mirror of
https://github.com/Richard-Sti/csiborgtools.git
synced 2024-12-22 17:38:02 +00:00
Add SFI groups (#121)
* Update LOS sampling * Bring SD back to speed * Update nb * Update script * Update requirements * Update flow chi2 calc * Update script * Add support for SFI groups * Update plot * Update script * Update script
This commit is contained in:
Richard Stiskalek
GitHub
parent
27c1f9249b
commit
29b37172fa
7 changed files with 253 additions and 135 deletions
|
|
@ -254,7 +254,7 @@ class DataLoader:
|
|||
for key in f.keys():
|
||||
arr[key] = f[key][:]
|
||||
elif catalogue in ["LOSS", "Foundation", "SFI_gals", "2MTF",
|
||||
"Pantheon+"]:
|
||||
"Pantheon+", "SFI_gals_masked", "SFI_groups"]:
|
||||
with File(catalogue_fpath, 'r') as f:
|
||||
grp = f[catalogue]
|
||||
|
||||
|
|
@ -746,10 +746,13 @@ class SD_PV_validation_model(BaseFlowValidationModel):
|
|||
self._r_xrange = r_xrange
|
||||
|
||||
# Get the various vmapped functions
|
||||
self._f_ptilde_wo_bias = lambda mu, err: calculate_ptilde_wo_bias(r_xrange, mu, err, r2_xrange, True) # noqa
|
||||
self._f_simps = lambda y: simps(y, dr) # noqa
|
||||
self._f_zobs = lambda beta, Vr, vpec_rad: predict_zobs(r_xrange, beta, Vr, vpec_rad, Omega_m) # noqa
|
||||
|
||||
self._vmap_ptilde_wo_bias = vmap(lambda mu, err: calculate_ptilde_wo_bias(r_xrange, mu, err, r2_xrange, True)) # noqa
|
||||
self._vmap_simps = vmap(lambda y: simps(y, dr))
|
||||
self._vmap_zobs = vmap(lambda beta, Vr, vpec_rad: predict_zobs(r_xrange, beta, Vr, vpec_rad, Omega_m), in_axes=(None, 0, 0)) # noqa
|
||||
self._vmap_ll_zobs = vmap(lambda zobs, zobs_pred, sigma_v: calculate_ll_zobs(zobs, zobs_pred, sigma_v), in_axes=(0, 0, None)) # noqa
|
||||
|
||||
# Distribution of external velocity components
|
||||
self._Vext = dist.Uniform(-500, 500)
|
||||
|
|
@ -757,12 +760,15 @@ class SD_PV_validation_model(BaseFlowValidationModel):
|
|||
self._alpha = dist.LogNormal(*lognorm_mean_std_to_loc_scale(1.0, 0.5)) # noqa
|
||||
self._beta = dist.Normal(1., 0.5)
|
||||
# Distribution of velocity uncertainty sigma_v
|
||||
self._sv = dist.LogNormal(*lognorm_mean_std_to_loc_scale(150, 100))
|
||||
self._sigma_v = dist.LogNormal(*lognorm_mean_std_to_loc_scale(150, 100)) # noqa
|
||||
self._h = dist.LogNormal(*lognorm_mean_std_to_loc_scale(1., 0.25))
|
||||
|
||||
self._Omega_m = Omega_m
|
||||
|
||||
def predict_zobs_single(self, **kwargs):
|
||||
raise NotImplementedError("This method is not implemented yet.")
|
||||
|
||||
def __call__(self, sample_alpha=False):
|
||||
def __call__(self, sample_alpha=True, sample_beta=True):
|
||||
"""
|
||||
The simple distance NumPyro PV validation model.
|
||||
|
||||
|
|
@ -771,29 +777,39 @@ class SD_PV_validation_model(BaseFlowValidationModel):
|
|||
sample_alpha : bool, optional
|
||||
Whether to sample the density bias parameter `alpha`, otherwise
|
||||
it is fixed to 1.
|
||||
sample_beta : bool, optional
|
||||
Whether to sample the velocity bias parameter `beta`, otherwise
|
||||
it is fixed to 1.
|
||||
"""
|
||||
Vx = numpyro.sample("Vext_x", self._Vext)
|
||||
Vy = numpyro.sample("Vext_y", self._Vext)
|
||||
Vz = numpyro.sample("Vext_z", self._Vext)
|
||||
|
||||
alpha = numpyro.sample("alpha", self._alpha) if sample_alpha else 1.0
|
||||
beta = numpyro.sample("beta", self._beta)
|
||||
sigma_v = numpyro.sample("sigma_v", self._sv)
|
||||
beta = numpyro.sample("beta", self._beta) if sample_beta else 1.0
|
||||
sigma_v = numpyro.sample("sigma_v", self._sigma_v)
|
||||
|
||||
h = numpyro.sample("h", self._h)
|
||||
|
||||
Vext_rad = project_Vext(Vx, Vy, Vz, self._RA, self._dec)
|
||||
|
||||
def scan_body(ll, i):
|
||||
# Calculate p(r) and multiply it by the galaxy bias
|
||||
ptilde = self._vmap_ptilde_wo_bias(self._r_hMpc, self._e2_rhMpc)
|
||||
ptilde *= self._los_density**alpha
|
||||
ptilde = self._f_ptilde_wo_bias(
|
||||
h * self._r_hMpc[i], h**2 * self._e2_rhMpc[i])
|
||||
ptilde *= self._los_density[i]**alpha
|
||||
|
||||
# Normalization of p(r)
|
||||
pnorm = self._vmap_simps(ptilde)
|
||||
pnorm = self._f_simps(ptilde)
|
||||
|
||||
# Calculate p(z_obs) and multiply it by p(r)
|
||||
zobs_pred = self._vmap_zobs(beta, Vext_rad, self._los_velocity)
|
||||
ptilde *= self._vmap_ll_zobs(self._z_obs, zobs_pred, sigma_v)
|
||||
zobs_pred = self._f_zobs(beta, Vext_rad[i], self._los_velocity[i])
|
||||
ptilde *= calculate_ll_zobs(self._z_obs[i], zobs_pred, sigma_v)
|
||||
|
||||
ll = jnp.sum(jnp.log(self._vmap_simps(ptilde) / pnorm))
|
||||
return ll + jnp.log(self._f_simps(ptilde) / pnorm), None
|
||||
|
||||
ll = 0.
|
||||
ll, __ = scan(scan_body, ll, jnp.arange(self.ndata))
|
||||
numpyro.factor("ll", ll)
|
||||
|
||||
|
||||
|
|
@ -1258,7 +1274,7 @@ def get_model(loader, zcmb_max=None, verbose=True):
|
|||
los_overdensity[mask], los_velocity[mask], RA[mask], dec[mask],
|
||||
zCMB[mask], mB[mask], x1[mask], c[mask], e_mB[mask], e_x1[mask],
|
||||
e_c[mask], loader.rdist, loader._Omega_m)
|
||||
elif kind in ["SFI_gals", "2MTF"]:
|
||||
elif kind in ["SFI_gals", "2MTF", "SFI_gals_masked"]:
|
||||
keys = ["RA", "DEC", "z_CMB", "mag", "eta", "e_mag", "e_eta"]
|
||||
RA, dec, zCMB, mag, eta, e_mag, e_eta = (loader.cat[k] for k in keys)
|
||||
|
||||
|
|
@ -1271,6 +1287,15 @@ def get_model(loader, zcmb_max=None, verbose=True):
|
|||
los_overdensity[mask], los_velocity[mask], RA[mask], dec[mask],
|
||||
zCMB[mask], mag[mask], eta[mask], e_mag[mask], e_eta[mask],
|
||||
loader.rdist, loader._Omega_m)
|
||||
elif kind == "SFI_groups":
|
||||
keys = ["RA", "DEC", "z_CMB", "r_hMpc", "e_r_hMpc"]
|
||||
RA, dec, zCMB, r_hMpc, e_r_hMpc = (loader.cat[k] for k in keys)
|
||||
|
||||
mask = (zCMB < zcmb_max)
|
||||
model = SD_PV_validation_model(
|
||||
los_overdensity[mask], los_velocity[mask], RA[mask], dec[mask],
|
||||
zCMB[mask], r_hMpc[mask], e_r_hMpc[mask], loader.rdist,
|
||||
loader._Omega_m)
|
||||
else:
|
||||
raise ValueError(f"Catalogue `{kind}` not recognized.")
|
||||
|
||||
|
|
|
|||
File diff suppressed because one or more lines are too long
|
|
@ -35,8 +35,10 @@ def read_samples(catalogue, simname, ksmooth, include_calibration=False,
|
|||
|
||||
if catalogue in ["LOSS", "Foundation", "Pantheon+"]:
|
||||
alpha_cal, beta_cal, mag_cal, e_mu_intrinsic = [], [], [], []
|
||||
elif catalogue in ["2MTF", "SFI_gals"]:
|
||||
elif catalogue in ["2MTF", "SFI_gals", "SFI_gals_masked"]:
|
||||
a, b, e_mu_intrinsic = [], [], []
|
||||
elif catalogue == "SFI_groups":
|
||||
h = []
|
||||
else:
|
||||
raise ValueError(f"Catalogue {catalogue} not recognized.")
|
||||
|
||||
|
|
@ -86,6 +88,8 @@ def read_samples(catalogue, simname, ksmooth, include_calibration=False,
|
|||
a.append(f[f"sim_{nsim}/a"][:])
|
||||
b.append(f[f"sim_{nsim}/b"][:])
|
||||
e_mu_intrinsic.append(f[f"sim_{nsim}/e_mu_intrinsic"][:])
|
||||
elif catalogue == "SFI_groups":
|
||||
h.append(f[f"sim_{nsim}/h"][:])
|
||||
else:
|
||||
raise ValueError(f"Catalogue {catalogue} not recognized.")
|
||||
|
||||
|
|
@ -95,8 +99,10 @@ def read_samples(catalogue, simname, ksmooth, include_calibration=False,
|
|||
|
||||
if catalogue in ["LOSS", "Foundation", "Pantheon+"]:
|
||||
alpha_cal, beta_cal, mag_cal, e_mu_intrinsic = np.hstack(alpha_cal), np.hstack(beta_cal), np.hstack(mag_cal), np.hstack(e_mu_intrinsic) # noqa
|
||||
elif catalogue in ["2MTF", "SFI_gals"]:
|
||||
elif catalogue in ["2MTF", "SFI_gals", "SFI_gals_masked"]:
|
||||
a, b, e_mu_intrinsic = np.hstack(a), np.hstack(b), np.hstack(e_mu_intrinsic) # noqa
|
||||
elif catalogue == "SFI_groups":
|
||||
h = np.hstack(h)
|
||||
else:
|
||||
raise ValueError(f"Catalogue {catalogue} not recognized.")
|
||||
|
||||
|
|
@ -114,9 +120,12 @@ def read_samples(catalogue, simname, ksmooth, include_calibration=False,
|
|||
if catalogue in ["LOSS", "Foundation", "Pantheon+"]:
|
||||
data += [alpha_cal, beta_cal, mag_cal, e_mu_intrinsic]
|
||||
names += ["alpha_cal", "beta_cal", "mag_cal", "e_mu_intrinsic"]
|
||||
elif catalogue in ["2MTF", "SFI_gals"]:
|
||||
elif catalogue in ["2MTF", "SFI_gals", "SFI_gals_masked"]:
|
||||
data += [a, b, e_mu_intrinsic]
|
||||
names += ["a", "b", "e_mu_intrinsic"]
|
||||
elif catalogue == "SFI_groups":
|
||||
data += [h]
|
||||
names += ["h"]
|
||||
else:
|
||||
raise ValueError(f"Catalogue {catalogue} not recognized.")
|
||||
|
||||
|
|
@ -162,6 +171,7 @@ def names_to_latex(names, for_corner=False):
|
|||
"beta": r"$\beta$",
|
||||
"Vmag": r"$V_{\rm ext}$",
|
||||
"sigma_v": r"$\sigma_v$",
|
||||
"h": r"$h$",
|
||||
}
|
||||
|
||||
labels = copy(names)
|
||||
|
|
|
|||
|
|
@ -58,7 +58,8 @@ def get_los(catalogue_name, simname, comm):
|
|||
if comm.Get_rank() == 0:
|
||||
folder = "/mnt/extraspace/rstiskalek/catalogs"
|
||||
|
||||
if catalogue_name in ["LOSS", "Foundation", "SFI_gals", "2MTF",
|
||||
if catalogue_name in ["LOSS", "Foundation", "SFI_gals",
|
||||
"SFI_gals_masked", "SFI_groups", "2MTF",
|
||||
"Pantheon+"]:
|
||||
fpath = join(folder, "PV_compilation_Supranta2019.hdf5")
|
||||
with File(fpath, 'r') as f:
|
||||
|
|
|
|||
|
|
@ -1,13 +1,13 @@
|
|||
nthreads=3
|
||||
nthreads=5
|
||||
memory=40
|
||||
on_login=${1}
|
||||
on_login=0
|
||||
queue="berg"
|
||||
env="/mnt/users/rstiskalek/csiborgtools/venv_csiborg/bin/python"
|
||||
file="field_los.py"
|
||||
|
||||
simname=${1}
|
||||
catalogue=${2}
|
||||
nsims="-1"
|
||||
simname="csiborg2_main"
|
||||
MAS="SPH"
|
||||
grid=1024
|
||||
|
||||
|
|
|
|||
|
|
@ -52,7 +52,7 @@ def get_model(args, nsim_iterator, get_model_kwargs):
|
|||
if args.catalogue == "A2":
|
||||
fpath = join(folder, "A2.h5")
|
||||
elif args.catalogue in ["LOSS", "Foundation", "Pantheon+", "SFI_gals",
|
||||
"2MTF"]:
|
||||
"2MTF", "SFI_groups", "SFI_gals_masked"]:
|
||||
fpath = join(folder, "PV_compilation_Supranta2019.hdf5")
|
||||
else:
|
||||
raise ValueError(f"Unknown catalogue: `{args.catalogue}`.")
|
||||
|
|
@ -107,10 +107,13 @@ def run_model(model, nsteps, nburn, nchains, nsim, dump_folder,
|
|||
# Calculate the chi2
|
||||
keys = list(thinned_samples.keys())
|
||||
nsamples = len(thinned_samples[keys[0]])
|
||||
try:
|
||||
zobs_mean, zobs_std = model.predict_zobs(thinned_samples)
|
||||
nu = model.ndata - len(keys)
|
||||
chi2 = [np.sum((zobs_mean[:, i] - model._z_obs)**2 / zobs_std[:, i]**2) / nu # noqa
|
||||
for i in range(nsamples)]
|
||||
except NotImplementedError:
|
||||
chi2 = [0. for _ in range(nsamples)]
|
||||
|
||||
gof = csiborgtools.numpyro_gof(model, mcmc, model_kwargs)
|
||||
|
||||
|
|
|
|||
|
|
@ -1,5 +1,5 @@
|
|||
memory=4
|
||||
on_login=1
|
||||
on_login=0
|
||||
nthreads=${1}
|
||||
ksmooth=${2}
|
||||
|
||||
|
|
@ -7,8 +7,8 @@ queue="berg"
|
|||
env="/mnt/users/rstiskalek/csiborgtools/venv_csiborg/bin/python"
|
||||
file="flow_validation.py"
|
||||
|
||||
catalogue="LOSS"
|
||||
simname="Carrick2015"
|
||||
catalogue="SFI_groups"
|
||||
simname="csiborg2_main"
|
||||
|
||||
|
||||
pythoncm="$env $file --catalogue $catalogue --simname $simname --ksmooth $ksmooth"
|
||||
|
|
|
|||
Loading…
Reference in a new issue