def setup_dr(self):
model, options, parameters, data, covariance, __, __, __, __ = gf.setup_mcmc_case_dr(
)
RDR = covariance._RDR
invR = covariance._invR
old_set = ParameterSet(theta=np.random.rand(2),
ss=np.array([10.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5]))
new_set = ParameterSet(theta=np.random.rand(2),
ss=np.array([8.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5]))
priorobj = PriorFunction(
priorfun=model.prior_function,
mu=parameters._thetamu[parameters._parind[:]],
sigma=parameters._thetasigma[parameters._parind[:]])
sosobj = SumOfSquares(model, data, parameters)
DR = DelayedRejection()
DR._initialize_dr_metrics(options=options)
accept, out_set, outbound = DR.run_delayed_rejection(
old_set=old_set,
new_set=new_set,
RDR=RDR,
ntry=2,
parameters=parameters,
invR=invR,
sosobj=sosobj,
priorobj=priorobj)
return accept, out_set, outbound
def test_alphafun(self):
invR = []
invR.append(np.array([[0.4, 0.1], [0., 0.2]]))
invR.append(np.array([[0.4, 0.1], [0., 0.2]]) / 4)
invR.append(np.array([[0.4, 0.1], [0., 0.2]]) / 5)
trypath = []
trypath.append(
ParameterSet(theta=0.1,
ss=np.array([10.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
trypath.append(
ParameterSet(theta=0.2,
ss=np.array([8.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
trypath.append(
ParameterSet(theta=0.2,
ss=np.array([8.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
trypath.append(
ParameterSet(theta=0.2,
ss=np.array([8.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
__, options, __, __ = gf.setup_mcmc()
DR = DelayedRejection()
DR._initialize_dr_metrics(options=options)
alpha = DR.alphafun(trypath=trypath, invR=invR)
self.assertIsInstance(alpha,
np.ndarray,
msg='Expect numpy array return')
self.assertEqual(alpha.size, 1, msg='Expect single element array')
def test_set_based_on_accept_true(self):
next_set = ParameterSet(theta=np.random.random_sample(size=(2, 1)),
ss=0.4)
old_set = ParameterSet(theta=np.random.random_sample(size=(2, 1)),
ss=0.6)
out_set = update_set_based_on_acceptance(accept=True,
old_set=old_set,
next_set=next_set)
self.assertEqual(out_set, next_set)
def setup_trypath(self, iq=0):
trypath = []
trypath.append(ParameterSet(theta=0.1))
trypath.append(ParameterSet(theta=0.2))
trypath.append(ParameterSet(theta=0.3))
trypath.append(ParameterSet(theta=0.4))
stage = len(trypath) - 2
y1, y2, y3, y4 = extract_state_elements(iq=iq,
stage=stage,
trypath=trypath)
return trypath, y1, y2, y3, y4
def test_logposteriorratio_with_2d(self):
trypath = []
trypath.append(
ParameterSet(theta=0.1,
ss=np.array([10.2, 5.1]),
sigma2=np.array([0.5, 0.6]),
prior=np.array([0.5, 0.75])))
trypath.append(
ParameterSet(theta=0.2,
ss=np.array([8.2, 7.5]),
sigma2=np.array([0.5, 1.2]),
prior=np.array([0.5, 0.25])))
zq = log_posterior_ratio(x1=trypath[0], x2=trypath[-1])
self.assertTrue(isinstance(zq, float), msg='Expect float return')
def test_chain_not_accepted_within_bounds(self):
accept = 0
outsidebounds = 0
CL = {
'theta': np.array([1.0, 2.0]),
'ss': 1.0,
'prior': 0.0,
'sigma2': 0.0
}
parset = ParameterSet(theta=CL['theta'],
ss=CL['ss'],
prior=CL['prior'],
sigma2=CL['sigma2'])
mcstat = gf.setup_initialize_chains(CL)
mcstat._MCMC__rejected = {
'total': 10,
'in_adaptation_interval': 4,
'outside_bounds': 1
}
mcstat._MCMC__update_chain(accept=accept,
new_set=parset,
outsidebounds=outsidebounds)
self.assertTrue(np.array_equal(mcstat._MCMC__chain[-1, :],
mcstat._MCMC__old_set.theta),
msg=str('theta added to end of chain - {}'.format(
mcstat._MCMC__chain[-1, :])))
def test_set_outsidebounds(self):
next_set = ParameterSet()
next_set, outbound = set_outside_bounds(next_set = next_set)
self.assertEqual(next_set.alpha, 0, msg = 'next_set.alpha should be 0')
self.assertEqual(next_set.prior, 0, msg = 'next_set.prior should be 0')
self.assertEqual(next_set.ss, np.inf, msg = 'next_set.ss should be np.inf')
self.assertEqual(outbound, 1, msg = 'outbound should be 1')
def test_nth_stage_logpropratio(self):
iq = 0
trypath = []
trypath.append(
ParameterSet(theta=0.1,
ss=np.array([10.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
trypath.append(
ParameterSet(theta=0.2,
ss=np.array([8.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
zq = nth_stage_log_proposal_ratio(iq=iq, trypath=trypath, invR=None)
self.assertTrue(np.array_equal(zq, np.zeros([1])),
msg='Expect arrays to match')
def test_unpack_set(self):
CL = {'theta': 1.0, 'ss': 1.0, 'prior': 0.0, 'sigma2': 0.0}
parset = ParameterSet(theta=CL['theta'],
ss=CL['ss'],
prior=CL['prior'],
sigma2=CL['sigma2'])
MA = Metropolis()
oldpar, ss, oldprior, sigma2 = MA.unpack_set(parset)
NL = {'theta': oldpar, 'ss': ss, 'prior': oldprior, 'sigma2': sigma2}
self.assertDictEqual(CL, NL)
def test_nth_stage_logpropratio_invR(self):
iq = 0
invR = [np.array([[0.4, 0.1], [0., 0.2]])]
trypath = []
trypath.append(
ParameterSet(theta=0.1,
ss=np.array([10.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
trypath.append(
ParameterSet(theta=0.2,
ss=np.array([8.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
trypath.append(
ParameterSet(theta=0.2,
ss=np.array([8.2]),
sigma2=np.array([0.5]),
prior=np.array([0.5])))
zq = nth_stage_log_proposal_ratio(iq=iq, trypath=trypath, invR=invR)
self.assertTrue(isinstance(zq, float), msg='Expect float return')
def setup_rms(cls, CL):
sos_object, prior_object, parameters = gf.setup_mcmc_case_mh()
MS = Metropolis()
R = np.array([[0.4, 0.2], [0, 0.3]])
parset = ParameterSet(theta=CL['theta'],
ss=CL['ss'],
prior=CL['prior'],
sigma2=CL['sigma2'])
accept, _, outbound, npar_sample_from_normal = MS.run_metropolis_step(
old_set=parset,
parameters=parameters,
R=R,
prior_object=prior_object,
sos_object=sos_object)
return accept, outbound
def setup_initialize_chains(CL, updatesigma=True, nsos=1):
mcstat = setup_mcmc_case_cp()
mcstat.simulation_options.updatesigma = updatesigma
mcstat.model_settings.nsos = nsos
mcstat._MCMC__old_set = ParameterSet(theta=CL['theta'],
ss=CL['ss'],
prior=CL['prior'],
sigma2=CL['sigma2'])
mcstat._MCMC__chain_index = mcstat.simulation_options.nsimu - 1
mcstat._MCMC__initialize_chains(
chainind=0,
nsimu=mcstat.simulation_options.nsimu,
npar=mcstat.parameters.npar,
nsos=mcstat.model_settings.nsos,
updatesigma=mcstat.simulation_options.updatesigma,
sigma2=mcstat.model_settings.sigma2)
return mcstat
def test_chain_accepted(self):
accept = 1
outsidebounds = 0
CL = {
'theta': np.array([1.0, 2.0]),
'ss': 1.0,
'prior': 0.0,
'sigma2': 0.0
}
parset = ParameterSet(theta=CL['theta'],
ss=CL['ss'],
prior=CL['prior'],
sigma2=CL['sigma2'])
mcstat = gf.setup_initialize_chains(CL)
mcstat._MCMC__update_chain(accept=accept,
new_set=parset,
outsidebounds=outsidebounds)
self.assertTrue(np.array_equal(mcstat._MCMC__chain[-1, :],
parset.theta),
msg=str('theta added to end of chain - {}'.format(
mcstat._MCMC__chain[-1, :])))
self.assertEqual(mcstat._MCMC__old_set,
parset,
msg='old_set updated to new set')
def test_PS_default_match(self):
PS = ParameterSet()
PSD = PS.__dict__
for (k,v) in PSD.items():
self.assertEqual(v, None, msg = str('Default {} is None'.format(k)))
def test_PS_set_theta(self):
x = 1.2
key = ['theta']
PS = ParameterSet(theta = x)
self.standard_check(key, x, PS)
def test_PS_set_ss(self):
x = 1.2
key = ['ss']
PS = ParameterSet(ss = x)
self.standard_check(key, x, PS)
def test_PS_set_prior(self):
x = 1.2
key = ['prior']
PS = ParameterSet(prior = x)
self.standard_check(key, x, PS)
def test_PS_set_theta_and_prior(self):
x = 1.2
keys = ['theta', 'prior']
PS = ParameterSet(theta = x, prior = x)
self.standard_check(keys, x, PS)
def test_PS_set_alpha(self):
x = 1.2
key = ['alpha']
PS = ParameterSet(alpha = x)
self.standard_check(key, x, PS)
def test_PS_set_sigma2(self):
x = 1.2
key = ['sigma2']
PS = ParameterSet(sigma2 = x)
self.standard_check(key, x, PS)