def doMCMC(n, nxx, nxy, nyy, x):
# Optional setting for reproducibility
use_seed = False
d = nxx.shape[0]
ns = 2000
if use_seed: # optional setting for reproducibility
seed = 42
# Disable printing
sys.stdout = open(os.devnull, 'w')
# Sufficient statistics
NXX = shared(nxx)
NXY = shared(nxy)
NYY = shared(nyy)
# Define model and perform MCMC sampling
with Model() as model:
# Fixed hyperparameters for priors
b0 = Deterministic('b0', th.zeros((d), dtype='float64'))
ide = Deterministic('ide', th.eye(d, m=d, k=0, dtype='float64'))
# Priors for parameters
l0 = Gamma('l0', alpha=2.0, beta=2.0)
l = Gamma('l', alpha=2.0, beta=2.0)
b = MvNormal('b', mu=b0, tau=l0 * ide, shape=d)
# Custom log likelihood
def logp(xtx, xty, yty):
return (n / 2.0) * th.log(l / (2 * np.pi)) + (-l / 2.0) * (
th.dot(th.dot(b, xtx), b) - 2 * th.dot(b, xty) + yty)
# Likelihood
delta = DensityDist('delta',
logp,
observed={
'xtx': NXX,
'xty': NXY,
'yty': NYY
})
# Inference
print('doMCMC: start NUTS')
step = NUTS()
if use_seed:
trace = sample(ns, step, progressbar=True, random_seed=seed)
else:
trace = sample(ns, step, progressbar=True)
# Enable printing
sys.stdout = sys.__stdout__
# Compute prediction over posterior
return np.mean([np.dot(x, trace['b'][i]) for i in range(ns)], 0)
def fit(self, base_models_predictions, true_targets,
model_identifiers=None):
ba = BayesianAverage()
weight_vector = ba.fit(base_models_predictions, true_targets)
default = True
base_models_predictions = base_models_predictions.transpose()
n_basemodels = base_models_predictions.shape[2]
with Model() as basic_model:
#define prior
HalfNormal('weights', sd=1, shape=n_basemodels)
#define likelihood function
ensemble_pred = np.dot(base_models_predictions, weight_vector)
Categorical('likelihood', p=ensemble_pred.transpose(), observed=true_targets)
with basic_model:
start = find_MAP(model=basic_model)
if not default:
step = Metropolis()
step = NUTS()
trace = sample(self.n_samples, step=step, start=start)
trace = trace[5000:]
self.sampled_weights = trace["weights"]
map_estimate = find_MAP(model=basic_model, fmin=optimize.fmin_powell)
print(map_estimate)
from pymc3 import NUTS, sample
from pymc3 import traceplot
with basic_model:
# obtain starting values via MAP
start = find_MAP(fmin=optimize.fmin_powell)
# instantiate sampler
step = NUTS(scaling=start)
# draw 2000 posterior samples
trace = sample(2000, step, start=start)
trace['alpha'][-5:]
traceplot(trace)
plt.show()
from pymc3 import summary
summary(trace)
n = 500
p = 0.3
def gaith(datagaith):
# datagaith=pd.DataFrame(datagaith)
data = datagaith.iloc[:-1, 0].values
# print(data)
# Fit a normal distribution to the data:
mu, std = norm.fit(data)
# Plot the histogram.
plt.hist(data, bins=25, normed=True, alpha=0.6, color='g')
# Plot the PDF.
xmin, xmax = plt.xlim()
x = np.linspace(xmin, xmax, 100)
pdf = norm.pdf(x, mu, std)
# plt.plot(x, pdf, 'k', linewidth=2)
# title = "Fit results: mu = %.2f, std = %.4f" % (mu, std)
# plt.title(title)
# plt.show()
# prior
prior = datagaith.iloc[-1, 0]
stdprior = 0.002
x = np.linspace(prior - .01, prior + .01, 100)
p = norm.pdf(x, prior, stdprior)
# plt.plot(x, p, 'k', linewidth=2)
# title = "Fit results: mu = %.2f, std = %.4f" % (prior, stdprior)
# plt.title(title)
# plt.show()
# with pm.Model():
# mu1 =pdf
#
# niter = 20
# with pm.Model():
# mu = mu
# sd = std
# y = pm.Normal('y', mu=mu, sd=sd, observed=data)
# start = pm.find_MAP(fmin=optimize.fmin_powell)
# print("************************")
# print(start)
# step = pm.NUTS(scaling=start)
# trace = pm.sample(niter, start=start, step=step)
model = pm.Model()
with model:
mu1 = pm.Normal("mu1", mu=mu, sd=std, shape=10)
with model:
# step = pm.NUTS()
# trace = pm.sample(2000, tune=1000, init=None, step=step, cores=2)
# obtain starting values via MAP
start = pm.find_MAP(fmin=optimize.fmin_powell)
# instantiate sampler
step = NUTS(scaling=start)
# draw 2000 posterior samples
trace = sample(10, step, start=start)
# for i in summary(trace).iterrows():
# print(str(i))
# if (i[1]) == (min(summary(trace)['mc_error'])):
# print(i)
a = summary(trace).loc[summary(trace)['mc_error'] == min(
summary(trace)['mc_error'])]
return a, mu, prior
'Q_ratematrixoneway': Q_raw_log,
'B_logodds': B_lo,
'B0_logodds': B0_lo,
'S': S_start,
'X': X_start,
'Z_logodds': Z_lo,
'L_logodds': L_lo
}
#teststart = {'Q_ratematrixoneway': Q_raw_log, 'B_logodds':B_lo, 'B0_logodds':B0_lo, 'S':S_start, 'X':X_start, 'Z_logodds':Z_lo, 'L_logodds':L_lo, 'pi_stickbreaking':np.ones(M)/float(M)}
#start = {'Q_ratematrixoneway': Q_raw_log, 'B_logodds':B_lo, 'B0_logodds':B0_lo, 'S':S_start, 'X':X_start, 'Z_logodds':Z_lo, 'L_logodds':L_start}
with model:
#import pdb; pdb.set_trace()
steps = []
steps.append(NUTS(vars=[pi]))
#steps.append(NUTS(vars=[pi], scaling=np.ones(M-1)*0.058))
#steps.append(Metropolis(vars=[pi], scaling=0.058, tune=False))
steps.append(NUTS(vars=[Q], scaling=np.ones(M - 1, dtype=float) * 10.))
#steps.append(Metropolis(vars=[Q], scaling=0.2, tune=False))
steps.append(
ForwardS(vars=[S], nObs=nObs, T=T, N=N, observed_jumps=obs_jumps))
steps.append(NUTS(vars=[B0, B]))
#steps.append(Metropolis(vars=[B0], scaling=0.2, tune=False))
#steps.append(NUTS(vars=[B]))
#steps.append(Metropolis(vars=[B], scaling=0.198, tune=False))
steps.append(ForwardX(vars=[X], N=N, T=T, K=K, D=D, Dd=Dd, O=O, nObs=nObs))
#steps.append(NUTS(vars=[Z], scaling=np.ones(K*D)))
steps.append(Metropolis(vars=[Z], scaling=0.0132, tune=False))
steps.append(NUTS(vars=[L], scaling=np.ones(D)))
#steps.append(Metropolis(vars=[L],scaling=0.02, tune=False, ))
plt.ylabel("Disaster count")
plt.xlabel("Year")
plt.show()
from pymc3 import DiscreteUniform, Poisson, switch, Model, Exponential, NUTS, Metropolis, sample, traceplot
with Model() as disaster_model:
switchpoint = DiscreteUniform('switchpoint',
lower=year.min(),
upper=year.max(),
testval=1900)
# Priors for pre- and post-switch rates number of disasters
early_rate = Exponential('early_rate', 1)
late_rate = Exponential('late_rate', 1)
# Allocate appropriate Poisson rates to years before and after current
rate = switch(switchpoint >= year, early_rate, late_rate)
disasters = Poisson('disasters', rate, observed=disaster_data)
step1 = NUTS([early_rate, late_rate])
# Use Metropolis for switchpoint, and missing values since it accommodates discrete variables
step2 = Metropolis([switchpoint, disasters.missing_values[0]])
trace = sample(10000, step=[step1, step2])
traceplot(trace)
sd = StudentT('sd', 10, 2, 5**-2)
vals = Normal('vals', p, sd=sd, observed=rate)
# <markdowncell>
# Model Fitting
# -------------
# <codecell>
with model:
s = find_MAP(vars=[sd, y])
step = NUTS(scaling=s)
trace = sample(100, step, s)
s = trace[-1]
step = NUTS(scaling=s)
def run(n=3000):
if n == "short":
n = 150
with model:
trace = sample(n, step, s)
# <codecell>
for i, country in enumerate(countries):
from pymc3 import traceplot
traceplot(trace)
plt.show()
'''
NUTS Sampler
'''
from pymc3 import NUTS, sample
with basic_model:
# Use starting ML point
start = MLpoint
hess = hessian(useToAs)
#Set scaling using hessian
step = NUTS()
# draw 2000 posterior samples
trace = sample(2000, start=start)
from pymc3 import traceplot
traceplot(trace)
plt.show()
accept = np.float64(np.sum(trace['phase'][1:] != trace['phase'][:-1]))
print "Acceptance Rate: ", accept / trace['phase'].shape[0]
'B_logodds': B_lo,
'B0_logodds': B0_lo,
'S': S_start,
'X': X_start,
'Z_anchoredbeta': Z_lo,
'L_logodds': L_lo
}
#start = {'Q_ratematrixoneway': Q_raw_log, 'B_logodds':B_lo, 'B0_logodds':B0_lo, 'S':S_start, 'X':X_start, 'Z_logodds':Z_lo, 'L_logodds':L_lo}
with model:
steps = []
if 'pi' in args.constantVars:
steps.append(Constant(vars=[pi]))
else:
steps.append(NUTS(vars=[pi]))
#steps.append(Metropolis(vars=[pi], scaling=0.058, tune=False))
if 'Q' in args.constantVars:
steps.append(Constant(vars=[Q]))
else:
steps.append(NUTS(
vars=[Q], scaling=np.ones(M - 1, dtype=float) *
10.)) #steps.append(Metropolis(vars=[Q], scaling=0.2, tune=False))
if 'S' in args.constantVars:
steps.append(Constant(vars=[S]))
else:
steps.append(
ForwardS(vars=[S], nObs=nObs, T=T, N=N, observed_jumps=obs_jumps))
if 'B0' in args.constantVars:
steps.append(Constant(vars=[B0]))
if 'B' in args.constantVars:
