def test_clamped_model_sample_multple(self):
"""
Test that clamped model sample with multiple clamped variables gets the
correct energy.
"""
checkerboard = IsingModel(
J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1},
h={}
)
checkerboard.clamp(1, 1)
checkerboard.clamp(3, -1)
sample = IsingSample(checkerboard, (-1, 1))
assert sample.assignment.as_tuple == (-1, 1, 1, -1)
assert sample.energy == 0
checkerboard = IsingModel(
J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1},
h={1: 4}
)
checkerboard.clamp(0, -1)
checkerboard.clamp(1, 1)
checkerboard.clamp(2, 1)
checkerboard.clamp(3, -1)
sample = IsingSample(checkerboard, tuple())
assert sample.assignment.as_tuple == (-1, 1, 1, -1)
assert sample.energy == 4
def test_variables_space_correct(self):
"""
Tests that IsingModel correctly determines the k-local and 2-local
space of variables correctly.
"""
h = {0: 20.4, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
assert model.variables == set([0, 1, 2])
# Present in h, but not present in J
h = {0: 20.4, 1: 10, 2: -5, 3: 13}
J = {(0, 1): -4.5, (1, 3): 5}
model = IsingModel(J, h)
assert model.variables == set([0, 1, 2, 3])
# Present in J, but not present in h
h = {0: 20.4, 1: 10}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
assert model.variables == set([0, 1, 2])
def test_clamp_variable_invalid_value(self):
"""
Make sure invalid variable cannot be clamped.
"""
h = {0: 4.5, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
# Attempt to clamp with some retarded value
with pytest.raises(ValueError):
model.clamp(2, 5)
with pytest.raises(ValueError):
model.clamp(2, -3)
with pytest.raises(ValueError):
model.clamp(2, 1.01)
with pytest.raises(ValueError):
model.clamp(2, -0.99)
with pytest.raises(ValueError):
model.clamp(2, 0)
with pytest.raises(ValueError):
model.clamp(2, -0.5)
def test_initialize_by_adding(self):
"""
Initialize SamplePool with a list of existing samples.
"""
simple = IsingModel(
J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1},
h={0: -3}
)
samples = [
IsingSample(simple, [-1, 1, 1, 1]),
IsingSample(simple, [1, -1, 1, -1]),
IsingSample(simple, [1, 1, -1, 1]),
IsingSample(simple, [-1, -1, -1, -1])
]
pool = SamplePool()
for sample in samples:
pool.add(sample)
assert len(pool) == 4
assert pool[0] == samples[1]
# Make sure the entire order is correct
assert pool[:1000] == sorted(samples, reverse=True)
def generate(self):
h = {}
J = {}
for (node1, node2) in self.graph.edges:
value = (-1) * self.binpix(*node1) * self.binpix(*node2)
J[self.variable(*node1), self.variable(*node2)] = value
return IsingModel(J, h)
def test_clamp_variable_simple(self):
"""
Makes sure that variable clamping transforms IsingModel in an expected manner.
"""
h = {0: 4.5, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
model.clamp(0, 1)
assert model.h_clamped == {1: 5.5, 2: -5}
assert model.J_clamped == {(1, 2): 5}
model = IsingModel(J, h)
model.clamp(0, -1)
assert model.h_clamped == {1: 14.5, 2: -5}
assert model.J_clamped == {(1, 2): 5}
def test_biased_checkerboard(self):
"""
Test that biased checkerboard (at one qubit) returns only one solution.
"""
sampler = GibbsSampler()
biased_checkerboard = IsingModel(J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1}, h={0: 2})
result = sampler.sample(biased_checkerboard, 10000)
solutions = [s.as_tuple for s in result]
assert (-1, 1, -1, 1) in solutions
assert len(solutions) == 1
# Enforce the opposite checkerboard
biased_checkerboard = IsingModel(J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1}, h={0: -2})
result = sampler.sample(biased_checkerboard, 10000)
solutions = [s.as_tuple for s in result]
assert (1, -1, 1, -1) in solutions
assert len(solutions) == 1
def generate(self, decimals):
J = {}
h = {}
for (node1, node2) in self.graph.edges:
J[self.variable(*node1),
self.variable(*node2)] = round(2 * random.random() - 1, decimals)
for node in self.graph.nodes:
h[self.variable(*node)] = round(2 * random.random() - 1, decimals)
return IsingModel(J, h)
def test_sample_energy(self):
"""
Test that the energy value computation is correct.
"""
h = {}
J = {(0, 1): -2.5, (1, 2): 3}
model = IsingModel(J, h)
assert IsingSample(model, {0: -1, 1: -1, 2: 1}).energy == pytest.approx(-5.5)
h = {0: 5.2, 1: 4, 2: -8}
J = {}
model = IsingModel(J, h)
assert IsingSample(model, {0: -1, 1: 1, 2: -1}).energy == pytest.approx(6.8)
h = {0: 5.2, 1: 4, 2: -8}
J = {(0, 1): -2.5, (1, 2): 3}
model = IsingModel(J, h)
assert IsingSample(model, {0: -1, 1: -1, 2: -1}).energy == pytest.approx(-0.7)
def test_initialization(self):
"""
Tests that IsingModel can be created.
"""
h = {0: 20.4, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
# Something is very broken if even this test does not work
model = IsingModel(J, h)
assert model is not None
def test_unbiased_checkerboard(self):
"""
Test that checkerboard problem returns both valid solutions.
"""
sampler = GibbsSampler()
checkerboard = IsingModel(J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1}, h={})
result = sampler.sample(checkerboard, 10000)
solutions = [s.as_tuple for s in result]
assert (-1, 1, -1, 1) in solutions
assert (1, -1, 1, -1) in solutions
def test_clamp_variable_not_present(self):
"""
Make sure invalid variable cannot be clamped.
"""
h = {0: 4.5, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
with pytest.raises(ValueError):
model.clamp(8, 1)
def test_clamped_model_sample(self):
"""
Test that clamped model sample gets the correct energy.
"""
checkerboard = IsingModel(
J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1},
h={}
)
checkerboard.clamp(0, 1)
sample = IsingSample(checkerboard, (-1, 1, -1))
assert sample.assignment.as_tuple == (1, -1, 1, -1)
assert sample.assignment == {0: 1, 1: -1, 2: 1, 3: -1}
assert sample.energy == -4
checkerboard = IsingModel(
J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1},
h={0: 5}
)
checkerboard.clamp(0, 1)
sample = IsingSample(checkerboard, (-1, 1, -1))
assert sample.assignment.as_tuple == (1, -1, 1, -1)
assert sample.assignment == {0: 1, 1: -1, 2: 1, 3: -1}
assert sample.energy == 1
checkerboard = IsingModel(
J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1},
h={0: -3, 3: 2}
)
checkerboard.clamp(3, -1)
sample = IsingSample(checkerboard, (-1, -1, -1))
assert sample.assignment.as_tuple == (-1, -1, -1, -1)
assert sample.assignment == {0: -1, 1: -1, 2: -1, 3: -1}
assert sample.energy == 5
def test_dwave_conversion_big_J_element(self):
"""
Test that the conversion to the D-Wave fills elements in the h vector
all the way up to maximum J variable.
"""
h = {0: 20.4, 1: 10, 4: -5}
J = {(0, 1): -4.5, (1, 2): 5, (0, 3): 2}
h_dwave, J_dwave = IsingModel(J, h).as_dwave()
assert h_dwave == [20.4, 10, 0, 0, -5]
assert J_dwave == {(0, 1): -4.5, (1, 2): 5, (0, 3): 2}
def test_dwave_conversion_h_inter(self):
"""
Test that the conversion to the D-Wave fills in missing elements in the
h dictionary.
"""
h = {0: 20.4, 1: 10, 3: -5}
J = {(0, 1): -4.5, (1, 2): 5}
h_dwave, J_dwave = IsingModel(J, h).as_dwave()
assert h_dwave == [20.4, 10, 0, -5]
assert J_dwave == {(0, 1): -4.5, (1, 2): 5}
def test_serialize_deserialize(self):
"""
Test that deserialized version of a serialized IsingModel defines the
same model.
"""
h = {0: 20.4, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 3): 5}
serialized = IsingModel(J, h).serialize()
deserialized = IsingModel.deserialize(serialized)
assert deserialized.h == h
assert deserialized.J == J
def test_sample_ordering(self):
"""
Tests that IsingSamples are ordered according to their negative energy.
"""
h = {0: 4.5, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
assert IsingSample(model, (1, 1, 1)) < IsingSample(model, (1, -1, 1))
assert IsingSample(model, (1, 1, -1)) < IsingSample(model, (-1, -1, 1))
assert IsingSample(model, (-1, 1, 1)) < IsingSample(model, (1, 1, 1))
assert IsingSample(model, (1, 1, 1)) < IsingSample(model, (-1, -1, 1))
def test_valid_sample(self):
"""
Create valid samples for a given model.
"""
h = {0: 20.4, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
# Something is very broken if even this test does not work
sample = IsingSample(model, {0: 1, 1: -1, 2: 1})
sample = IsingSample(model, {0: 1, 1: 1, 2: 1}, occurences=54)
sample = IsingSample(model, {0: -1, 1: 1, 2: -1}, occurences=23)
assert sample is not None
def test_invalid_sample_length(self):
"""
Attempt to create a sample with invalid number of variables.
"""
h = {0: 20.4, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
with pytest.raises(ValueError):
sample = IsingSample(model, {0: -1, 1: 1})
with pytest.raises(ValueError):
sample = IsingSample(model, {0: 1, 1: 1, 2: 1, 3: -1})
with pytest.raises(ValueError):
sample = IsingSample(model, {})
def test_clamp_multiple_variables(self):
"""
Make sure clamping multiple variables work.
"""
h = {0: 4, 1: 10, 2: -5, 3: 4, 4: -5}
J = {(0, 1): -6, (1, 2): 5, (0, 3): 3, (0, 2): -4}
model = IsingModel(J, h)
model.clamp(0, -1)
model.clamp(2, 1)
model.clamp(4, -1)
assert model.h_clamped == {1: 21, 3: 1}
assert model.J_clamped == {}
def test_clamp_variable_already_clamped(self):
"""
Make sure invalid variable cannot be clamped.
"""
h = {0: 4.5, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
# Clamp the variable
model.clamp(2, -1)
# Attempt to clamp the variable again
with pytest.raises(ValueError):
model.clamp(2, 1)
def test_sample_equality(self):
"""
Test that the samples are equal iff assignments are equal.
"""
h = {0: 4.5, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
# Equal assignments imply equal samples
assert IsingSample(model, (-1, 1, 1)) == IsingSample(model, (-1, 1, 1))
# Unequal assignments imply unequal samples
assert IsingSample(model, (1, 1, 1)) != IsingSample(model, (1, -1, 1))
assert IsingSample(model, (1, -1, 1)) != IsingSample(model, (1, 1, -1))
assert IsingSample(model, (-1, 1, 1)) != IsingSample(model, (1, 1, 1))
def test_clamp_not_affecting_original(self):
"""
Make sure that the original ising model definition is not affected by
clamping variables.
"""
h = {0: 4, 1: 10, 2: -5, 3: 4, 4: -5}
J = {(0, 1): -6, (1, 2): 5, (0, 3): 3, (0, 2): -4}
model = IsingModel(J, h)
model.clamp(0, -1)
model.clamp(2, 1)
model.clamp(4, -1)
assert model.h == {0: 4, 1: 10, 2: -5, 3: 4, 4: -5}
assert model.J == {(0, 1): -6, (1, 2): 5, (0, 3): 3, (0, 2): -4}
def test_sample_deserialize_serialize(self):
"""
Tests that a sample can be deserialized and then serialized into a
duplicate object containing same content.
"""
h = {0: 4.5, 1: 10, 2: -5}
J = {(0, 1): -4.5, (1, 2): 5}
model = IsingModel(J, h)
sample = IsingSample(model, (-1, 1, -1))
deserialized = IsingSample.deserialize(model, sample.serialize())
# Assert equality
assert sample == deserialized
assert sample.energy == deserialized.energy
assert sample.occurences == deserialized.occurences
def test_occurrence_growing(self):
"""
Check that the number of occurrences is growing when same sample is
submitted.
"""
simple = IsingModel(
J={(0, 1): 1, (1, 2): 1, (2, 3): 1, (3, 0): 1},
h={0: -3}
)
pool = SamplePool()
pool.add(IsingSample(simple, [1, 1, 1, 1], occurences=3))
assert len(pool) == 1
assert pool[0].occurences == 3
pool.add(IsingSample(simple, [1, 1, 1, 1], occurences=2))
assert len(pool) == 1
assert pool[0].occurences == 5
pool.add(IsingSample(simple, [1, 1, 1, 1]))
assert len(pool) == 1
assert pool[0].occurences == 6
pool.add(IsingSample(simple, [-1, -1, -1, -1], occurences=15))
assert len(pool) == 2
assert pool[0].occurences == 6
assert pool[1].occurences == 15
pool.add(IsingSample(simple, [-1, -1, -1, -1], occurences=5))
assert len(pool) == 2
assert pool[0].occurences == 6
assert pool[1].occurences == 20