list(list(range(n)), name="node")
maxcut_func = mdl.sum(w[i, j] * mdl.node_vars[i] *
(1 - mdl.node_vars[j]) for i in range(n)
for j in range(n))
mdl.maximize(maxcut_func)
qubit_op, offset = docplex.get_operator(mdl)
aqua_globals.random_seed = seed
# Run quantum algorithm QAOA on qasm simulator
spsa = SPSA(max_trials=250)
qaoa = QAOA(qubit_op, spsa, p=5, max_evals_grouped=4)
quantum_instance = QuantumInstance(
backend,
shots=1024,
seed_simulator=seed,
seed_transpiler=seed,
optimization_level=0,
)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result["eigvecs"][0])
result["solution"] = max_cut.get_graph_solution(x)
result["solution_objective"] = max_cut.max_cut_value(x, w)
result["maxcut_objective"] = result["energy"] + offset
"""
print("energy:", result["energy"])
print("time:", result["eval_time"])
print("max-cut objective:", result["energy"] + offset)
print("solution:", max_cut.get_graph_solution(x))
print("solution objective:", max_cut.max_cut_value(x, w))
"""
return result
def test_change_operator_size(self):
""" QAOA change operator size test """
aqua_globals.random_seed = 0
qubit_op, _ = max_cut.get_operator(
np.array([[0, 1, 0, 1], [1, 0, 1, 0], [0, 1, 0, 1], [1, 0, 1, 0]]))
qaoa = QAOA(qubit_op.to_opflow(), COBYLA(), 1)
quantum_instance = QuantumInstance(
BasicAer.get_backend('statevector_simulator'),
seed_simulator=aqua_globals.random_seed,
seed_transpiler=aqua_globals.random_seed)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result.eigenstate)
graph_solution = max_cut.get_graph_solution(x)
with self.subTest(msg='QAOA 4x4'):
self.assertIn(''.join([str(int(i)) for i in graph_solution]),
{'0101', '1010'})
try:
qubit_op, _ = max_cut.get_operator(
np.array([
[0, 1, 0, 1, 0, 1],
[1, 0, 1, 0, 1, 0],
[0, 1, 0, 1, 0, 1],
[1, 0, 1, 0, 1, 0],
[0, 1, 0, 1, 0, 1],
[1, 0, 1, 0, 1, 0],
]))
qaoa.operator = qubit_op.to_opflow()
except Exception as ex: # pylint: disable=broad-except
self.fail("Failed to change operator. Error: '{}'".format(str(ex)))
return
result = qaoa.run()
x = sample_most_likely(result.eigenstate)
graph_solution = max_cut.get_graph_solution(x)
with self.subTest(msg='QAOA 6x6'):
self.assertIn(''.join([str(int(i)) for i in graph_solution]),
{'010101', '101010'})
def test_portfolio_qaoa(self):
""" portfolio test with QAOA """
qaoa = QAOA(self.qubit_op, COBYLA(maxiter=500), initial_point=[0., 0.])
backend = BasicAer.get_backend('statevector_simulator')
quantum_instance = QuantumInstance(backend=backend,
seed_simulator=self.seed,
seed_transpiler=self.seed)
result = qaoa.run(quantum_instance)
selection = sample_most_likely(result.eigenstate)
value = portfolio.portfolio_value(selection, self.muu, self.sigma,
self.risk, self.budget, self.penalty)
np.testing.assert_array_equal(selection, [0, 1, 1, 0])
self.assertAlmostEqual(value, -0.00679917)
def test_qaoa_random_initial_point(self):
""" QAOA random initial point """
aqua_globals.random_seed = 10598
w = nx.adjacency_matrix(
nx.fast_gnp_random_graph(5, 0.5,
seed=aqua_globals.random_seed)).toarray()
qubit_op, _ = max_cut.get_operator(w)
qaoa = QAOA(qubit_op, NELDER_MEAD(disp=True), 1)
quantum_instance = QuantumInstance(
BasicAer.get_backend('qasm_simulator'),
seed_simulator=aqua_globals.random_seed,
seed_transpiler=aqua_globals.random_seed,
shots=4096)
_ = qaoa.run(quantum_instance)
np.testing.assert_almost_equal([2.5179, 0.3528],
qaoa.optimal_params,
decimal=4)
def test_qaoa_qc_mixer_no_parameters(self):
""" QAOA test with a mixer as a parameterized circuit with zero parameters. """
seed = 0
aqua_globals.random_seed = seed
qubit_op, _ = max_cut.get_operator(W1)
qubit_op = qubit_op.to_opflow()
num_qubits = qubit_op.num_qubits
mixer = QuantumCircuit(num_qubits)
# just arbitrary circuit
mixer.rx(np.pi / 2, range(num_qubits))
qaoa = QAOA(qubit_op, optimizer=COBYLA(), p=1, mixer=mixer)
backend = BasicAer.get_backend('statevector_simulator')
quantum_instance = QuantumInstance(backend,
seed_simulator=seed,
seed_transpiler=seed)
result = qaoa.run(quantum_instance)
# we just assert that we get a result, it is not meaningful.
self.assertIsNotNone(result.eigenstate)
def simulate_optimize(n_shots, p_steps, qubitOp, G):
backend = BasicAer.get_backend('qasm_simulator')
quantum_instance = QuantumInstance(backend, shots=n_shots)
qaoa = QAOA(qubitOp, ESCH(max_evals=100), p=p_steps)
result = qaoa.run(quantum_instance)
# QAOA converts the problem into finding the maximum eigenval/eigenvec pair
solution = max_cut.sample_most_likely(
result['eigvecs'][0]) #returns vector with highest counts
#print('energy:', result['energy'])
print('solution:', solution)
# For p steps, there should be 2*p parameters (beta,gamma)
#print('optimal parameters', result['opt_params'])
cost = 0
for i in range(len(G.nodes)):
for j in range(len(G.nodes)):
cost = cost + adj[i, j] * solution[i] * (1 - solution[j])
print("Sample profit = {}".format(cost))
return result, solution
pos = nx.spring_layout(G)
w = my_graphs.adjacency_matrix(G)
print("\nAdjacency matrix\n", w, "\n")
# setting p
p = 1
# ... QAOA ...
# Create an Ising Hamiltonian with docplex.
mdl = Model(name='max_cut')
mdl.node_vars = mdl.binary_var_list(list(range(n)), name='node')
maxcut_func = mdl.sum(w[i, j] * mdl.node_vars[i] * (1 - mdl.node_vars[j])
for i in range(n) for j in range(n))
mdl.maximize(maxcut_func)
qubit_op, offset = docplex.get_operator(mdl)
# Run quantum algorithm QAOA on qasm simulator
optimizer = NELDER_MEAD()
qaoa = QAOA(qubit_op, optimizer, p=p)
backend = Aer.get_backend('qasm_simulator')
quantum_instance = QuantumInstance(backend, shots=1000)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result.eigenstate)
print('energy:', result.eigenvalue.real)
print('time:', result.optimizer_time, 's')
print('max-cut objective:', result.eigenvalue.real + offset)
print('solution:', max_cut.get_graph_solution(x))
print('solution objective:', max_cut.max_cut_value(x, w))
print('angles:', result.optimal_point)
def run_experiment(
G,
p,
optimizer,
backend,
n_shots=100, #important for running time, max is 8192
print_result=False,
skip_qobj_validation=False,
noise_model=None,
coupling_map=None,
basis_gates=None,
):
'''Runs (Qiskit) QAOA experiment with the given parameters
Inputs
- G, graph
- p, number of angles
- optimizer, classical optimizer
- backend
- number of shots (n_shots) - Note, this parameter is important for running time but also affects accuracy if too low
- ...
Returns a dictionary with the following keys
- energy
- time (in seconds)
- iterations
- max-cut objective
- solution
- solution objective
- eigenstate distribution
- angles
'''
n = len(G) # number of nodes
w = adjacency_matrix(G) # calculating adjacency matrix from graph
# ... QAOA ...
# Create an Ising Hamiltonian with docplex.
mdl = Model(name='max_cut')
mdl.node_vars = mdl.binary_var_list(list(range(n)), name='node')
maxcut_func = mdl.sum(w[i, j] * mdl.node_vars[i] * (1 - mdl.node_vars[j])
for i in range(n) for j in range(n))
mdl.maximize(maxcut_func)
qubit_op, offset = docplex.get_operator(mdl)
# Construct a circuit from the model
qaoa = QAOA(qubit_op, optimizer, p=p)
quantum_instance = QuantumInstance(
backend,
shots=n_shots,
skip_qobj_validation=skip_qobj_validation,
coupling_map=coupling_map,
basis_gates=basis_gates,
noise_model=noise_model)
# Run quantum algorithm QAOA on the backend
result = qaoa.run(quantum_instance)
x = sample_most_likely(result.eigenstate)
# Results
energy = result.eigenvalue.real
time = result.optimizer_time
iterations = result.optimizer_evals
objective = result.eigenvalue.real + offset # why offset?
solution = max_cut.get_graph_solution(x)
solution_objective = max_cut.max_cut_value(x, w)
distribution = result.eigenstate
angles = result.optimal_point
if print_result:
print('energy:', energy)
print('time:', time, 's')
print('max-cut objective:', objective)
print('solution:', solution)
print('solution objective:', solution_objective)
print('angles:', angles)
return {
'energy': energy,
'time': time,
'iterations': iterations,
'max-cut objective': objective,
'solution': solution,
'solution objective': solution_objective,
'distribution': distribution,
'angles': angles,
'n_shots': n_shots
}
def test_readme_sample(self):
""" readme sample test """
# pylint: disable=import-outside-toplevel,redefined-builtin
def print(*args):
""" overloads print to log values """
if args:
self.log.debug(args[0], *args[1:])
# --- Exact copy of sample code ----------------------------------------
import networkx as nx
import numpy as np
from docplex.mp.model import Model
from qiskit import BasicAer
from qiskit.aqua import aqua_globals, QuantumInstance
from qiskit.aqua.algorithms import QAOA
from qiskit.aqua.components.optimizers import SPSA
from qiskit.optimization.applications.ising import docplex, max_cut
from qiskit.optimization.applications.ising.common import sample_most_likely
# Generate a graph of 4 nodes
n = 4
graph = nx.Graph()
graph.add_nodes_from(np.arange(0, n, 1))
elist = [(0, 1, 1.0), (0, 2, 1.0), (0, 3, 1.0), (1, 2, 1.0),
(2, 3, 1.0)]
graph.add_weighted_edges_from(elist)
# Compute the weight matrix from the graph
w = np.zeros([n, n])
for i in range(n):
for j in range(n):
temp = graph.get_edge_data(i, j, default=0)
if temp != 0:
w[i, j] = temp['weight']
# Create an Ising Hamiltonian with docplex.
mdl = Model(name='max_cut')
mdl.node_vars = mdl.binary_var_list(list(range(n)), name='node')
maxcut_func = mdl.sum(w[i, j] * mdl.node_vars[i] *
(1 - mdl.node_vars[j]) for i in range(n)
for j in range(n))
mdl.maximize(maxcut_func)
qubit_op, offset = docplex.get_operator(mdl)
# Run quantum algorithm QAOA on qasm simulator
seed = 40598
aqua_globals.random_seed = seed
spsa = SPSA(max_trials=250)
qaoa = QAOA(qubit_op, spsa, p=5)
backend = BasicAer.get_backend('qasm_simulator')
quantum_instance = QuantumInstance(backend,
shots=1024,
seed_simulator=seed,
seed_transpiler=seed)
result = qaoa.run(quantum_instance)
x = sample_most_likely(result.eigenstate)
print('energy:', result.eigenvalue.real)
print('time:', result.optimizer_time)
print('max-cut objective:', result.eigenvalue.real + offset)
print('solution:', max_cut.get_graph_solution(x))
print('solution objective:', max_cut.max_cut_value(x, w))
# ----------------------------------------------------------------------
self.assertListEqual(
max_cut.get_graph_solution(x).tolist(), [1, 0, 1, 0])
self.assertAlmostEqual(max_cut.max_cut_value(x, w), 4.0)
# (("GBP", "EUR"), 1.),
# (("EUR", "GBP"), 1.3),
# no arb
# (("EUR", "GBP"), 0.88),
# (("GBP", "EUR"), 1.13),
# (("EUR", "CAD"), 1.47),
# (("CAD", "EUR"), 0.68),
# (("GBP", "CAD"), 1.65),
# (("CAD", "GBP"), 0.6),
# GBP -> EUR -> CAD -> GBP makes you money. The other cycles do not.
(("EUR", "GBP"), 0.88),
(("GBP", "EUR"), 1.13),
(("EUR", "CAD"), 1.58),
(("CAD", "EUR"), 0.61),
(("GBP", "CAD"), 1.65),
(("CAD", "GBP"), 0.6),
))
op = get_cost_hamiltonian(rates, 1, 1)
#%%
ee = ExactEigensolver(op)
result = ee.run()
print(bin(result['eigvecs'][0].argmax()))
#%%
p = 1
optimizer = COBYLA()
qaoa = QAOA(op, optimizer, p)
backend = BasicAer.get_backend('statevector_simulator')
quantum_instance = QuantumInstance(backend)
r2 = qaoa.run(quantum_instance)
print(bin(np.absolute(r2['eigvecs'][0]).argmax()))
def libraryoptimize(qubo,
edges,
cedges,
n,
p,
plotoptangles=False,
modulo=False):
if (qubo == "qubo1"):
H = makeNPMatrix1(edges, cedges, n)
if (qubo == "qubo2"):
H = makeNPMatrix2(edges, cedges, n)
if (qubo == "qubo3"):
H = makeNPMatrix2(edges, cedges, n)
optimizer = COBYLA()
qaoa_mes = QAOA(H,
p=p,
optimizer=optimizer,
quantum_instance=Aer.get_backend('statevector_simulator'))
results = qaoa_mes.run()
qc1 = qaoa_mes.get_optimal_circuit()
print(type(results.optimal_parameters))
print(type(results.optimal_parameters.items()))
i = 0
for key, value in results.optimal_parameters.items():
if (i == 0):
gamma = value
i += 1
else:
beta = value
print("beta", beta)
print("gamma", gamma)
print('optimal params: ', results.optimal_parameters)
print('optimal value: ', results.optimal_value)
qcl = qaoa_mes.get_optimal_circuit()
optgammasbetas = []
isgamma = 0
for key, value in results.optimal_parameters.items():
if (modulo):
if (isgamma < p):
optgammasbetas.append(value % (2 * math.pi))
isgamma += 1
else:
optgammasbetas.append(value % (math.pi))
else:
optgammasbetas.append(value)
if (plotoptangles):
plottheoptangles(optgammasbetas, p)
if (qubo == "qubo1"):
qc = makeCircuit1(n, edges, cedges, optgammasbetas, p)
if (qubo == "qubo2"):
qc = makeCircuit2(n, edges, cedges, optgammasbetas, p)
if (qubo == "qubo3"):
qc = makeCircuit3(n, edges, cedges, optgammasbetas, p)
ans = ClassicalRegister(n)
sol = QuantumRegister(n)
QAOA2 = QuantumCircuit(sol, ans)
QAOA2.append(qc1, range(n))
return qc, optgammasbetas