-
Notifications
You must be signed in to change notification settings - Fork 0
/
tools_lens.py
executable file
·142 lines (104 loc) · 4.59 KB
/
tools_lens.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
# from external
import numpy as np
import healpy as hp
import sys
import pickle
import tqdm
# from ACT modules
from pixell import enmap
# from cmblensplus/wrap/
import curvedsky
# from cmblensplus/utils/
import misctools
import quad_func
# local
import local
import tools_cmb
def load_input_plm(fpalm,lmax,verbose=False):
if verbose: print('load input phi alms') # true phi
# load input phi alms
alm = np.complex128(hp.fitsfunc.read_alm(fpalm))
# convert order of (l,m) to healpix format
alm = curvedsky.utils.lm_healpy2healpix(alm,5100)[:lmax+1,:lmax+1]
# convert to kappa
L = np.linspace(0,lmax,lmax+1)
alm = L[:,None]*(L[:,None]+1)*alm/2.
return alm
def aps(qobj,rlzs,fpalm,wn,verbose=True,mean_sub=True):
# Compute aps of reconstructed lensing map
for q in tqdm.tqdm(qobj.qlist,ncols=100,desc='aps'):
cl = np.zeros((len(rlzs),4,qobj.olmax+1))
W2, W4 = wn[2], wn[4]
for ii, rlz in enumerate(tqdm.tqdm(rlzs,ncols=100,desc='each rlz ('+q+'):')):
# load reconstructed kappa and curl alms with mean-field correction
glm, clm = quad_func.load_rec_alm(qobj,q,rlz,mean_sub=mean_sub)
# load kappa
if rlz != 0:
klm = load_input_plm(fpalm[rlz],qobj.olmax)
else:
klm = 0.*glm
# compute cls
cl[ii,0,:] = curvedsky.utils.alm2cl(qobj.olmax,glm)/W4
cl[ii,1,:] = curvedsky.utils.alm2cl(qobj.olmax,clm)/W4
cl[ii,2,:] = curvedsky.utils.alm2cl(qobj.olmax,glm,klm)/W2
cl[ii,3,:] = curvedsky.utils.alm2cl(qobj.olmax,klm)
np.savetxt(qobj.f[q].cl[rlz],np.concatenate((qobj.l[None,:],cl[ii,:,:])).T)
# save sim mean
if rlzs[1]>=1 and len(rlzs)>1:
np.savetxt(qobj.f[q].mcls,np.concatenate((qobj.l[None,:],np.mean(cl[1:,:,:],axis=0),np.std(cl[1:,:,:],axis=0))).T)
def interface(qid,run=['norm','qrec','n0','mean','aps'],mean_sub=True,kwargs_ov={},kwargs_cmb={},kwargs_qrec={}):
aobj = local.init_analysis_params(qid=qid,**kwargs_cmb)
if qid == 'diff_dn':
aobj_c = local.init_analysis_params(qid='comb_dn',**kwargs_cmb)
else:
# same as aobj
aobj_c = local.init_analysis_params(qid=qid,**kwargs_cmb)
if aobj.fltr == 'cinv':
# filter
ep = 1e-30
# noise
if aobj.qid in local.wqids:
nl = 0.
qids = local.get_subqids(aobj.qid)
for q in qids:
if q in local.boss_d:
bobj = local.init_analysis_params(qid=q,fltr='none',wind='com16',ivar='base')
if q in local.boss_n or q in local.s_16_d:
bobj = local.init_analysis_params(qid=q,fltr='none',wind='base',ivar='base')
nl += 1. / ( np.loadtxt(bobj.fscl['n'],unpack=True)[1] + ep )
nl = 1./(nl+ep)
# load wfactors
print('computing W factor')
W = 1.
for q in qids:
W *= 1. - enmap.to_healpix(tools_cmb.load_survey_mask(q),nside=aobj.nside)
wn = np.zeros(5)
wn[0] = np.mean(1.-W)
wn[:] = wn[0]
print('W factor:',wn[0])
else:
# white
bl = tools_cmb.beam_func(aobj.lmax,aobj.qid)
nl = ( local.qid_wnoise(qid) / bl )**2
wn = tools_cmb.get_wfactors([aobj.qid],aobj.ascale,wind=aobj.wind,ivar=aobj.ivar,ptsr=aobj.ptsr,fltr=aobj.fltr)[aobj.qid]
wn[:] = wn[0]
# corrected factors
cnl = aobj.lcl[0,:] + nl
#wcl = (np.loadtxt(aobj.fscl['c'])).T[1] # wiener-fileterd CMB aps
#ocl, ifl = quad_func.cinv_empirical_fltr(aobj.lcl[0,:],wcl,cnl)
#ocl = np.reshape( aobj.lcl[0,:]**2/(wcl+ep) ,(1,aobj.lmax+1) )
ocl = np.reshape( cnl ,(1,aobj.lmax+1) )
ifl = np.reshape( aobj.lcl[0,:], (1,aobj.lmax+1) )
#wn = tools_cmb.get_wfactors([aobj.qid],1.,wind='base',ptsr=aobj.ptsr,fltr='cinv')[aobj.qid]
else:
# load wfactors
wn = tools_cmb.get_wfactors([aobj.qid],aobj.ascale,wind=aobj.wind,ivar=aobj.ivar,ptsr=aobj.ptsr,fltr=aobj.fltr)[aobj.qid]
# filter
ocl = np.ones((3,aobj_c.lmax+1))
ocl[0,:] = (np.loadtxt(aobj_c.fscl['c'])).T[1]
ifl = ocl.copy()
dirs = local.data_directory()
qobj = quad_func.reconstruction(dirs['local'],aobj.ids,rlz=aobj.rlz,stag=aobj.stag,run=run,wn=wn,lcl=aobj.lcl,ocl=ocl,ifl=ifl,falm=aobj.falm['c'],**kwargs_ov,**kwargs_qrec)
# Aps of reconstructed phi
if 'aps' in run:
aps(qobj,aobj.rlz,aobj.fiklm,wn,mean_sub=mean_sub)