def test_generateneighbors():
J = sps.dok_matrix((5, 5), dtype=np.float64)
#
# We're going to build this graph:
#
# 0 --- 1
# | / |
# | 4 |
# | / |
# 3 --- 2
#
J[0, 1] = 1
J[1, 2] = 1
J[2, 3] = 1
J[3, 4] = 1
J[0, 3] = 1
J[1, 4] = 1
J[0, 4] = 1
J[0, 0] = 1
J[1, 1] = 1
J[2, 2] = 1
J[3, 3] = 1
J[4, 4] = 1
true_nb = np.array([[[1., 1.], [0., 1.], [4., 1.], [3., 1.]],
[[0., 1.], [2., 1.], [4., 1.], [1., 1.]],
[[1., 1.], [2., 1.], [3., 1.], [0., 0.]],
[[3., 1.], [2., 1.], [0., 1.], [4., 1.]],
[[4., 1.], [1., 1.], [0., 1.], [3., 1.]]])
generated_nb = tools.GenerateNeighbors(nspins=5,
J=J,
maxnb=4,
savepath=None)
assert (np.all(true_nb == generated_nb))
def setUp(self):
# Define some parameters
self.nspins = 8
self.preannealingtemp = 1.0
self.annealingtemp = 0.01
self.annealingsteps = 10
self.annealingmcsteps = 3
self.trotterslices = 5
self.fieldstart = 0.5
self.fieldend = 1e-8
# Random number generator
self.rng = np.random.RandomState(123)
# Construct Ising matrix
kvp = [
[1, 2, 1.0],
[1, 4, 1.0],
[1, 5, 1.0],
[2, 3, 1.0],
[2, 6, 1.0],
[3, 4, 1.0],
[3, 7, 1.0],
[4, 8, 1.0],
[1, 1, 1.0],
[2, 2, 1.0],
[3, 3, 1.0],
[4, 4, 1.0],
[5, 5, -1.0],
[6, 6, -1.0],
[7, 7, -1.0],
[8, 8, -1.0]
]
self.isingJ = sps.dok_matrix((self.nspins,self.nspins))
for i,j,val in kvp:
self.isingJ[i-1,j-1] = val
# get the neighbors data structure (tested elsewhere)
self.nbs = tools.GenerateNeighbors(self.nspins,
self.isingJ,
4, None)
for b in [bin(x)[2:].rjust(nspins, '0') for x in range(2**nspins)]:
bvec = np.array([int(k) for k in b])
svec = bitstr2spins(b)
bstr = reduce(lambda x, y: x + y, [str(k) for k in bvec])
results.append([sa.ClassicalIsingEnergy(svec, isingJ), bstr])
for res in sorted(results):
print res
# Initialize random state
spinVector = np.array([2 * rng.randint(2) - 1 for k in range(nspins)],
dtype=np.float)
spinVector_original = spinVector.copy()
configurations = np.tile(spinVector, (trotterslices, 1)).T
# Generate list of nearest-neighbors for each spin
neighbors = tools.GenerateNeighbors(nspins, isingJ, 4)
# Generate annealing schedules
tannealingsched = np.linspace(preannealingtemp, annealingtemp, annealingsteps)
annealingsched = np.linspace(fieldstart, fieldend, annealingsteps)
# Try using SA (deterministic start)
print("SA results using same starting state:")
print("Starting state: ", sa.ClassicalIsingEnergy(spinVector, isingJ),
getbitstr(spinVector))
for sa_itr in range(trotterslices):
spinVector = spinVector_original.copy()
sa.Anneal(tannealingsched, preannealingmcsteps, spinVector, neighbors, rng)
print(sa.ClassicalIsingEnergy(spinVector, isingJ), getbitstr(spinVector))
# Try using SA (random start)
print("SA results using random state (start and end):")
for sa_itr in range(trotterslices):
spinVector = np.array([2 * rng.randint(2) - 1 for k in range(nspins)],
fieldstart = 1.5
fieldend = 1e-8
seed = None
trotterslices = 5
# Random number generator
rng = np.random.RandomState(seed)
# Read from textfile directly to be sure
inputfname = 'ising_instances/santoro_80x80.txt'
loaded = np.loadtxt(inputfname)
isingJ = sps.dok_matrix((nspins, nspins))
for i, j, val in loaded:
isingJ[i - 1, j - 1] = val
# Get list of nearest-neighbors for each spin
neighbors = tools.GenerateNeighbors(
nspins, isingJ, 4, 'ising_instances/santoro_80x80_neighbors.npy')
# neighbors = np.load('ising_instances/santoro_80x80_neighbors.npy')
# initialize the state
spinVector = np.array([2 * rng.randint(2) - 1 for k in range(nspins)],
dtype=np.float)
configurations = np.tile(spinVector, (trotterslices, 1)).T
# Try just using SA
tannealingsched = np.linspace(preannealingtemp, annealingtemp, annealingsteps)
cProfile.runctx(
"sa.Anneal(tannealingsched, annealingmcsteps, "
"spinVector, neighbors, rng)", globals(), locals(), "sa.prof")
s = pstats.Stats("sa.prof")
s.strip_dirs().sort_stats("time").print_stats()
# Now do PIQA
annealingsched = np.linspace(fieldstart, fieldend, annealingsteps)
def bitstr2spins(vec):
""" Take a bitstring and return a spinvector. """
a = [int(k) for k in vec]
return tools.bits2spins(a)
for b in [bin(x)[2:].rjust(nspins, '0') for x in range(2**nspins)]:
bvec = np.array([int(k) for k in b])
svec = bitstr2spins(b)
bstr = reduce(lambda x, y: x + y, [str(k) for k in bvec])
results.append([sa.ClassicalIsingEnergy(svec, isingJ), bstr])
for res in sorted(results): #[:100]:
print res
print("\n")
# Generate list of nearest-neighbors for each spin
neighbors = np.asarray(tools.GenerateNeighbors(nspins, isingJ, 5))
# Generate annealing schedules
tannealingsched = np.linspace(preannealingtemp, annealingtemp, annealingsteps)
# Generate random states to compare
svecs = np.asarray([[2 * rng.randint(2) - 1 for k in range(nspins)]
for j in xrange(samples)],
dtype=np.float)
# Try using SA (random start)
for sa_itr in xrange(samples):
# spinVector = np.array([ 2*rng.randint(2)-1 for k in range(nspins) ],
# dtype=np.float)
spinVector = svecs[sa_itr].copy()
starten, startstate = (sa.ClassicalIsingEnergy(spinVector, isingJ),
getbitstr(spinVector))
sa.Anneal(tannealingsched, annealingmcsteps, spinVector, neighbors, rng)
bitstr = getbitstr(spinVector)