def fix_plots(filename):
fit_instructions = {}
galaxy, model_components = import_spectrum(filename)
# Calculate redshift constraints.
z_low, z_high = model_components["redshift"] - 0.001, model_components["redshift"] + 0.001
fit = pipes.fit(galaxy, fit_instructions, run="r4_exponential_burst")
fit.fit(verbose=False)
fig, ax = fit.plot_sfh_posterior(save=False, show=False)
plt.savefig("pipes/plots/r4_exponential_burst/"+filename+"_sfh.pdf", bbox_inches = 'tight')
fit = pipes.fit(galaxy, fit_instructions, run="r4_dblplaw_burst")
fit.fit(verbose=True)
fig, ax = fit.plot_sfh_posterior(save=False, show=False)
plt.savefig("pipes/plots/r4_dblplaw_burst/"+filename+"_sfh.pdf", bbox_inches = 'tight')
fit = pipes.fit(galaxy, fit_instructions, run="r4_delayed_burst")
fit.fit(verbose=False)
fig, ax = fit.plot_sfh_posterior(save=False, show=False)
plt.savefig("pipes/plots/r4_delayed_burst/"+filename+"_sfh.pdf", bbox_inches = 'tight')
fit = pipes.fit(galaxy, fit_instructions, run="r4_lognormal_burst")
fit.fit(verbose=False)
fig, ax = fit.plot_sfh_posterior(save=False, show=False)
fig.set_size_inches(12,4)
plt.savefig("pipes/plots/r4_lognormal_burst/"+filename+"_sfh.pdf", bbox_inches = 'tight')
def get_best_fit():
for filename in datafiles:
fit_instructions = {}
chi_squ_vals = {}
galaxy, model_components = import_spectrum(filename)
# Calculate redshift constraints.
z_low, z_high = model_components["redshift"] - 0.001, model_components["redshift"] + 0.001
fit = pipes.fit(galaxy, fit_instructions, run="r4_exponential_burst")
fit.fit(verbose=False)
chi_squ_vals = {"r4_exponential_burst" : chi_squared(galaxy, fit)}
fit = pipes.fit(galaxy, fit_instructions, run="r4_dblplaw_burst")
fit.fit(verbose=True)
chi_squ_vals["r4_dblplaw_burst"] = chi_squared(galaxy, fit)
fit = pipes.fit(galaxy, fit_instructions, run="r4_delayed_burst")
fit.fit(verbose=False)
chi_squ_vals["r4_delayed_burst"] = chi_squared(galaxy, fit)
fit = pipes.fit(galaxy, fit_instructions, run="r4_lognormal_burst")
fit.fit(verbose=False)
chi_squ_vals["r4_lognormal_burst"] = chi_squared(galaxy, fit)
# Get the functional form with the lowest chi-squared value.
best_func = min(chi_squ_vals, key=lambda k: chi_squ_vals[k])
# Select the fit with lowest chi-squared value and plot it.
print(filename)
print(best_func)
print(chi_squ_vals[best_func])
def plot_model_posterior_quantities(filename):
fit_instructions = {}
chi_squ_vals = {}
galaxy, model_components = import_spectrum(filename)
# Calculate redshift constraints.
z_low, z_high = model_components["redshift"] - 0.001, model_components["redshift"] + 0.001
fit = pipes.fit(galaxy, fit_instructions, run="r4_exponential_burst")
fit.fit(verbose=False)
chi_squ_vals = {"r4_exponential_burst" : chi_squared(galaxy, fit)}
fit = pipes.fit(galaxy, fit_instructions, run="r4_dblplaw_burst")
fit.fit(verbose=True)
chi_squ_vals["r4_dblplaw_burst"] = chi_squared(galaxy, fit)
fit = pipes.fit(galaxy, fit_instructions, run="r4_delayed_burst")
fit.fit(verbose=False)
chi_squ_vals["r4_delayed_burst"] = chi_squared(galaxy, fit)
fit = pipes.fit(galaxy, fit_instructions, run="r4_lognormal_burst")
fit.fit(verbose=False)
chi_squ_vals["r4_lognormal_burst"] = chi_squared(galaxy, fit)
# Get the functional form with the lowest chi-squared value.
best_func = min(chi_squ_vals, key=lambda k: chi_squ_vals[k])
fit = pipes.fit(galaxy, fit_instructions, run=best_func)
# Select the fit with lowest chi-squared value and plot it.
print(filename)
print(best_func)
for key in chi_squ_vals.keys():
print(key, chi_squ_vals[key])
if best_func == "r4_exponential_burst":
labels = ["exponential2:age", "exponential2:massformed", "exponential1:age", "exponential1:massformed"]
if best_func == "r4_delayed_burst":
labels = ["exponential2:age", "exponential2:massformed", "delayed:age", "delayed:massformed"]
if best_func == "r4_lognormal_burst":
labels = ["exponential2:age", "exponential2:massformed", "lognormal:tmax", "lognormal:massformed"]
if best_func == "r4_dblplaw_burst":
labels = ["exponential2:age", "exponential2:massformed", "dblplaw:tau", "dblplaw:massformed"]
post_quantities = dict(zip(labels, [fit.posterior.samples[l] for l in labels]))
tex_on = True
update_rcParams()
fig, axes = plt.subplots(1, 4)
for i in range(4):
hist1d(post_quantities[labels[i]], axes[i], smooth=True, label=labels[i])
fixed_aspect_ratio(1,axes[i])
fig.set_figwidth(6.75)
plt.savefig("report/img/"+filename+"_posterior.pdf", bbox_inches = 'tight', pad_inches = 0)
def corner():
# from matplotlib import rcParams
# rcParams.update({'figure.autolayout': True})
galaxy, model_components = import_spectrum("phil_model_4")
fit_instructions = {}
fit = pipes.fit(galaxy, fit_instructions, run="r4_exponential_burst")
fit.fit(verbose=False)
plot_corner(fit, names=["exponential2:massformed", "exponential1:tau"], save=True)
def print_posterior_file(filename, run_name):
fit_instructions = {}
chi_squ_vals = {}
galaxy, model_components = import_spectrum(filename)
# Calculate redshift constraints.
z_low, z_high = model_components["redshift"] - 0.001, model_components["redshift"] + 0.001
fit = pipes.fit(galaxy, fit_instructions, run=run_name)
fit.fit(verbose=False)
# chi_squ_vals = {"r4_exponential_burst" : chi_squared(galaxy, fit)}
print_posterior(fit)
def fix_h5_enumeration():
for oldname, newname in zip(oldfiles, datafiles):
for run in runs:
try:
oldpath = "pipes/posterior/" + run + "/" + oldname + ".h5"
newpath = "pipes/posterior/" + run + "/" + newname + ".h5"
os.rename(oldpath, newpath)
except FileNotFoundError:
pass
if os.path.isfile(newpath):
galaxy, model_components = import_spectrum(newname)
fit_instructions = {}
fit = pipes.fit(galaxy, fit_instructions, run=run)
fit.fit(verbose=False)
fit.plot_sfh_posterior()
else:
pass
def fitmodel(ID, saveplot=False):
filt_list = np.loadtxt("filters/goodss_filt_list.txt", dtype="str")
galaxy = pipes.galaxy(ID,
load_goodss,
spectrum_exists=False,
filt_list=filt_list)
exp = {}
exp["age"] = (0.1, 15.)
exp["tau"] = (0.3, 12.)
exp["massformed"] = (2., 15.)
exp["metallicity"] = (0., 2.5)
dust = {}
dust["type"] = "Cardelli"
dust["Av"] = (0., 2.)
fit_instructions = {}
fit_instructions["redshift"] = (0., 10.)
fit_instructions["exponential"] = exp
fit_instructions["dust"] = dust
fit = pipes.fit(galaxy, fit_instructions)
fit.fit(verbose=False)
if saveplot == True:
fig = fit.plot_spectrum_posterior(
save=True, show=False) #####doesnt work, unclear why
fig = fit.plot_sfh_posterior(save=True, show=False)
fig = fit.plot_corner(save=True, show=False)
else:
fig = fit.plot_spectrum_posterior(
save=False, show=True) #####doesnt work, unclear why
fig = fit.plot_sfh_posterior(save=False, show=True)
fig = fit.plot_corner(save=False, show=True)
z_low, z_high = redshift - 0.001, redshift + 0.001
try:
f = open("pipes/posterior/exponential_burst_r4/"+filename+".h5")
f.close()
fit_instructions = {
"redshift" : (z_low, z_high), # z varies tight_layout around z_obs.
"t_bc" : (0.013, 0.021), # Constraints from Murray 2011.
"veldisp" : (50.0, 450.0), # Constrained by Faber-Jackson. TODO: Lookup Minkowski 1962!
"exponential1" : exponential1, # Add the exp SFH component.
"exponential2" : exponential2, # Add the burst component.
"dust" : dust,
"nebular" : nebular
}
# Do a fit with both an old exponential component and recent burst.
fit = pipes.fit(galaxy, fit_instructions, run="exponential_burst_r4")
fit.fit(verbose=False)
# Create a dictionary for storying posterior sample distribution widths.
chi_squ_vals["exponential_burst_r4"] = chi_squared(galaxy, fit)
# Select the fit with lowest chi-ssquared value and plot it.
plt.tight_layout()
fig = fit.plot_sfh_posterior(save=True, show=False)
except IOError:
print("Run not exponential_burst_r4 not accessible for "+filename)
try:
f = open("pipes/posterior/dblplaw_burst_r4/"+filename+".h5")
f.close()
fit_instructions = {
"redshift" : (z_low, z_high), # z varies tight_layout around z_obs.
"t_bc" : (0.013, 0.021), # Constraints from Murray 2011.
def main(segid_to_test):
print("Running BAGPIPES on galaxy SegID:", segid_to_test)
galaxy = pipes.galaxy(str(segid_to_test),
load_data,
photometry_exists=False)
galaxy.plot()
dblplaw = {}
dblplaw["tau"] = (0., 15.)
dblplaw["alpha"] = (0.01, 1000.)
dblplaw["beta"] = (0.01, 1000.)
dblplaw["alpha_prior"] = "log_10"
dblplaw["beta_prior"] = "log_10"
dblplaw["massformed"] = (1., 15.)
dblplaw["metallicity"] = (0.1, 2.)
dblplaw["metallicity_prior"] = "log_10"
nebular = {}
nebular["logU"] = -3.
dust = {}
dust["type"] = "CF00"
dust["eta"] = 2.
dust["Av"] = (0., 2.0)
dust["n"] = (0.3, 2.5)
dust["n_prior"] = "Gaussian"
dust["n_prior_mu"] = 0.7
dust["n_prior_sigma"] = 0.3
fit_instructions = {}
fit_instructions["redshift"] = (0.1, 2.5)
fit_instructions["t_bc"] = 0.01
#fit_instructions["redshift_prior"] = "Gaussian"
#fit_instructions["redshift_prior_mu"] = 0.5
#fit_instructions["redshift_prior_sigma"] = 0.2
fit_instructions["dblplaw"] = dblplaw
fit_instructions["nebular"] = nebular
fit_instructions["dust"] = dust
fit_instructions["veldisp"] = (1., 1000.) #km/s
fit_instructions["veldisp_prior"] = "log_10"
calib = {}
calib["type"] = "polynomial_bayesian"
calib["0"] = (0.5, 1.5)
# Zero order is centred on 1,
# at which point there is no change to the spectrum.
calib["0_prior"] = "Gaussian"
calib["0_prior_mu"] = 1.0
calib["0_prior_sigma"] = 0.25
calib["1"] = (-0.5, 0.5) # Subsequent orders are centred on zero.
calib["1_prior"] = "Gaussian"
calib["1_prior_mu"] = 0.
calib["1_prior_sigma"] = 0.25
calib["2"] = (-0.5, 0.5)
calib["2_prior"] = "Gaussian"
calib["2_prior_mu"] = 0.
calib["2_prior_sigma"] = 0.25
fit_instructions["calib"] = calib
noise = {}
noise["type"] = "white_scaled"
noise["scaling"] = (1., 10.)
noise["scaling_prior"] = "log_10"
fit_instructions["noise"] = noise
fit = pipes.fit(galaxy, fit_instructions, run="spectroscopy")
fit.fit(verbose=False)
fig = fit.plot_spectrum_posterior(save=True, show=True)
fig = fit.plot_calibration(save=True, show=True)
fig = fit.plot_sfh_posterior(save=True, show=True)
fig = fit.plot_corner(save=True, show=True)
print("\a \a \a")
return None
print("Running initial exponential fit for {}...".format(filename)),
# Create (or reset) the fit instructions dictionary.
fit_instructions = {
"redshift": (0.0, 0.1), # Obs. redshift from 0-10.
"t_bc": (0.005, 0.015),
"veldisp": (150.0, 200.0),
"exponential1": exponential1, # Add the exp SFH component.
"exponential2": exponential2, # Add the burst component.
"dust": dust
}
# Do an initial fit with only an exponential compontent.
galaxy, model_components = import_spectrum(filename)
fit = pipes.fit(galaxy, fit_instructions, run="r2_exponential_burst")
fit.fit(verbose=False)
# Create a dictionary for storying posterior sample distribution widths.
chi_squ_vals = {"r2_exponential_burst": chi_squared(galaxy, fit)}
fit_instructions.pop("exponential1")
fit_instructions["dblplaw"] = dblplaw
fit = pipes.fit(galaxy, fit_instructions, run="r2_dblplaw_burst")
fit.fit(verbose=True)
chi_squ_vals["r2_dblplaw_burst"] = chi_squared(galaxy, fit)
fit_instructions.pop("dblplaw", None)
fit_instructions["delayed"] = delayed
fit = pipes.fit(galaxy, fit_instructions, run="r2_delayed_burst")
fit.fit(verbose=False)
for filename in datafiles:
print("Running initial exponential fit for {}...".format(filename)),
# Create (or reset) the fit instructions dictionary.
fit_instructions = {
"redshift": (0.0, 0.1), # Obs. redshift from 0-10.
"t_bc": (0.005, 0.015),
"veldisp": (150.0, 200.0),
"exponential": exponential, # Add the exp SFH component.
"dust": dust
}
# Do an initial fit with only an exponential compontent, over a large parameter space.
galaxy, model_components = import_spectrum(filename)
fit = pipes.fit(galaxy, fit_instructions, run="r1_exponential_noburst")
fit.fit(verbose=False)
# Now run a new fit with other functional components with age constrained to one standard deviation around the exponential age.
age_lower_bound = np.percentile(fit_instructions["exponential"]["age"], 16)
age_upper_bound = np.percentile(fit_instructions["exponential"]["age"], 84)
# Create a dictionary for storying posterior sample distribution widths.
chi_squ_vals = {"exponential": chi_squared(galaxy, fit)}
fit_instructions.pop("exponential", None)
delayed["age"] = (age_lower_bound, age_upper_bound)
fit_instructions["delayed"] = delayed
fit = pipes.fit(galaxy, fit_instructions, run="r1_delayed_noburst")
fit.fit(verbose=False)
chi_squ_vals["delayed"] = chi_squared(galaxy, fit)
from loaddata import load_goodss
from loaddata import load_vandels_spec
import bagpipes as pipes
import numpy as np
filt_list = np.loadtxt("filters/goodss_filt_list.txt", dtype="str")
galaxy = pipes.galaxy(14697,load_goodss,spectrum_exists=False,filt_list=filt_list)
dust = {}
dust["type"] = "Cardelli"
dust["Av"] = (0.,2.)
power = {}
power["tau"] = (0., 15.)
power["alpha"] = (.1, 1000.)
power["beta"] = (.1, 1000.)
power["alpha_prior"] = "log_10"
power["beta_prior"] = "log_10"
power["massformed"] = (2., 15.)
power["metallicity"] = (0., 2.5)
secondfit = {}
secondfit["redshift"] = (0.,10.)
secondfit["dblplaw"] = power
secondfit["dust"] = dust
fit = pipes.fit(galaxy, secondfit, run="dblplaw_sfh")
fit.fit(verbose=False)
fig = fit.plot_sfh_posterior(save=False, show=True)
fig = fit.plot_corner(save=False, show=True)
def fit_various_models(galaxy, redshift, fit_instructions_dict, catalog):
"""
This is the function to fit different models to the
galaxy object. It will create the following directories:
results_dir/
catalog/
results/
And for every run (model) a different directory will be
made to store the results.
FUNCTION:
=========
It will save a .h5 file with all the SED fitting results
information for every galaxy and they are as follows:
(for more info look at Bagpipes)
fit.posterior.get_basic_quantities()
fit.posterior.get_advanced_quantities()
results = fit.results
added_results = fit.posterior.samples
results["samples"] = added_results
results["samples"]["ages"] = fit.posterior.sfh.ages
results["filters_list"] = galaxy.filt_list
results["model_components"] = fit.fitted_model.model_components
INPUT
======
galaxy (Galaxy obj) : galaxy obj from bagpipes
redshift (float): redshift of the Galaxy
fit_instructions_dict (dict): A dictionary of --->
{run (model) : fit_instructions}
catalog (string) : catalog string like GOODSS, GOODSN, UDS, ...
OUTPUT
======
None
"""
# Create the main results directory if not exists
if not os.path.isdir(res):
os.mkdir(res)
results_dir = res + catalog + "/"
if not os.path.isdir(results_dir):
os.mkdir(results_dir)
if not os.path.isdir(results_dir + "results"):
os.mkdir(results_dir + "results")
# Create the error log file if not exists
log_error = results_dir + "results/.no_fitting"
if not os.path.exists(log_error):
with open(log_error, "w+") as f:
f.close()
# Create directories for every models if not exists
for run in fit_instructions_dict:
if not os.path.isdir(results_dir + "results/" + run):
os.mkdir(results_dir + "results/" + run)
# print(run)
for run, fit_info in fit_instructions_dict.items():
run = run + "_" + catalog
try:
fit_info["redshift"] = redshift
fit = pipes.fit(galaxy=galaxy, fit_instructions=fit_info, run=run)
fit.fit(verbose=False)
fit.posterior.get_basic_quantities()
fit.posterior.get_advanced_quantities()
results = fit.results
added_results = fit.posterior.samples
results["samples"] = added_results
results["samples"]["ages"] = fit.posterior.sfh.ages
results["filters_list"] = galaxy.filt_list
results["model_components"] = fit.fitted_model.model_components
# Saving the results of the SED fitting with all of the main
# physical quantities.
path_to_save = results_dir + "results/" + run \
+ "/" + galaxy.ID + ".h5"
if not os.path.exists(path_to_save):
dd.io.save(path_to_save, results)
except:
with open(log_error, "a") as f:
f.write("No fitting for " + run + " " + fit.galaxy.ID)
return None
