def openTets(fn, tets, t0=None, t1=None, lm=None, detectHash=True, chooseSpikes=_motion_):
"""
Open a file with marks and positions
stets if I give
"""
if lm is None:
with open(fn, "rb") as f:
lm = _pkl.load(f)
f.close()
if type(tets) is int:
tets = [tets]
elif type(tets[0]) is str:
inds = []
for i in xrange(len(tets)):
inds.append(lm.tetlist.index(tets[i]))
tets = inds
t0 = 0 if (t0 is None) else t0
t1 = lm.marks.shape[0] if (t1 is None) else t1
if chooseSpikes == _motion_:
minds = lm.minds
elif chooseSpikes == _all_:
minds = _N.array([[t0, t1]])
else:
lminds = []
if lm.minds[0, 0] > 0:
lminds.append([0, lm.minds[0, 0]])
for i in xrange(lm.minds.shape[0] - 1):
lminds.append([lm.minds[i, 1], lm.minds[i+1, 0]])
if lm.minds[-1, 1] > 0:
lminds.append([lm.minds[-1, 1], t1])
minds = _N.array(lminds)
inds = _N.where((minds[:, 0] >= t0) & (minds[:, 0] <= t1))[0]
if detectHash:
allnhmks = []
allhmks = []
else:
allmarks = []
for tet in tets:
marks = []
pos = []
for t in inds:
mksl = lm.marks[minds[t, 0]:minds[t, 1], tet]
shsl = lm.pos[minds[t, 0]:minds[t, 1]]
nn = _N.where(_N.equal(mksl, None) == False)[0]
for n in nn:
for l in xrange(len(mksl[n])):
marks.append(mksl[n][l])
pos.append(shsl[n])
posmarks = _N.empty((len(marks), 5))
posmarks[:, 1:] = _N.array(marks)
posmarks[:, 0] = _N.array(pos)
if detectHash:
# separate the hash
nhid, hid, gmms = _fu.sepHashEM(posmarks)
nhmks = posmarks[nhid]
hmks = posmarks[hid]
allnhmks.append(nhmks)
allhmks.append(hmks)
else:
allmarks.append(posmarks)
if detectHash:
return allnhmks, allhmks, lm, gmms
else:
return allmarks, lm
def init0(self, pos, mk, n1, n2):
"""
M total number of clusters
MS number of clusters assigned to signal
MF number of clusters used for initial fit
M - MF If doing touchup, number of clusters to assign to this
"""
print "init0"
oo = self
k = oo.k
MF = oo.M
###
#
# parameters of cluster covariance. Becomes prior
# priors
oo.PR_cov_PSI = _N.tile(_N.eye(k)*0.5, oo.M).T.reshape(oo.M, k, k)
oo.PR_cov_nu = _N.ones(oo.M, dtype=_N.int)*(k+1)
# parameters of cluster mean. Becomes prior
oo.PR_mu_mu = _N.zeros((oo.M, k))
oo.PR_mu_sg = _N.tile(_N.eye(k)*30, oo.M).T.reshape(oo.M, k, k)
oo.iPR_mu_sg= _N.linalg.inv(oo.PR_mu_sg)
# parameters of cluster weights
oo.PR_m_alp = _N.ones(oo.M) * (1./oo.M)
_x = _N.empty((n2-n1, k))
_x[:, 0] = pos
_x[:, 1:] = mk
N = n2-n1
# Gibbs sampling
################ init cluster centers
##########################
BINS = 20
bins = _N.linspace(-6, 6, BINS+1)
blksz = 20
unonhash, hashsp, gmms = _fu.sepHashEM(_x)
#unonhash, hashsp, hashThr = _fu.sepHash(_x)
if (len(unonhash) > 0) and (len(hashsp) > 0):
labS, labH, clstrs = _fu.emMKPOS_sep(_x[unonhash], _x[hashsp])
elif len(unonhash) == 0:
labS, labH, clstrs = _fu.emMKPOS_sep(None, _x[hashsp])
else:
labS, labH, clstrs = _fu.emMKPOS_sep(_x[unonhash], None)
print "**** clusters sig: %(1)d hash: %(2)d" % {"1" : clstrs[0], "2" : clstrs[1]}
MS = 0
if len(labS) > 0:
MS = _N.max(_N.unique(labS)) + 1
##################
lab = _N.array(labS.tolist() + (labH + clstrs[0]).tolist())
x = _N.empty((n2-n1, k))
strt = 0
if len(unonhash) > 0:
x[0:len(unonhash)] = _x[unonhash]
strt = len(unonhash)
if len(hashsp) > 0:
x[strt:] = _x[hashsp]
# now assign the cluster we've found to Gaussian mixtures
covAll = _N.cov(x.T)
dcovMag= _N.diagonal(covAll)*0.005
print "look at the covariances"
for im in xrange(MF): # if lab < 0, these marks not used for init
kinds = _N.where(lab == im)[0] # inds
if len(kinds) >= 6: # problem when cov is not positive def.
oo.smu[0, im] = _N.mean(x[kinds], axis=0)
oo.scov[0, im] = _N.cov(x[kinds], rowvar=0)
#print "im %(im)d %(c).3e" % {"im" : im, "c" : _N.linalg.det(oo.scov[0, im])}
if _N.linalg.det(oo.scov[0, im]) < 0:
print len(kinds)
print x[kinds]
oo.sm[0, im] = float(len(kinds)+1) / (N+MF)
else:
if len(kinds) > 2:
oo.smu[0, im] = _N.mean(x[kinds], axis=0)
else:
oo.smu[0, im] = _N.mean(x, axis=0)
oo.scov[0, im] = covAll*0.125
oo.sm[0, im] = float(len(kinds)+1) / (N+MF)
def openTets(fn,
tets,
t0=None,
t1=None,
lm=None,
detectHash=True,
chooseSpikes=_motion_):
"""
Open a file with marks and positions
stets if I give
"""
if lm is None:
with open(fn, "rb") as f:
lm = _pkl.load(f)
f.close()
if type(tets) is int:
tets = [tets]
elif type(tets[0]) is str:
inds = []
for i in range(len(tets)):
inds.append(lm.tetlist.index(tets[i]))
tets = inds
t0 = 0 if (t0 is None) else t0
t1 = lm.marks.shape[0] if (t1 is None) else t1
if chooseSpikes == _motion_:
minds = lm.minds
elif chooseSpikes == _all_:
minds = _N.array([[t0, t1]])
else:
lminds = []
if lm.minds[0, 0] > 0:
lminds.append([0, lm.minds[0, 0]])
for i in range(lm.minds.shape[0] - 1):
lminds.append([lm.minds[i, 1], lm.minds[i + 1, 0]])
if lm.minds[-1, 1] > 0:
lminds.append([lm.minds[-1, 1], t1])
minds = _N.array(lminds)
inds = _N.where((minds[:, 0] >= t0) & (minds[:, 0] <= t1))[0]
if detectHash:
allnhmks = []
allhmks = []
else:
allmarks = []
for tet in tets:
marks = []
pos = []
for t in inds:
mksl = lm.marks[minds[t, 0]:minds[t, 1], tet]
shsl = lm.pos[minds[t, 0]:minds[t, 1]]
nn = _N.where(_N.equal(mksl, None) == False)[0]
for n in nn:
for l in range(len(mksl[n])):
marks.append(mksl[n][l])
pos.append(shsl[n])
posmarks = _N.empty((len(marks), 5))
posmarks[:, 1:] = _N.array(marks)
posmarks[:, 0] = _N.array(pos)
if detectHash:
# separate the hash
nhid, hid, gmms = _fu.sepHashEM(posmarks)
nhmks = posmarks[nhid]
hmks = posmarks[hid]
allnhmks.append(nhmks)
allhmks.append(hmks)
else:
allmarks.append(posmarks)
if detectHash:
return allnhmks, allhmks, lm, gmms
else:
return allmarks, lm
