Initial import

scripts/sample_analysis/build_means.py

@@ -0,0 +1,57 @@
+import h5py as h5
+import cosmotool as ct
+import numpy as np
+import ares_tools as at
+
+
+#result = at.read_chain_avg_dev(".",["final_density", "v_field"],step=20*4,prefix="output",do_dev=True,start=20000)
+if False:
+  mean_p,dev_p = at.read_chain_complex_avg_dev(".",lambda ff: (1+ff['final_density'][...])*ff['v_field'][...],step=10,prefix="output",do_dev=True,start=200, pattern="%s_%04d.h5")
+  mean_d,dev_d= at.read_chain_complex_avg_dev(".",lambda ff: ff['final_density'][...],step=10,prefix="output",do_dev=True,start=200,pattern="%s_%04d.h5")
+  np.savez("means.npz", mean_p=mean_p, dev_p=dev_p, mean_d=mean_d,dev_d=dev_d)
+else:
+  x=np.load("means.npz")
+  mean_d = x['mean_d']
+  mean_p = x['mean_p']
+
+with h5.File("output_2200.h5", mode="r") as ff:
+  one_d = ff['final_density'][...]
+  one_p = ff['v_field'][...]#(ff['final_density'][...]+1)*ff['v_field'][...]
+
+Nside=256
+L=4000
+N=256
+Dmin=10
+Dmax=128
+
+#ix=np.arange(N)*L/N - 0.5*L
+ix=np.arange(N)*L/N
+
+# These are the box corners
+corner_x=-2200
+corner_y=-2000
+corner_z=-300
+
+shifter=-np.array([corner_x,corner_y,corner_z])*N/L - 0.5*N
+shifter2=np.array([corner_x,corner_y,corner_z])
+mm=ct.spherical_projection(Nside, mean_d, Dmin, Dmax,shifter=shifter,integrator_id=1,booster=100)
+one_mm=ct.spherical_projection(Nside, one_d, Dmin, Dmax,shifter=shifter,integrator_id=1,booster=100)
+
+x = ix[:,None,None].repeat(N,axis=1).repeat(N,axis=2) + shifter2[0]
+y = ix[None,:,None].repeat(N,axis=0).repeat(N,axis=2) + shifter2[1]
+z = ix[None,None,:].repeat(N,axis=0).repeat(N,axis=1) + shifter2[2]
+r = np.sqrt(x**2+y**2+z**2)
+
+cond = r>0
+pr = np.where(cond, (mean_p[0,...] * x + mean_p[1,...] * y + mean_p[2,...]*z)/r,0)
+one_pr = np.where(cond,(one_p[0,...] * x + one_p[1,...] * y + one_p[2,...]*z)/r,0)
+
+one = np.ones(pr.shape)
+
+mpr=ct.spherical_projection(Nside, pr, Dmin, Dmax,shifter=shifter,integrator_id=1,booster=100)
+one_mpr=ct.spherical_projection(Nside, one_pr, Dmin, Dmax,shifter=shifter,integrator_id=1,booster=100)
+mdist=ct.spherical_projection(Nside, one, Dmin, Dmax,shifter=shifter,integrator_id=1,booster=100)
+
+np.savez("sky.npz", d=mm,pr=mpr,r=r,dist=mdist,one_d=one_mm,one_pr=one_mpr)
+
+

scripts/sample_analysis/show_log_likelihood.py

@@ -0,0 +1,117 @@
+from tqdm import tqdm
+import math
+import os
+import numpy as np
+import matplotlib as mpl
+import matplotlib.cm as cm
+mpl.use('Agg')
+import ares_tools as at
+import matplotlib.pyplot as plt
+import h5py as h5
+import numba as nb
+
+startMC=0
+suffix="ref"
+
+ss = at.analysis(".")
+opts=dict(Nbins=256,range=(0,ss.kmodes.max()))
+
+names=[]
+PP=[]
+Fmax=0
+while True:
+    try:
+      os.stat("mcmc_%d.h5" % Fmax)
+    except:
+      break
+    names.append(Fmax)
+    Fmax += 1
+print(Fmax)
+
+def handle_likelihood():
+
+  def bias_func(rho, bias):
+    a=ne.evaluate('exp(b0*log(rho) - b1*rho**(-b2))', dict(b0=bias[0],b1=bias[1],b2=bias[2],rho=rho))
+    return a 
+  
+  
+  @nb.jit(parallel=True,nopython=True)
+  def compute_likelihood(S,nmean,bias,density,data):
+    N0,N1,N2 = density.shape
+    alpha,r0,eps = bias
+    L = 0
+    for p in nb.prange(N0*N1*N2):
+          k = p % N2
+          j = (p//N2) % N1
+          i = (p//N2//N1)
+          if S[i,j,k] <= 0:
+            continue
+  
+          rho = 1+density[i,j,k]+1e-6
+          x = r0*rho**(-eps)
+          lrho = math.log(rho)
+          rho_g = nmean * rho**alpha * math.exp(-x)
+          log_rho_g = math.log(nmean) + lrho*alpha - x 
+          lam = S[i,j,k] * rho_g
+          log_lam = math.log(S[i,j,k]) + log_rho_g
+          L += data[i,j,k]*log_lam - lam
+  
+    return L
+  
+  try:
+    Likelihood = list(np.load("Like_%s.npy" % suffix))
+    loc_names = names[len(Likelihood):]
+  except:
+    Likelihood = []
+    loc_names = list(names)
+  print(loc_names)
+  if len(loc_names) == 0:
+    return
+  
+  data = []
+  selection = []
+  for d_no in range(16):
+    print("Load data %d" % (d_no,))
+    data.append(ss.get_data(d_no))
+    selection.append(ss.get_mask(d_no))
+  
+  for mc_id in tqdm(loc_names):
+    with h5.File("mcmc_%d.h5" % mc_id, mode="r") as f:
+      density = f['/scalars/BORG_final_density'][...]
+      L=[]
+      for i,(D,S) in enumerate(zip(data,selection)):
+        nmean = f['/scalars/galaxy_nmean_%d' % i][0]
+        bias = f['/scalars/galaxy_bias_%d' % i][...]
+        L.append( compute_likelihood(S, nmean, bias, density, D) )
+      Likelihood.append(L)
+  
+  np.save("Like.npy", Likelihood)
+
+
+def handle_power():
+  Pref = ss.rebin_power_spectrum(startMC, **opts)
+  try:
+    data = np.load("power_%s.npz" % suffix,allow_pickle=True)
+    print("Found previous run")
+    loc_names = names[len(data['P']):]
+    print(loc_names)
+    PP = list(data['P'])
+  except Exception as exc:
+    print(exc)
+    print("no previous run")
+    PP = []
+    loc_names = list(names)
+  print(loc_names)
+  if len(loc_names) == 0:
+    return
+  
+  for i in tqdm(loc_names):
+    PP.append(ss.compute_power_shat_spectrum(i, **opts))
+  
+  bins = 0.5*(Pref[2][1:]+Pref[2][:-1])
+  
+  np.savez("power_%s.npz" % suffix, bins=bins, P=PP, startMC=startMC, Fmax=Fmax, Pref=Pref)
+  
+
+#handle_likelihood()
+handle_power()