def significance_test(size, bag, samplespace, factors, E_k):
tt = [incexc(i, bag, size) for i in samplespace]
if any([count < 0 for count in tt]):
raise Exception("Value Error: -ve %s" % tt)
n = float(bag[()])
q_ref = [v/n for v in tt]
K = E_k(q_ref)
model = maxentropy.model(factors, samplespace)
model.fit(K, algorithm="CG") # The algorithm can be 'CG', 'BFGS', 'LBFGSB', 'Powell', or 'Nelder-Mead'.
p = model.probdist()
KLdiv = sum([ p_i * (log2(p_i) - log2(q_i)) for p_i, q_i in zip(q_ref, p)])
return (KLdiv, p[-1], q_ref[-1])
def f0(x):
return x in samplespace
def f1(x):
return x == 'dans' or x == 'en'
def f2(x):
return x == 'dans' or x == a_grave
f = [f0, f1, f2]
model = maxentropy.model(f, samplespace)
# Now set the desired feature expectations
K = [1.0, 0.3, 0.5]
model.verbose = True
# Fit the model
model.fit(K)
# Output the distribution
print "\nFitted model parameters are:\n" + str(model.params)
print "\nFitted distribution is:"
p = model.probdist()
for j in range(len(model.samplespace)):
x = model.samplespace[j]
"""
from scipy import maxentropy
import numpy as np
samplespace = [1., 2., 3., 4., 5., 6.]
def sump(x):
return x in samplespace
def meanp(x):
return np.mean(x)
# Set the constraints
# 1) We have a proper probability
# 2) The mean is equal to...
F = [sump, meanp]
model = maxentropy.model(F, samplespace)
# set the desired feature expectations
K = np.ones((5,2))
K[:,1] = [2.,3.,3.5,4.,5.]
model.verbose = False
for i in range(K.shape[0]):
model.fit(K[i])
# Output the distribution
print("\nFitted model parameters are:\n" + str(model.params))
print("\nFitted distribution is:")
p = model.probdist()
for j in range(len(model.samplespace)):
bag[(0,1,2)] = val
tt = [incexc(i, bag, size) for i in samplespace]
if any([count < 0 for count in tt]):
print val, '-ve'
continue
if debug:
for pattern, count in zip(samplespace, tt):
print pattern, count
den = float(bag[()])#float(sum(tt))
q_ref = [v/den for v in tt]
K = E_k(q_ref)
model = maxentropy.model(factors, samplespace)
t1 = time.clock()
model.fit(K, algorithm="CG") # The algorithm can be 'CG', 'BFGS', 'LBFGSB', 'Powell', or 'Nelder-Mead'.
t2 = time.clock()
if debug:
print "Training time = %.4f sec" % (t2 - t1)
print "\nFitted model parameters are:\n" + str(model.params)
p = model.probdist()
K_fit = E_k(p)
KLdiv = sum([ p_i * (log2(p_i) - log2(q_i)) for p_i, q_i in zip(q_ref, p)])
if debug:
print "Comparison of distribution: fit (ref)"
for i in samplespace:
xs, logprobs = sampler.sample(n, return_probs=2)
F = maxentropy.sparsefeaturematrix(f, xs, SPARSEFORMAT)
yield F, logprobs
print "Generating an initial sample ..."
model.setsampleFgen(sampleFgen(sampler, f, n))
model.verbose = True
# Fit the model
model.avegtol = 1e-4
model.fit(K, algorithm=algorithm)
# Output the true distribution
print "\nFitted model parameters are:\n" + str(model.params)
smallmodel = maxentropy.model(f, samplespace)
smallmodel.setparams(model.params)
print "\nFitted distribution is:"
p = smallmodel.probdist()
for j in range(len(smallmodel.samplespace)):
x = smallmodel.samplespace[j]
print ("\tx = %-15s" %(x + ":",) + " p(x) = "+str(p[j])).encode('utf-8')
# Now show how well the constraints are satisfied:
print
print "Desired constraints:"
print "\tp['dans'] + p['en'] = 0.3"
print ("\tp['dans'] + p['" + a_grave + "'] = 0.5").encode('utf-8')
print
print "Actual expectations under the fitted model:"
