def test_minimize(self):
print('Finding maximum likelihood')
sampler_args = {
"method":'Powell',
"tol":None,
"callback":None,
"options":{'xtol':1E-4,'ftol':1E-4,'maxiter':None,'maxfev':None,'direc':None}
}
rdict=clean_pixels(self.maplike,run_minimize,**sampler_args)
print("Results: ",)
if rdict['ML_success'] :
print(" Successful maximization")
else :
print(" Unsuccessful maximization")
print(" Param truth: ",self.true_params)
print(" Param ML: ",rdict['params_ML'])
print("\n")
def test_emcee(self):
# Calculate the p value and reduced chi squred for the true parameter values
# in the 4 pixels above.
sampler_args = {
"nwalkers": 20,
"nsamps": 500,
"nburn": 100,
"verbose": True
}
samples=clean_pixels(self.maplike,run_emcee,**sampler_args)['chains']
ranges = lambda v: (v[1], v[2]-v[1], v[1]-v[0])
params_mcmc = list(map(ranges, zip(*np.percentile(samples, [16, 50, 84], axis=0))))
print("Results:", )
print("Param medians: ", [p[0] for p in params_mcmc])
print("Param spread, 14th to 84th percentile: ", [p[1] + p[2] for p in params_mcmc])
labels = [r"$\beta_s$", r"$\beta_d$", r"$T_d$", r"$\beta_c$"]
fig = corner.corner(samples, labels=labels, truths=self.true_params)
plt.show()
print("\n")
def test_fisher(self):
print('Fisher sampler')
sampler_args = {
"ml_first": True,
"ml_method":'Powell',
"ml_options":{'xtol':1E-4,'ftol':1E-4,'maxiter':None,'maxfev':None,'direc':None}
}
rdict=clean_pixels(self.maplike,run_fisher,**sampler_args)
covar=np.linalg.inv(rdict['fisher_m'])
pbias=np.dot(covar,rdict['fisher_v'])
print("Results: ",)
if rdict['ML_success'] :
print(" Successful maximization")
else :
print(" Unsuccessful maximization")
print(" Param truth: ",self.true_params)
print(" Param ML: ",rdict['params_cent'])
print(" Param sigma: ",np.sqrt(np.diag(covar)))
print(" Paramb bias: ",pbias)
print("\n")
'var_prior_mean': var_prior_mean,
'var_prior_width': var_prior_width,
'var_prior_type': ['tophat' for b in var_pars]
}, sky_true, instrument)
if W_FWHM:
ml_rw = mpl.MapLike(
{
'data': mps_d_rw[isim][:, ips, :],
'noisevar': mps_nv[:, ips, :]**2,
'fixed_pars': fixed_pars,
'var_pars': var_pars,
'var_prior_mean': var_prior_mean,
'var_prior_width': var_prior_width,
'var_prior_type': ['tophat' for b in var_pars]
}, sky_true, instrument)
rdict = clean_pixels(ml, run_minimize, **sampler_args)
if VERBOSE > 1:
print(" - %d function evaluations" % (rdict['ML_nev']))
if W_FWHM:
amps = ml_rw.get_amplitude_mean(rdict['params_ML']).reshape(
[2, npix_here, 3])
cova = np.linalg.inv(
ml_rw.get_amplitude_covariance(rdict['params_ML'])).reshape(
[2, npix_here, 3, 3])
else:
amps = ml.get_amplitude_mean(rdict['params_ML']).reshape(
[2, npix_here, 3])
cova = np.linalg.inv(
ml.get_amplitude_covariance(rdict['params_ML'])).reshape(
[2, npix_here, 3, 3])
spec_o[isim, 0][ips] = rdict['params_ML'][0]
def clean_maps(k, fn, sdir):
testmap = hp.read_map(f'{fn}maps_sky_signal.fits', field=np.arange(12), verbose=False)
Qs = testmap[::2, sat_mask>0]
Us = testmap[1::2, sat_mask>0]
skymaps = np.array([np.transpose(Qs), np.transpose(Us)])
nu_ref_sync_p=23.
beta_sync_fid=-3.
curv_sync_fid=0.
nu_ref_dust_p=353.
beta_dust_fid=1.5
temp_dust_fid=19.6
spec_i=np.zeros([2, npix]);
spec_o=np.zeros([2, npix]);
amps_o=np.zeros([3, 2, npix]);
cova_o=np.zeros([6, 2, npix]);
bs=beta_sync_fid; bd=beta_dust_fid; td=temp_dust_fid; cs=curv_sync_fid;
sbs=3.0; sbd=3.0;
spec_i[0]=bs; spec_i[1]=bd
fixed_pars={'nu_ref_d':nu_ref_dust_p,'nu_ref_s':nu_ref_sync_p,'T_d':td}
var_pars=['beta_s','beta_d']
var_prior_mean=[bs,bd]
var_prior_width=[sbs,sbd]
sky_true=sky.SkyModel(['syncpl', 'dustmbb', 'unit_response'])
nus = [27., 39., 93., 145., 225., 280.]
bps=np.array([{'nu':np.array([n-0.5,n+0.5]),'bps':np.array([1])} for n in nus])
instrument=ins.InstrumentModel(bps)
ml=mpl.MapLike({'data': skymaps,
'noisevar':np.ones_like(skymaps),
'fixed_pars':fixed_pars,
'var_pars':var_pars,
'var_prior_mean':var_prior_mean,
'var_prior_width':var_prior_width,
'var_prior_type':['tophat' for b in var_pars]},
sky_true,
instrument)
sampler_args = {
"method" : 'Powell',
"tol" : None,
"callback" : None,
"options" : {'xtol':1E-4,'ftol':1E-4,'maxiter':None,'maxfev':None,'direc':None}
}
rdict = clean_pixels(ml, run_minimize, **sampler_args)
Sbar = ml.f_matrix(rdict['params_ML']).T
Sninv = np.linalg.inv(np.dot(Sbar.T, Sbar))
P = np.diag([1., 1., 0.])
Q = np.identity(6) - Sbar.dot(P).dot(Sninv).dot(Sbar.T)
reducedmaps = np.einsum('ab, cdb', Q, skymaps).reshape((12, -1))
filled_maps = np.zeros((12, hp.nside2npix(nside)))
filled_maps[:, sat_mask>0] = reducedmaps
np.savez(f'{sdir}{sname}_hybrid_params_{k}', params=rdict['params_ML'], Sbar=Sbar, Q=Q)
hp.write_map(f'{sdir}{sname}_residualmaps_{k}.fits', filled_maps, overwrite=True)
return