def classifyByGMMHMM(seq, models, configs):
Y = []
for config in configs:
_rawdata_type = config["logType"]
_event_type = config["eventType"]
_motion_type = config["motionType"]
_sound_type = config["soundType"]
_location_type = config["locationType"]
d = Dataset(
rawdata_type=_rawdata_type,
event_type=_event_type,
motion_type=_motion_type,
sound_type=_sound_type,
location_type=_location_type
)
# Initiation of data need prediction.
y = np.array(d._convetNumericalSequence(seq))
Y.append(y)
_GMMHMMs = []
for model in models:
_GMMs = []
for gmm in model["gmmParams"]["params"]:
_GMM = GMM(
n_components=model["nMix"],
covariance_type=model["covarianceType"]
)
_GMM.covars_ = np.array(gmm["covars"])
_GMM.means_ = np.array(gmm["means"])
_GMM.weights_ = np.array(gmm["weights"])
_GMMs.append(_GMM)
_GMMHMM = GMMHMM(
n_components=model["nComponent"],
n_mix=model["nMix"],
startprob=np.array(model["hmmParams"]["startProb"]),
transmat=np.array(model["hmmParams"]["transMat"]),
gmms=_GMMs,
covariance_type=model["covarianceType"]
)
_GMMHMMs.append(_GMMHMM)
results = []
# for _GMMHMM in _GMMHMMs:
# res = _GMMHMM.score(Y)
# results.append(res)
for i in range(0, len(models)):
res = _GMMHMMs[i].score(Y[i])
results.append(res)
return results
def adaptation(self, ti):
kadp = np.ones(self.nadp)
kadp[0] = STD_RATIO1
kadp[1:3] = STD_RATIO2
kadp[3:5] = STD_RATIO3
sumX = 0
sumX2 = 0
for iadp in range(self.nadp):
nCompo = self.gmdistr.get_params()['n_components']
tmp_gmdistr = GMM(n_components=nCompo, covariance_type='full')
tmp_gmdistr.weights_ = self.gmdistr.weights_
tmp_gmdistr.means_ = self.gmdistr.means_
tmp_gmdistr.covars_ = self.gmdistr.covars_ * kadp[iadp]**2
subSmp = MainSmp(self.nsmp, tmp_gmdistr, self.rv_array,
self.r_array, self.sl_array, self.gd_array,
self.gcov_array)
i = 0
while i <= MAX_NUM_UPDATE:
try:
smps = subSmp.sample()
f = subSmp.priorPDF(smps)
g = 1 - subSmp.condAvailability(smps, ti)
hv, responsibilities = subSmp.score_samples(smps)
gf2hv = g * f / hv
self.updateParameter(smps, responsibilities, gf2hv)
i += 1
break
except ValueError:
i += 1
continue
sumX += np.sum(gf2hv)
sumX2 += np.sum(gf2hv**2)
npre = self.nadp * self.nsmp
pfpre = 1. / npre * sumX
Spf2pre = 1. / (npre * (npre - 1)) * (sumX2 - 2 * sumX * pfpre +
npre * pfpre**2)
return pfpre, Spf2pre
def setPreSmp(self, r_array, sl_array, rate_array, gd_array, gcov_array):
# initial peaks
corrTarget = np.eye(self.ncomp)
rv_list = []
for icomp in xrange(self.ncomp):
cov_variable = Normal('cov', mean=0., stdv=COVT0_COV)
rv_list.append(cov_variable.rv)
rv_array = np.array(rv_list)
peaks0 = getInitialPeaks(rv_array, self.ncomp, corrTarget=corrTarget, options={'disp':False, 'anneal':True})
weights0 = 1./self.ncomp * np.ones(self.ncomp)
covar0 = getInitialCovar(rv_array, self.ncomp)
# initial values
gmdistr = GMM(n_components=self.ncomp , covariance_type='full')
gmdistr.weights_ = weights0
gmdistr.means_ = peaks0
gmdistr.covars_ = covar0
self.preSmp = PreSmp(self.analysis_type, self.nadp, self.nsmp, gmdistr, rv_array, r_array,
sl_array, rate_array, gd_array, gcov_array)
def getKopt(self, k_array, ti):
pf = np.zeros(k_array.shape)
for k in k_array:
nCompo = self.gmdistr.get_params()['n_components']
tmp_gmdistr = GMM(n_components=nCompo,covariance_type='full')
tmp_gmdistr.weights_ = self.gmdistr.weights_
tmp_gmdistr.means_ = self.gmdistr.means_
tmp_gmdistr.covars_ = self.gmdistr.covars_ * k**2
subSmp = MainSmp(self.nsmp, tmp_gmdistr, self.rv_array, self.r_array, self.sl_array, self.gd_array, self.gcov_array)
smps = subSmp.sample()
indx = np.where(k_array==k)[0][0]
pf[indx], dummy = subSmp.getPf(smps, ti)
if indx>0 and pf[indx] < KRATIO_CRITERIA * np.mean(pf[:indx]):
k_indx = indx-1
break
else:
k_indx = -1
return k_array[k_indx]
def getKopt(self, k_array, ti):
pf = np.zeros(k_array.shape)
for k in k_array:
nCompo = self.gmdistr.get_params()['n_components']
tmp_gmdistr = GMM(n_components=nCompo, covariance_type='full')
tmp_gmdistr.weights_ = self.gmdistr.weights_
tmp_gmdistr.means_ = self.gmdistr.means_
tmp_gmdistr.covars_ = self.gmdistr.covars_ * k**2
subSmp = MainSmp(self.nsmp, tmp_gmdistr, self.r_array,
self.sl_array, self.gd_array)
smps = subSmp.sample()
indx = np.where(k_array == k)[0][0]
pf[indx], dummy = subSmp.getPf(smps, ti)
if indx > 0 and pf[indx] < KRATIO_CRITERIA * np.mean(pf[:indx]):
k_indx = indx - 1
break
else:
k_indx = -1
return k_array[k_indx]
def adaptation(self, ti):
kadp = np.ones(self.nadp)
kadp[0] = STD_RATIO1
kadp[1:3] = STD_RATIO2
kadp[3:5] = STD_RATIO3
sumX = 0; sumX2 = 0;
for iadp in range(self.nadp):
nCompo = self.gmdistr.get_params()['n_components']
tmp_gmdistr = GMM(n_components=nCompo,covariance_type='full')
tmp_gmdistr.weights_ = self.gmdistr.weights_
tmp_gmdistr.means_ = self.gmdistr.means_
tmp_gmdistr.covars_ = self.gmdistr.covars_ * kadp[iadp]**2
subSmp = MainSmp(self.nsmp, tmp_gmdistr, self.rv_array, self.r_array, self.sl_array, self.gd_array, self.gcov_array)
i = 0
while i<=MAX_NUM_UPDATE:
try:
smps = subSmp.sample()
f = subSmp.priorPDF(smps)
g = 1 - subSmp.condAvailability(smps, ti)
hv, responsibilities = subSmp.score_samples(smps)
gf2hv = g*f/hv
self.updateParameter(smps, responsibilities, gf2hv)
i += 1
break
except ValueError:
i += 1
continue
sumX += np.sum(gf2hv)
sumX2 += np.sum(gf2hv**2)
npre = self.nadp * self.nsmp
pfpre = 1./npre * sumX
Spf2pre = 1./(npre*(npre-1)) * (sumX2 - 2*sumX*pfpre + npre * pfpre**2)
return pfpre, Spf2pre
def __init__(self, _models):
super(GMMHMMClassifier, self).__init__(_models)
self.gmmhmms = {}
self.predict_data_ = None
for label, value in _models.iteritems():
_model = value['param']
hmm_params = _model['hmmParams']
gmm_params = _model['gmmParams']
n_iter = _model.get('nIter', 50)
transmat = np.array(hmm_params['transMat'])
transmat_prior = np.array(hmm_params['transMatPrior'])
n_component = hmm_params['nComponent']
startprob = np.array(hmm_params['startProb'])
startprob_prior = np.array(hmm_params['startProbPrior'])
n_mix = gmm_params['nMix']
covariance_type = gmm_params['covarianceType']
gmms = gmm_params.get('gmms', None)
gmm_obj_list = []
if not gmms:
gmm_obj_list = None
else:
for gmm in gmms:
gmm_obj = GMM(n_components=gmm['nComponent'], covariance_type=gmm['covarianceType'])
gmm_obj.covars_ = np.array(gmm['covars'])
gmm_obj.means_ = np.array(gmm['means'])
gmm_obj.weights_ = np.array(gmm['weights'])
gmm_obj_list.append(gmm_obj)
gmmhmm = GMMHMM(n_components=n_component, n_mix=n_mix, gmms=gmm_obj_list,
n_iter=n_iter, covariance_type=covariance_type,
transmat=transmat, transmat_prior=transmat_prior,
startprob=startprob, startprob_prior=startprob_prior)
self.gmmhmms[label] = {'gmmhmm': gmmhmm, 'status_set': value['status_set']}
if __name__ == '__main__':
ti = 1
nCompo = 4
nsmp = 400
nadp = 20
nk = 100
muV0 = np.array([[2768.03327821769, 1519.21878505205, 2060.98121494796, 812.166721782312],
[812.166721782312, 2060.98121494796, 1519.21878505205, 2768.03327821769]]).T
muV0 = np.load('peaks0.npy')
piV0 = np.array([0.25, 0.25, 0.25, 0.25])
covarV0 = np.array([[[80300.0, 0.0],[0.0, 80300.0]], [[80300.0, 0.0],[0.0, 80300.0]],
[[80300.0, 0.0],[0.0, 80300.0]], [[80300.0, 0.0],[0.0, 80300.0]]])
gmdistr = GMM(n_components=4, covariance_type='full')
gmdistr.weights_ = piV0
gmdistr.means_ = muV0
gmdistr.covars_ = covarV0
resistRv = Lognormal('r', mean=1790.1, stdv=283.37)
rDistr = resistRv.rv
#r_array = np.array([stats.norm(loc=1790.1, scale=283.37), stats.norm(loc=1790.1, scale=283.37)])
r_array = np.array([rDistr, rDistr])
sl_array = np.array([stats.norm(loc=301.1, scale=120.44)])
ti_mean = 15.91
a = 0.0135; b = 0.8580
gd = lambda x: (1-a*(x-ti_mean)**b) if x>ti_mean else 1.0
gd_array = np.array([gd, gd])
k_array = np.arange(2.0, 1.0-0.05/2, -0.05)
testSmp = PreSmp(nadp, nsmp, gmdistr, r_array, sl_array, gd_array)
pf, Spf2 = testSmp.adaptation(ti)
piVopt = testSmp.gmdistr.weights_
muVopt = testSmp.gmdistr.means_
nk = 100
muV0 = np.array([[
2768.03327821769, 1519.21878505205, 2060.98121494796, 812.166721782312
], [
812.166721782312, 2060.98121494796, 1519.21878505205, 2768.03327821769
]]).T
muV0 = np.load('peaks0.npy')
piV0 = np.array([0.25, 0.25, 0.25, 0.25])
covarV0 = np.array([[[80300.0, 0.0], [0.0, 80300.0]],
[[80300.0, 0.0], [0.0, 80300.0]],
[[80300.0, 0.0], [0.0, 80300.0]],
[[80300.0, 0.0], [0.0, 80300.0]]])
gmdistr = GMM(n_components=4, covariance_type='full')
gmdistr.weights_ = piV0
gmdistr.means_ = muV0
gmdistr.covars_ = covarV0
resistRv = Lognormal('r', mean=1790.1, stdv=283.37)
rDistr = resistRv.rv
#r_array = np.array([stats.norm(loc=1790.1, scale=283.37), stats.norm(loc=1790.1, scale=283.37)])
r_array = np.array([rDistr, rDistr])
sl_array = np.array([stats.norm(loc=301.1, scale=120.44)])
ti_mean = 15.91
a = 0.0135
b = 0.8580
gd = lambda x: (1 - a * (x - ti_mean)**b) if x > ti_mean else 1.0
gd_array = np.array([gd, gd])
k_array = np.arange(2.0, 1.0 - 0.05 / 2, -0.05)
testSmp = PreSmp(nadp, nsmp, gmdistr, r_array, sl_array, gd_array)
pf, Spf2 = testSmp.adaptation(ti)
piVopt = testSmp.gmdistr.weights_
def getPreSmpObject(analysis_type, new_service_time, frp_mean_history, frp_cov_history):
# time array
time_array = np.arange(RELIABILITY_DT,SERVICE_LIFE+RELIABILITY_DT,RELIABILITY_DT)
# resistance array and live load array
flexure_mean_history = frp_mean_history[0]
flexure_cov_history = frp_cov_history[0]
shear_mean_history = frp_mean_history[1]
shear_cov_history = frp_cov_history[1]
rm_mean = flexure_mean_history[0]
rv_mean = shear_mean_history[0]
r_array = np.array([rm_mean, rv_mean])
slmRv = Normal('slm', mean=M_LLIM_MEAN, stdv=M_LLIM_MEAN*M_LLIM_COV)
slvRv = Normal('slv', mean=V_LLIM_MEAN, stdv=V_LLIM_MEAN*V_LLIM_COV)
sldRv = Normal('sld', mean=M_LLIM_DECK_MEAN, stdv=M_LLIM_DECK_MEAN*M_LLIM_DECK_COV)
slmDistr = slmRv.rv
slvDistr = slvRv.rv
sldDistr = sldRv.rv
sl_array = np.array([slmDistr, slvDistr, sldDistr])
rate_m = LL_ARRIVAL_RATE
rate_v = LL_ARRIVAL_RATE
rate_d = LL_ARRIVAL_RATE_DECK
rate_array = np.array([rate_m, rate_v, rate_d])
# function gd(t) and gcov(t)
#def gd_func(x, a, b):
# return a+b*x
def gd_func(x, a, b, c, d, e):
return a+b*x**1+c*x**2+d*x**3+e*x**4
gt_flex = flexure_mean_history/flexure_mean_history[0] / np.sqrt(1+flexure_cov_history**2)
gt_shear = shear_mean_history/shear_mean_history[0] / np.sqrt(1+shear_cov_history**2)
gd_flex_popt, pcov = curve_fit(gd_func, new_service_time, gt_flex)
gd_shear_popt,pcov = curve_fit(gd_func, new_service_time, gt_shear)
gd_flexure = lambda x: gd_func(x, gd_flex_popt[0], gd_flex_popt[1], gd_flex_popt[2], gd_flex_popt[3], gd_flex_popt[4])
gd_shear = lambda x: gd_func(x, gd_shear_popt[0], gd_shear_popt[1], gd_shear_popt[2], gd_shear_popt[3], gd_shear_popt[4])
#gd_flexure = lambda x: gd_func(x, gd_flex_popt[0], gd_flex_popt[1])
#gd_shear = lambda x: gd_func(x, gd_shear_popt[0], gd_shear_popt[1])
def gcov_func(x, a, b): # dummy function, not used
return a+b*x
at_flex = np.sqrt(np.log(flexure_cov_history**2+1)/COVT0_COV**2)
at_shear = np.sqrt(np.log(shear_cov_history**2+1)/COVT0_COV**2)
gcov_flex_popt, pcov = curve_fit(gcov_func, new_service_time, at_flex)
gcov_shear_popt,pcov = curve_fit(gcov_func, new_service_time, at_shear)
#gcov_flexure = lambda x: gcov_func(x, gcov_flex_popt[0], gcov_flex_popt[1], gcov_flex_popt[2], gcov_flex_popt[3], gcov_flex_popt[4])
#gcov_shear = lambda x: gcov_func(x, gcov_shear_popt[0], gcov_shear_popt[1], gcov_shear_popt[2], gcov_shear_popt[3], gcov_shear_popt[4])
gcov_flexure = lambda x: at_flex[0] + 0*x
gcov_shear = lambda x: at_shear[0] + 0*x
gd_array = np.array([gd_flexure, gd_shear])
gcov_array = np.array([gcov_flexure, gcov_shear])
#plt.close('all')
#plt.figure()
#plt.plot(new_service_time, gt_flex, 'bo',
# new_service_time, gt_shear, 'rs',
# new_service_time, gd_flexure(new_service_time), 'b-',
# new_service_time, gd_shear(new_service_time), 'r-')
#plt.figure()
#plt.plot(new_service_time, at_flex, 'bo',
# new_service_time, at_shear, 'rs',
# new_service_time, gcov_flexure(new_service_time), 'b-',
# new_service_time, gcov_shear(new_service_time), 'r-')
#plt.show()
# initial peaks
corrTarget = np.eye(2)
cov_variable1 = Normal('cov1', mean=0., stdv=COVT0_COV)
cov_rv1 = cov_variable1.rv
cov_variable2 = Normal('cov2', mean=0., stdv=COVT0_COV)
cov_rv2 = cov_variable2.rv
rv_array = np.array([cov_rv1, cov_rv2])
peaks0 = getInitialPeaks(rv_array, NUM_COMPONENT, corrTarget=corrTarget, options={'disp':False, 'anneal':True})
weights0 = 1./NUM_COMPONENT * np.ones(NUM_COMPONENT)
covar0 = getInitialCovar(rv_array, NUM_COMPONENT)
# initial values
gmdistr = GMM(n_components=NUM_COMPONENT , covariance_type='full')
gmdistr.weights_ = weights0
gmdistr.means_ = peaks0
gmdistr.covars_ = covar0
# CE-based smpling
nTimePoint = time_array.size
#pfpre = np.zeros(nTimePoint)
#pfmain = np.zeros(nTimePoint)
Spf2pre = np.zeros(nTimePoint)
Spf2main = np.zeros(nTimePoint)
kopt = np.zeros(nTimePoint)
#pf = np.zeros(nTimePoint)
#Spf2 = np.zeros(nTimePoint)
# preliminary sampling
preSmp = PreSmp(analysis_type, NUM_ADAPTATION, NUM_PRE_SMP, gmdistr, rv_array, r_array, sl_array, rate_array, gd_array, gcov_array)
return preSmp
def trainingGMMHMM(
dataset, # training dataset.
n_c, # number of hmm's components (ie. hidden states)
n_m, # number of gmm's mixtures (ie. Gaussian model)
start_prob_prior=None, # prior of start hidden states probabilities.
trans_mat_prior=None, # prior of transition matrix.
start_prob=None, # the start hidden states probabilities.
trans_mat=None, # the transition matrix.
gmms=None, # models' params of gmm
covar_type='full',
n_i=50):
# Initiation of dataset.
# d = Dataset(dataset)
X = dataset.getDataset()
# Initiation of GMM.
_GMMs = []
if gmms is None:
_GMMs = None
else:
for gmm in gmms:
_GMM = GMM(n_components=n_m, covariance_type=covar_type)
_GMM.covars_ = np.array(gmm["covars"])
_GMM.means_ = np.array(gmm["means"])
_GMM.weights_ = np.array(gmm["weights"])
_GMMs.append(_GMM)
# Initiation of GMMHMM.
model = GMMHMM(startprob_prior=np.array(start_prob_prior),
transmat_prior=np.array(trans_mat_prior),
startprob=np.array(start_prob),
transmat=np.array(trans_mat),
gmms=_GMMs,
n_components=n_c,
n_mix=n_m,
covariance_type=covar_type,
n_iter=n_i)
# Training.
model.fit(X)
# The result.
new_gmmhmm = {
"nComponent": n_c,
"nMix": n_m,
"covarianceType": covar_type,
"hmmParams": {
"startProb": model.startprob_.tolist(),
"transMat": model.transmat_.tolist()
},
"gmmParams": {
"nMix": n_m,
"covarianceType": covar_type,
"params": []
}
}
for i in range(0, n_m):
gaussian_model = {
"covars": model.gmms_[i].covars_.tolist(),
"means": model.gmms_[i].means_.tolist(),
"weights": model.gmms_[i].weights_.tolist()
}
new_gmmhmm["gmmParams"]["params"].append(gaussian_model)
return new_gmmhmm