def test_hmc():
# load test data
datadir = os.path.join(os.path.dirname(phasemodel.__file__), 'tests',
'testdata')
mdict = np.load(os.path.join(datadir, 'three_phases_v2.npz'))
for var in mdict.files:
globals()[var] = mdict[var]
sz = K_true.shape[0]
# convert coupling from complex 3x3 to real 6x6 matrix
M = kappa2m(K_true)
# some settings
opts = hmc2.opt(nsamples=10**3,
nomit=10**3,
steps=50,
stepadj=.15,
persistence=False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist, np.zeros(sz), opts, en.g_phasedist, M)
data = utils.smod(samps.T)
utils.toc()
# fit test data
K_fit = phasemodel.model.fit_model(data)
print K_true
print K_fit
K_error = (abs(K_true - K_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.1
""" % (K_error)
def test_hmc_biased():
# generate random coupling
dim = 5
K_true = np.random.randn(dim + 1,
dim + 1) + 1j * np.random.randn(dim + 1, dim + 1)
K_true[np.diag(np.ones(dim + 1, bool))] = 0
K_true = .5 * (K_true + np.conj(K_true.T))
# convert coupling from complex 3x3 to real 6x6 matrix
M = kappa2m(K_true)
# some settings
opts = hmc2.opt(nsamples=10**3,
nomit=10**3,
steps=50,
stepadj=.15,
persistence=False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist_biased, np.zeros(dim), opts,
en.g_phasedist_biased, M)
data = utils.smod(samps.T)
utils.toc()
# fit test data
K_fit = phasemodel.model.fit_model_biased(data)
print K_true
print K_fit
K_error = (abs(K_true - K_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.2
""" % (K_error)
def test_hmc():
# load test data
datadir = os.path.join(os.path.dirname(phasemodel.__file__),'tests','testdata')
mdict = np.load(os.path.join(datadir,'three_phases_v2.npz'))
for var in mdict.files:
globals()[var] = mdict[var]
sz = K_true.shape[0]
# convert coupling from complex 3x3 to real 6x6 matrix
M = kappa2m(K_true);
# some settings
opts = hmc2.opt(
nsamples = 10**3,
nomit = 10**3,
steps = 50,
stepadj = .15,
persistence = False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist,np.zeros(sz),opts,en.g_phasedist,M)
data = utils.smod(samps.T)
utils.toc()
# fit test data
K_fit = phasemodel.model.fit_model(data);
print K_true
print K_fit
K_error = (abs(K_true-K_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.1
"""%(K_error)
def test_hmc_biased():
# generate random coupling
dim = 5
K_true = np.random.randn(dim+1,dim+1)+1j*np.random.randn(dim+1,dim+1)
K_true[np.diag(np.ones(dim+1,bool))] = 0
K_true = .5*(K_true+np.conj(K_true.T))
# convert coupling from complex 3x3 to real 6x6 matrix
M = kappa2m(K_true);
# some settings
opts = hmc2.opt(
nsamples = 10**3,
nomit = 10**3,
steps = 50,
stepadj = .15,
persistence = False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist_biased,np.zeros(dim),opts,en.g_phasedist_biased,M)
data = utils.smod(samps.T)
utils.toc()
# fit test data
K_fit = phasemodel.model.fit_model_biased(data);
print K_true
print K_fit
K_error = (abs(K_true-K_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.2
"""%(K_error)
def test_hmc_gen():
# load test data
datadir = os.path.join(os.path.dirname(phasemodel.__file__), 'tests',
'testdata')
mdict = np.load(os.path.join(datadir, 'three_phases_gen_v2.npz'))
for var in mdict.files:
globals()[var] = mdict[var]
sz = M_true.shape[0] / 2
# some settings
opts = hmc2.opt(nsamples=10**3,
nomit=10**3,
steps=50,
stepadj=.15,
persistence=False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist, np.zeros(sz), opts, en.g_phasedist,
M_true)
data = utils.smod(samps.T)
utils.toc()
# fit test data
M_fit = phasemodel.model.fit_gen_model(data)
print M_true
print M_fit
M_error = (abs(M_true - M_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.2
""" % (M_error)
def test_hmc_gen():
# load test data
datadir = os.path.join(os.path.dirname(phasemodel.__file__),'tests','testdata')
mdict = np.load(os.path.join(datadir,'three_phases_gen_v2.npz'))
for var in mdict.files:
globals()[var] = mdict[var]
sz = M_true.shape[0]/2
# some settings
opts = hmc2.opt(
nsamples = 10**3,
nomit = 10**3,
steps = 50,
stepadj = .15,
persistence = False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist,np.zeros(sz),opts,en.g_phasedist,M_true)
data = utils.smod(samps.T)
utils.toc()
# fit test data
M_fit = phasemodel.model.fit_gen_model(data);
print M_true
print M_fit
M_error = (abs(M_true-M_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.2
"""%(M_error)
def test_hmc_biased():
from nose.exc import SkipTest
raise SkipTest("For some reason the generalized model with bias fails...")
# generate random coupling
dim = 5
M_true = np.random.randn(2 * dim + 2, 2 * dim + 2)
M_true = .5 * (M_true + M_true.T)
# M_true[:2,:2] = 0
# M_true[np.diag(np.ones(2*dim+2,bool))] = 0
M_true[:, :2] = 0
M_true[:2, :] = 0
for i in np.arange(M_true.shape[0] / 2):
M_true[2 * i:2 * i + 2, 2 * i:2 * i + 2] = 0
# generate random coupling
dim = 5
K_true = np.random.randn(dim + 1,
dim + 1) + 1j * np.random.randn(dim + 1, dim + 1)
K_true[np.diag(np.ones(dim + 1, bool))] = 0
K_true = .5 * (K_true + np.conj(K_true.T))
# convert coupling from complex 3x3 to real 6x6 matrix
M_true = phasemodel.kappa2m(K_true)
print M_true
#
# anti-symmetrize diagonal elements for estimation matrix
#
# for i in np.arange(2,M_true.shape[0]/2):
# s = M_true[2*i,2*i] + M_true[2*i+1,2*i+1]
# M_true[2*i,2*i] -= s/2
# M_true[2*i+1,2*i+1] -= s/2
# some settings
opts = hmc2.opt(nsamples=10**3,
nomit=10**3,
steps=50,
stepadj=.15,
persistence=False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist_biased, np.zeros(dim), opts,
en.g_phasedist_biased, M_true)
data = utils.smod(samps.T)
utils.toc()
# fit test data
# M_fit = phasemodel.fit_gen_model(data);
M_fit = phasemodel.fit_gen_model_biased(data)
print M_true
print M_fit
M_error = (abs(M_true - M_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.2
""" % (M_error)
def test_hmc_biased():
from nose.exc import SkipTest
raise SkipTest("For some reason the generalized model with bias fails...")
# generate random coupling
dim = 5
M_true = np.random.randn(2*dim+2,2*dim+2)
M_true = .5*(M_true+M_true.T)
# M_true[:2,:2] = 0
# M_true[np.diag(np.ones(2*dim+2,bool))] = 0
M_true[:,:2] = 0
M_true[:2,:] = 0
for i in np.arange(M_true.shape[0]/2):
M_true[2*i:2*i+2,2*i:2*i+2] = 0
# generate random coupling
dim = 5
K_true = np.random.randn(dim+1,dim+1)+1j*np.random.randn(dim+1,dim+1)
K_true[np.diag(np.ones(dim+1,bool))] = 0
K_true = .5*(K_true+np.conj(K_true.T))
# convert coupling from complex 3x3 to real 6x6 matrix
M_true = phasemodel.kappa2m(K_true);
print M_true
#
# anti-symmetrize diagonal elements for estimation matrix
#
# for i in np.arange(2,M_true.shape[0]/2):
# s = M_true[2*i,2*i] + M_true[2*i+1,2*i+1]
# M_true[2*i,2*i] -= s/2
# M_true[2*i+1,2*i+1] -= s/2
# some settings
opts = hmc2.opt(
nsamples = 10**3,
nomit = 10**3,
steps = 50,
stepadj = .15,
persistence = False)
# generate test data
utils.tic()
samps = hmc2.hmc2(en.f_phasedist_biased,np.zeros(dim),opts,en.g_phasedist_biased,M_true)
data = utils.smod(samps.T)
utils.toc()
# fit test data
# M_fit = phasemodel.fit_gen_model(data);
M_fit = phasemodel.fit_gen_model_biased(data);
print M_true
print M_fit
M_error = (abs(M_true-M_fit)).mean()
print """
mean-absolute-difference = %6.8f; expect: 0.2
"""%(M_error)
""" Test script for caching random number generation (for speedup) """ # make sure phasemodel package is in path import sys,os cwd = os.path.abspath(os.path.dirname(__file__)) sys.path.insert(0,os.path.join(cwd,"..","..")) import numpy as np from phasemodel import utils import os os.environ['C_INCLUDE_PATH']=np.get_include() import pyximport; pyximport.install() import random_cache as rand utils.tic() rand.test_random() utils.toc() utils.tic() rand.test_random_cached() utils.toc()
