def create_3Dlabels_Potts(condition_defs, beta, dims, mask):
labels = []
graph = graph_from_lattice(mask, kerMask=kerMask3D_6n, toroidal=True)
for c in condition_defs:
tmp = genPotts(graph, beta, 2)
tmp = np.reshape(tmp, dims)
labels.append(tmp)
return np.array(labels)
def create_3Dlabels_Potts(condition_defs, beta, dims, mask):
labels = []
graph = graph_from_lattice(mask, kerMask=kerMask3D_6n, toroidal=True)
for c in condition_defs:
tmp = genPotts(graph, beta, 2)
tmp = np.reshape(tmp, dims)
labels.append(tmp)
return np.array(labels)
def create_labels_Potts(condition_defs, beta, nb_voxels):
labels = []
nb_voxels = nb_voxels**.5 # assume square shape
shape = (nb_voxels, nb_voxels)
mask = np.ones(shape, dtype=int)
graph = graph_from_lattice(mask, kerMask=kerMask2D_4n, toroidal=True)
shape = (1, nb_voxels, nb_voxels)
for c in condition_defs:
tmp = genPotts(graph, beta, 2)
tmp = np.reshape(tmp, shape)
labels.append(tmp)
return np.array(labels)
def create_labels_Potts(condition_defs, beta, nb_voxels):
labels = []
nb_voxels = nb_voxels ** 0.5 # assume square shape
shape = (nb_voxels, nb_voxels)
mask = np.ones(shape, dtype=int)
graph = graph_from_lattice(mask, kerMask=kerMask2D_4n, toroidal=True)
shape = (1, nb_voxels, nb_voxels)
for c in condition_defs:
tmp = genPotts(graph, beta, 2)
tmp = np.reshape(tmp, shape)
labels.append(tmp)
return np.array(labels)
def test_single_PFPS_ML(self):
"""PF estimation method : path sampling. ML on p(label|beta).
"""
# generate a field:
beta = 0.4
nbClasses = 2
print 'generating potts ..., beta =', beta
labels = genPotts(self.g, beta, nbClasses)
# partition function estimation
gridLnz = Cpt_Vec_Estim_lnZ_Graph(self.g, nbClasses)
gridPace = gridLnz[1][1] - gridLnz[1][0]
# beta estimation
be, pb = beta_estim_obs_field(self.g, labels, gridLnz, 'ML')
print 'betaML:', be
assert abs(be-beta) <= gridPace
def test_beta_estim_jde_mcmc(self):
nbCond = 2
betaMax = 1.2
dBeta = .1
betas = np.arange(0, betaMax, dBeta)
height = 20
width = 20
graph = graph_from_lattice(np.ones((height,width),dtype=int),
kerMask2D_4n)
gridLnZ = cached_eval(Cpt_Vec_Estim_lnZ_Graph, (graph, 2))
nbIt = 3
bolds = []
outputs = []
for beta in betas:
labels = [genPotts(graph, beta).reshape((height,width))
for i in range(nbCond)]
b = simulate_bold_from_labels(labels)
bolds.append(b)
p = {'data':b.to_fmri_data(np.ones((height,width),dtype=int)),
'nbIterations':nbIt,
'betaSimu':beta,
#'noiseVarSimu':noiseVar,
'gridLnZ':gridLnZ}
#outs = cached_eval(runEstimationBetaEstim, (p,))
#outputs.append(outs)
#view_results(bolds, outputs, betas)
scm = "0;0;0.333333;1;0.651282;0;1;1#" \
"0;0;0.333333;1;0.661538;0;1;0#" \
"0;0;0.333333;1;0.664103;1;1;0"
cm = cmstring_to_mpl_cmap(scm)
import plt
for output, beta in zip(outputs,betas):
lmap = output['pm_Labels_marked'].data[0,:,:,0]
plt.figure()
img = plt.matshow(lmap, cmap=cm)
img.get_axes().set_axis_off()
sbeta = ('%1.2f'%beta).replace('.','_')
plt.savefig('./hidden_field_pm_labels_marked_'+sbeta)
def test_obs_field_ML(self):
""" Test estimation of beta with an observed field: a small 2D case.
Partition function estimation method : extrapolation scheme.
Use the ML on p(label|beta). PF estimation: Onsager
"""
shape = (6, 6)
mask = _np.ones(shape, dtype=int) # full mask
g = graph_from_lattice(mask, kerMask=kerMask2D_4n, toroidal=True)
# generate a field:
beta = 0.4
nbClasses = 2
labels = genPotts(g, beta, nbClasses)
# partition function estimation
dbeta = 0.05
betaGrid = _np.arange(0, 1.5, dbeta)
lpf = logpf_ising_onsager(labels.size, betaGrid)
betaML = beta_estim_obs_field(g, labels, (lpf, betaGrid), 'ML')
def test_obs_3Dfield_MAP(self):
""" Test estimation of beta with an observed field: a small 3D case.
Partition function estimation method : extrapolation scheme.
Use the MAP on p(beta|label).
"""
shape = (5, 5, 5)
mask = _np.ones(shape, dtype=int) # full mask
g = graph_from_lattice(mask, kerMask=kerMask3D_6n)
# generate a field:
beta = 0.4
nbClasses = 2
labels = genPotts(g, beta, nbClasses)
# partition function estimation
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses)
gridPace = gridLnz[1][1] - gridLnz[1][0]
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz)
def test_single_Onsager_MAP(self):
"""PF method: Onsager. MAP on p(label|beta).
"""
# generate a field:
beta = 0.4
nbClasses = 2
print 'generating potts ..., beta =', beta
labels = genPotts(self.g, beta, nbClasses)
# partition function estimation
dbeta = 0.05
betaGrid = np.arange(0, 1.5, dbeta)
lpf = logpf_ising_onsager(labels.size, betaGrid)
#lpf, betaGrid = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses)
print 'lpf:'
print lpf
betaMAP = beta_estim_obs_field(self.g, labels, (lpf, betaGrid), 'MAP')
print 'betaMAP:', betaMAP
def test_obs_field_ML(self):
""" Test estimation of beta with an observed field: a small 2D case.
Partition function estimation method : extrapolation scheme.
Use the ML on p(label|beta). PF estimation: Onsager
"""
shape = (6, 6)
mask = _np.ones(shape, dtype=int) # full mask
g = graph_from_lattice(mask, kerMask=kerMask2D_4n, toroidal=True)
# generate a field:
beta = 0.4
nbClasses = 2
labels = genPotts(g, beta, nbClasses)
# partition function estimation
dbeta = 0.05
betaGrid = _np.arange(0, 1.5, dbeta)
lpf = logpf_ising_onsager(labels.size, betaGrid)
betaML = beta_estim_obs_field(g, labels, (lpf, betaGrid), 'ML')
def test_obs_3Dfield_MAP(self):
""" Test estimation of beta with an observed field: a small 3D case.
Partition function estimation method : extrapolation scheme.
Use the MAP on p(beta|label).
"""
shape = (5, 5, 5)
mask = _np.ones(shape, dtype=int) # full mask
g = graph_from_lattice(mask, kerMask=kerMask3D_6n)
# generate a field:
beta = 0.4
nbClasses = 2
labels = genPotts(g, beta, nbClasses)
# partition function estimation
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses)
gridPace = gridLnz[1][1] - gridLnz[1][0]
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz)
def test_obs_2Dfield_MAP(self):
""" Test estimation of beta with an observed 2D field.
Partition function estimation method : extrapolation scheme.
Use the MAP on p(beta|label).
"""
shape = (6, 6)
mask = _np.ones(shape, dtype=int) # full mask
g = graph_from_lattice(mask, kerMask=kerMask2D_4n)
# generate a field:
beta = 0.4
nbClasses = 2
labels = genPotts(g, beta, nbClasses)
# partition function estimation
old_settings = np.seterr('raise')
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses)
gridPace = gridLnz[1][1] - gridLnz[1][0]
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz)
np.seterr(**old_settings)
def test_obs_2Dfield_MAP(self):
""" Test estimation of beta with an observed 2D field.
Partition function estimation method : extrapolation scheme.
Use the MAP on p(beta|label).
"""
shape = (6, 6)
mask = _np.ones(shape, dtype=int) # full mask
g = graph_from_lattice(mask, kerMask=kerMask2D_4n)
# generate a field:
beta = 0.4
nbClasses = 2
labels = genPotts(g, beta, nbClasses)
# partition function estimation
old_settings = np.seterr('raise')
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses)
gridPace = gridLnz[1][1] - gridLnz[1][0]
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz)
np.seterr(**old_settings)
def test_single_PFES_MAP(self):
"""PF estimation method : extrapolation scheme. MAP on p(beta|label).
"""
# generate a field:
beta = 0.4
nbClasses = 2
nbTests = 30
#print 'g:', self.g
# partition function estimation
gridLnz = Cpt_Vec_Estim_lnZ_Graph_fast3(self.g, nbClasses)
gridPace = gridLnz[1][1] - gridLnz[1][0]
#print 'generating potts ..., beta =', beta
bes = np.zeros(nbTests)
for t in xrange(nbTests):
labels = genPotts(self.g, beta, nbClasses)
# beta estimation
be, pb = beta_estim_obs_field(self.g, labels, gridLnz, 'MAP')
print 'betaMAP:', be
bes[t] = be
print 'mean betaMAP:', bes.mean(), bes.std()
assert abs(bes.mean()-beta) <= gridPace
def test_single_surface_PFPS_ML(self):
"""PF estimation method : path sampling. ML on p(label|beta).
topology from a surfacic RDI
"""
# generate a field:
beta = 0.4
nbClasses = 2
print 'generating potts ..., beta =', beta
# grab surfacic data:
from pyhrf.graph import graph_from_mesh, sub_graph, graph_is_sane
from pyhrf.tools._io.tio import Texture
from soma import aims
print 'import done'
roiId = 20
mfn = pyhrf.get_data_file_name('right_hemisphere.mesh')
print 'mesh file:', mfn
mesh = aims.read(mfn)
print 'mesh read'
triangles = [t.arraydata() for t in mesh.polygon().list()]
print 'building graph ... '
wholeGraph = graph_from_mesh(triangles)
roiMaskFile = pyhrf.get_data_file_name('roimask_gyrii_tuned.tex')
roiMask = Texture.read(roiMaskFile).data.astype(int)
mroi = np.where(roiMask==roiId)
g, nm = sub_graph(wholeGraph, mroi[0])
print "g:", len(g), len(g[0])
nnodes = len(g)
points = np.vstack([v.arraydata() for v in mesh.vertex().list()])
weights = [[1./dist(points[j],points[k]) for k in g[j]]
for j in xrange(nnodes)]
print "weights:", len(weights), len(weights[0])
if 1:
for j in xrange(nnodes):
s = sum(weights[j]) * 1.
for k in xrange(len(weights[j])):
weights[j][k] = weights[j][k]/s * len(weights[j])
labels = genPotts(g, beta, nbClasses, weights=weights)
print labels
# partition function estimation
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses,
GraphWeight=weights)
print 'gridLnz with weights:'
print gridLnz
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz, 'ML', weights)
print 'betaML:', be
weights = None
gridLnz = Cpt_Vec_Estim_lnZ_Graph(g, nbClasses,
GraphWeight=weights)
print 'gridLnz without weights:'
print gridLnz
# beta estimation
be, pb = beta_estim_obs_field(g, labels, gridLnz, 'ML', weights)
print 'betaML:', be
gridPace = gridLnz[1][1] - gridLnz[1][0]
assert abs(be-beta) <= gridPace
