def mcmc_CH(self, walkerRatio, n_run, n_burn, mean_start, sigma_start, threadCount=1, init_pos=None, mpi_monch=False):
"""
runs mcmc on the parameter space given parameter bounds with CosmoHammerSampler
returns the chain
"""
lowerLimit, upperLimit = self.cosmoParam.param_bounds
params = np.array([mean_start, lowerLimit, upperLimit, sigma_start]).T
chain = LikelihoodComputationChain(
min=lowerLimit,
max=upperLimit)
temp_dir = tempfile.mkdtemp("Hammer")
file_prefix = os.path.join(temp_dir, "logs")
# chain.addCoreModule(CambCoreModule())
chain.addLikelihoodModule(self.chain)
chain.setup()
store = InMemoryStorageUtil()
if mpi_monch is True:
sampler = MpiCosmoHammerSampler(
params=params,
likelihoodComputationChain=chain,
filePrefix=file_prefix,
walkersRatio=walkerRatio,
burninIterations=n_burn,
sampleIterations=n_run,
threadCount=1,
initPositionGenerator=init_pos,
storageUtil=store)
else:
sampler = CosmoHammerSampler(
params=params,
likelihoodComputationChain=chain,
filePrefix=file_prefix,
walkersRatio=walkerRatio,
burninIterations=n_burn,
sampleIterations=n_run,
threadCount=threadCount,
initPositionGenerator=init_pos,
storageUtil=store)
time_start = time.time()
if sampler.isMaster():
print('Computing the MCMC...')
print('Number of walkers = ', len(mean_start)*walkerRatio)
print('Burn-in itterations: ', n_burn)
print('Sampling itterations:', n_run)
sampler.startSampling()
if sampler.isMaster():
time_end = time.time()
print(time_end - time_start, 'time taken for MCMC sampling')
# if sampler._sampler.pool is not None:
# sampler._sampler.pool.close()
try:
shutil.rmtree(temp_dir)
except Exception as ex:
print(ex)
pass
return store.samples
def test_createChainContext(self):
chain = LikelihoodComputationChain()
p = np.array([1, 2])
ctx = chain.createChainContext(p)
assert ctx is not None
assert np.all(ctx.getParams() == p)
def test_modules(self):
chain = LikelihoodComputationChain()
assert len(chain.getCoreModules()) == 0
assert len(chain.getLikelihoodModules()) == 0
coreModule = DummyModule()
likeModule = DummyModule()
chain.addCoreModule(coreModule)
chain.addLikelihoodModule(likeModule)
assert len(chain.getCoreModules()) == 1
assert len(chain.getLikelihoodModules()) == 1
chain.setup()
assert coreModule.init
assert likeModule.init
like, data = chain([0])
assert coreModule.called
assert likeModule.compLike
assert like == DummyModule.like
assert len(data) == 1
assert data["data"] == DummyModule.data
def test_createChainContext_params(self):
keys = ["a", "b"]
params = Params((keys[0], 0), (keys[1], 1))
chain = LikelihoodComputationChain()
chain.params = params
p = np.array([1, 2])
ctx = chain.createChainContext(p)
assert ctx is not None
assert np.all(ctx.getParams().keys == keys)
assert np.all(ctx.getParams()[0] == p[0])
assert np.all(ctx.getParams()[1] == p[1])
def test_init(self):
self.sampler = CosmoHammerSampler(
params=self.params,
likelihoodComputationChain=LikelihoodComputationChain(),
filePrefix=self._getTempFilePrefix(),
walkersRatio=10,
burninIterations=1,
sampleIterations=1)
assert isinstance(self.sampler.storageUtil, SampleFileUtil)
assert isinstance(self.sampler.stopCriteriaStrategy,
IterationStopCriteriaStrategy)
assert isinstance(self.sampler.initPositionGenerator,
SampleBallPositionGenerator)
assert self.sampler.likelihoodComputationChain.params is not None
def __init__(self, data, nbins, noise=0., div=1.0, like_func='c'):
"""
:param data: load your data
:param nbins: number of k-modes in powerspectrum OR
number of triangle contributions in bispectrum (for covariance matrix)
:param noise: system noise, e.g. SKA, MWA noise response (if any), default 0.0,
:param div: likelihood normalization factor, default 1.0,
:param like_func: choose between complex likelihood function (use 'c'), and normal function (use 'n')
"""
chain = LikelihoodComputationChain(min=params[:, 1], max=params[:, 2])
chain.params = params
if like_func == 'n':
chain.addLikelihoodModule(LikeModule(data, nbins, noise, div))
else:
chain.addLikelihoodModule(ComplexLikeModule(data, nbins, noise))
self.chain = chain
def setup(self):
"""
Initialise some data vectors used for the comparisons by the other functions.
"""
self.params = np.array([[70, 40, 100, 3], [0.0226, 0.005, 0.1, 0.001],
[0.122, 0.01, 0.99, 0.01],
[2.1e-9, 1.48e-9, 5.45e-9, 1e-10],
[0.96, 0.5, 1.5, 0.02],
[0.09, 0.01, 0.8, 0.03], [1, 0, 2, 0.4]])
#the real means..
means = [70.704, 0.02256, 0.1115, 2.18474E-09, 0.9688, 0.08920]
# ...and non-trivial covariance matrix.
cov = np.array(
[[6.11E+00, 0, 0, 0, 0, 0], [7.19E-04, 3.26E-07, 0, 0, 0, 0],
[-1.19E-02, -3.37E-07, 3.14E-05, 0, 0, 0],
[-3.56E-11, 1.43E-14, 1.76E-13, 5.96E-21, 0, 0],
[2.01E-02, 6.37E-06, -2.13E-05, 3.66E-13, 1.90E-04, 0],
[1.10E-02, 2.36E-06, -1.92E-05, 8.70E-13, 7.32E-05, 2.23E-04]])
cov += cov.T - np.diag(cov.diagonal())
# Invert the covariance matrix
icov = np.linalg.inv(cov)
chain = LikelihoodComputationChain()
pseudoLikelihood = PseudoCmbModule(icov, means)
chain.addLikelihoodModule(pseudoLikelihood)
chain.setup()
posGen = FlatPositionGenerator()
self.sampler = CosmoHammerSampler(params=self.params,
likelihoodComputationChain=chain,
filePrefix=self._getTempFilePrefix(),
walkersRatio=10,
burninIterations=1,
sampleIterations=11,
initPositionGenerator=posGen,
storageUtil=InMemoryStorageUtil())
def mcmc_CH(self,
walkerRatio,
n_run,
n_burn,
mean_start,
sigma_start,
lowerLimit,
upperLimit,
X2_chain,
threadCount=1,
init_pos=None):
"""
runs mcmc on the parameter space given parameter bounds with CosmoHammerSampler
returns the chain
"""
params = np.array([mean_start, lowerLimit, upperLimit, sigma_start]).T
chain = LikelihoodComputationChain(min=lowerLimit, max=upperLimit)
# chain.addCoreModule(CambCoreModule())
chain.addLikelihoodModule(X2_chain)
chain.setup()
store = InMemoryStorageUtil()
sampler = CosmoHammerSampler(params=params,
likelihoodComputationChain=chain,
filePrefix="testHammer",
walkersRatio=walkerRatio,
burninIterations=n_burn,
sampleIterations=n_run,
threadCount=threadCount,
initPositionGenerator=init_pos,
storageUtil=store)
sampler.startSampling()
if sampler._sampler.pool is not None:
sampler._sampler.pool.close()
return store.samples
def test_isValid(self):
chain = LikelihoodComputationChain()
assert chain.isValid([0])
chain = LikelihoodComputationChain(min=[0])
assert chain.isValid([1])
assert chain.isValid([0])
assert not chain.isValid([-1])
chain = LikelihoodComputationChain(min=[0, 1])
assert chain.isValid([1, 2])
assert chain.isValid([0, 1])
assert not chain.isValid([-1, 1])
assert not chain.isValid([0, 0])
assert not chain.isValid([-1, 0])
chain = LikelihoodComputationChain(max=[1])
assert chain.isValid([1])
assert chain.isValid([0])
assert not chain.isValid([2])
chain = LikelihoodComputationChain(max=[1, 2])
assert chain.isValid([0, 1])
assert chain.isValid([1, 2])
assert not chain.isValid([2, 2])
assert not chain.isValid([1, 3])
assert not chain.isValid([2, 3])
chain = LikelihoodComputationChain(min=[0, 1], max=[1, 2])
assert chain.isValid([1, 2])
assert chain.isValid([0, 1])
assert chain.isValid([0, 1])
assert chain.isValid([1, 2])
assert not chain.isValid([-1, 1])
assert not chain.isValid([0, 0])
assert not chain.isValid([-1, 0])
assert not chain.isValid([2, 2])
assert not chain.isValid([1, 3])
assert not chain.isValid([2, 3])
like, data = chain([-1, 0])
assert like == -np.inf
assert len(data) == 0
like, data = chain([2, 3])
assert like == -np.inf
assert len(data) == 0
with h5py.File(init, 'r') as hf:
params = np.array(hf.get("params"))
inpos = np.array(hf.get("positions"))
szs = np.array(hf.get("szs_parms")).astype(int)
mins = params[:, 1]
maxs = params[:, 2]
nparam = len(mins)
#--------------------------------------------------------------------------------
#################################################################################################
####################################### LIKELIHOOD CHAIN ########################
#-------------------- Setup the Chain --------
# if mins, and maxs included, the code performs checkin
# and can calculate a null iteration of Generated Quantities in case of rejection
print 'Setting up chain'
chain = LikelihoodComputationChain()
coremodule = CoreModule(szs)
logPosterior = LogPosteriorModule(data, threads=k)
chain.addCoreModule(coremodule)
chain.addLikelihoodModule(logPosterior)
chain.setup()
#######################################################################################
############################### Particle Swarm Optimizer #############################
pso = MPSO(chain,
low=mins,
high=maxs,
particleCount=partCount,
req=req,
threads=1,
InPos=inpos)
smp = emcee.EnsembleSampler(
from cosmoHammer import CosmoHammerSampler
from cosmoHammer.util import InMemoryStorageUtil
from cosmoHammer.util import Params
from cosmoHammer.modules import PseudoCmbModule
from cosmoHammer.pso.ParticleSwarmOptimizer import ParticleSwarmOptimizer
#parameter start center, min, max, start width
params = Params(
("hubble", [70, 65, 80, 3]), ("ombh2", [0.0226, 0.01, 0.03, 0.001]),
("omch2", [0.122, 0.09, 0.2, 0.01]),
("scalar_amp", [2.1e-9, 1.8e-9, 2.35e-9, 1e-10]),
("scalar_spectral_index", [0.96, 0.8, 1.2, 0.02]),
("re_optical_depth", [0.09, 0.01, 0.1, 0.03]), ("sz_amp", [1, 0, 2, 0.4]))
chain = LikelihoodComputationChain(params[:, 1], params[:, 2])
chain.params = params
chain.addLikelihoodModule(PseudoCmbModule())
chain.setup()
# find the best fit value and update our params knowledge
print("find best fit point")
pso = ParticleSwarmOptimizer(chain, params[:, 1], params[:, 2])
psoTrace = np.array([pso.gbest.position.copy() for _ in pso.sample()])
params[:, 0] = pso.gbest.position
storageUtil = InMemoryStorageUtil()
sampler = CosmoHammerSampler(params=params,
likelihoodComputationChain=chain,
filePrefix="pseudoCmb_pso",
def mcmc_CH(self,
walkerRatio,
n_run,
n_burn,
mean_start,
sigma_start,
threadCount=1,
init_pos=None,
mpi=False):
"""
runs mcmc on the parameter space given parameter bounds with CosmoHammerSampler
returns the chain
"""
lowerLimit, upperLimit = self.lower_limit, self.upper_limit
mean_start = np.maximum(lowerLimit, mean_start)
mean_start = np.minimum(upperLimit, mean_start)
low_start = mean_start - sigma_start
high_start = mean_start + sigma_start
low_start = np.maximum(lowerLimit, low_start)
high_start = np.minimum(upperLimit, high_start)
sigma_start = (high_start - low_start) / 2
mean_start = (high_start + low_start) / 2
params = np.array([mean_start, lowerLimit, upperLimit, sigma_start]).T
chain = LikelihoodComputationChain(min=lowerLimit, max=upperLimit)
temp_dir = tempfile.mkdtemp("Hammer")
file_prefix = os.path.join(temp_dir, "logs")
#file_prefix = "./lenstronomy_debug"
# chain.addCoreModule(CambCoreModule())
chain.addLikelihoodModule(self.chain)
chain.setup()
store = InMemoryStorageUtil()
#store = None
if mpi is True:
sampler = MpiCosmoHammerSampler(params=params,
likelihoodComputationChain=chain,
filePrefix=file_prefix,
walkersRatio=walkerRatio,
burninIterations=n_burn,
sampleIterations=n_run,
threadCount=1,
initPositionGenerator=init_pos,
storageUtil=store)
else:
sampler = CosmoHammerSampler(params=params,
likelihoodComputationChain=chain,
filePrefix=file_prefix,
walkersRatio=walkerRatio,
burninIterations=n_burn,
sampleIterations=n_run,
threadCount=threadCount,
initPositionGenerator=init_pos,
storageUtil=store)
time_start = time.time()
if sampler.isMaster():
print('Computing the MCMC...')
print('Number of walkers = ', len(mean_start) * walkerRatio)
print('Burn-in iterations: ', n_burn)
print('Sampling iterations:', n_run)
sampler.startSampling()
if sampler.isMaster():
time_end = time.time()
print(time_end - time_start, 'time taken for MCMC sampling')
# if sampler._sampler.pool is not None:
# sampler._sampler.pool.close()
try:
shutil.rmtree(temp_dir)
except Exception as ex:
print(ex, 'shutil.rmtree did not work')
pass
#samples = np.loadtxt(file_prefix+".out")
#prob = np.loadtxt(file_prefix+"prob.out")
return store.samples, store.prob
else:
hmparams = None
else:
cosmo = hmparams = None
param_mapping = {}
nparams = len(fit_params.keys())
params = np.zeros((nparams, 4))
for key in fit_params.keys():
param_mapping[key] = fit_params[key][0]
params[fit_params[key][0], :] = fit_params[key][1:]
# Set up CosmoHammer
chain = LikelihoodComputationChain(
min=params[:, 1],
max=params[:, 2])
tracers = config['tracers']
trc_combs = []
if config['fit_comb'] == 'all':
logger.info('Fitting auto- and cross-correlations of tracers.')
i = 0
for tr_i in tracers:
for tr_j in tracers[:i + 1]:
# Generate the appropriate list of tracer combinations to plot
trc_combs.append([tr_j, tr_i])
i += 1
elif config['fit_comb'] == 'auto':
logger.info('Fitting auto-correlations of tracers.')