def get_pulsar_noise(pta, ret):
ndim = len(pta.params)
# Generate jump groups.
groups0 = [[i, i + 1] for i in range(0, ndim - 1, 2)]
groups0.extend([range(ndim)])
groups1 = [range(ndim)]
# This definiton can ensure a fast convergence (I tried).
# Start the first run. Notice the choice of starting points.
x0 = np.zeros(ndim)
x0[0:-4:2] = 1.
x0[1:-4:2] = -7.
x0[-4:] = [2, -13.5, 2, -13.5]
cov0 = np.diag(np.ones(ndim) * 0.05)
outDir0 = '/home/sdb/xuexiao/M80PTAchains/NoiseFixing/FirstRun/' + pta.pulsars[
0]
sampler = PTSampler(ndim,
pta.get_lnlikelihood,
pta.get_lnprior,
cov0,
groups=groups0,
outDir=outDir0,
verbose=False)
sampler.sample(x0, 250000, isave=1000)
chain0 = np.loadtxt(outDir0 + '/chain_1.txt')
# End of the first run.
# Start the second run.
outDir1 = '/home/sdb/xuexiao/M80PTAchains/NoiseFixing/SecondRun/' + pta.pulsars[
0]
x1 = chain0[np.where(chain0 == np.max(chain0[:, -3]))[0][0], :
-4] # Get the best fit parameters from the last run.
cov1 = np.load(
outDir0 + '/cov.npy'
) # Get the covariance matrix of proposal distribution from the last run.
sampler = PTSampler(ndim,
pta.get_lnlikelihood,
pta.get_lnprior,
cov1,
groups=groups1,
outDir=outDir1,
verbose=False)
sampler.sample(x1, 500000, isave=1000)
chain1 = np.loadtxt(outDir1 + '/chain_1.txt')
# End of the second run.
# Return the ln-likelihood value of the best fit(maximal likelihood).
MLHselect = chain1[np.where(chain1 == np.max(chain1[:, -3]))[0][0], :]
Dict = {pta.params[i].name: MLHselect[i] for i in range(ndim)}
ret.value = (Dict, pta.get_lnlikelihood(Dict), pta.get_lnprior(Dict))
model = tm + dp + wnb + dmn + spn + eph
nparams = [] # Get the number of parameters of each pulsar
signals = []
for psr in psrs:
signal = model(psr)
nparams.append(len(signal.params) -
5) # Subtracting common DPDM params
signals.append(signal)
PTA = signal_base.PTA(signals)
ndim = len(PTA.params) + 5
# Use the best fit noise parameters!
PTA.set_default_params(Dict)
# Set starting points at the already-known best fit points!
xs = {par.name: par.sample() for par in PTA.params}
for parname in Dict.keys():
if parname in xs.keys():
xs[parname] = Dict[parname]
x0 = np.hstack([xs[key] for key in sorted(xs.keys())])
sampler = PTSampler(
ndim,
PTA.get_lnlikelihood,
PTA.get_lnprior,
cov=np.diag(np.ones(ndim) * 0.25),
groups=None,
outDir='/home/sdb/xuexiao/EPHPTAchains/FixNoiseBayesian/')
sampler.sample(x0, 10000000, isave=1000)
# Get Starting Points and constant parameter values.
save1 = np.load('noisepars.npy')
save2 = np.load('noisepardict.npy')
save3 = np.load('dpdmpars-maxposprob.npy')
save4 = np.load('dpdmpardict.npy')
Dict = {save2[i]: save1[i] for i in range(len(save2))}
Dict.update({save4[i]: save3[i] for i in range(len(save4))})
# Use the best fit noise parameters!
PTA.set_default_params(Dict)
# Set starting points at the already-known best fit points!
xs = {par.name: par.sample() for par in PTA.params}
for parname in Dict.keys():
if parname in xs.keys():
xs[parname] = Dict[parname]
x0 = np.hstack([xs[key] for key in sorted(xs.keys())])
# Set groups
groups = [range(ndim), range(ndim - 5, ndim)]
# First sampling (prepare)
sampler = PTSampler(ndim,
PTA.get_lnlikelihood,
PTA.get_lnprior,
cov=np.diag(np.ones(ndim) * 0.025),
groups=groups,
outDir='/home/sdb/xuexiao/PTAchains/openBayesian/')
sampler.sample(x0, 10000000, isave=1000)