run_dir = rdir + '{}/'.format(ncomps)
mkpath(run_dir)
try:
new_groups = dt.loadGroups(run_dir + 'final/final_groups.npy')
new_meds = np.load(run_dir + 'final/final_med_errs.npy')
new_z = np.load(run_dir + 'final/final_membership.npy')
logging.info("Loaded from previous run")
except IOError:
new_groups, new_meds, new_z =\
em.fit_many_comps(star_pars, ncomps, rdir=run_dir, pool=pool,
bg_dens=BG_DENS,
)
new_groups = np.array(new_groups)
new_lnlike = em.get_overall_lnlikelihood(star_pars, new_groups,
bg_ln_ols=bg_ln_ols)
new_lnpost = em.get_overall_lnlikelihood(star_pars, new_groups,
bg_ln_ols=bg_ln_ols,
inc_posterior=True)
new_BIC = em.calc_bic(star_pars, ncomps, new_lnlike)
# handle multiple components
else:
logging.info("******************************************")
logging.info("********* FITTING {} COMPONENTS *********".\
format(ncomps))
logging.info("******************************************")
best_fits = []
lnlikes = []
lnposts = []
BICs = []
all_meds = []
# final_z_file = rdir + "final/final_membership.npy"
final_z_file = rdir + "memberships.npy"
final_groups_file = rdir + "final_groups.npy"
bg_hists_file = rdir + "bg_hists.npy"
data_file = ddir + assoc_name + "_xyzuvw.fits"
star_pars = gf.loadXYZUVW(data_file)
z = np.load(final_z_file)
# import pdb; pdb.set_trace()
groups = np.load(final_groups_file)
bg_hists = np.load(bg_hists_file)
bg_ln_ols = em.backgroundLogOverlaps(star_pars['xyzuvw'], bg_hists)
overall_lnlike, z = em.get_overall_lnlikelihood(star_pars, groups,
bg_ln_ols, return_memb_probs=True)
print("overall_lnlike with {} comps is: {:.5}".format(ncomps, overall_lnlike))
bic = calcBIC(star_pars, ncomps, overall_lnlike)
print("BIC is: {}".format(bic))
print("With {:.2} stars accounted for by background"\
.format(np.sum(z[:,-1])))
means = {}
means['fitted_then'] = [g.mean for g in groups]
means['fitted_now'] = [
torb.trace_cartesian_orbit(g.mean, g.age, single_age=True)
for g in groups
]
covs = {}
covs['fitted_then'] = np.array([
g.generateSphericalCovMatrix() for g in groups
# final_z_file = rdir + "final/final_membership.npy"
final_z_file = rdir + "memberships.npy"
final_groups_file = rdir + "final_groups.npy"
bg_hists_file = rdir + "bg_hists.npy"
data_file = ddir + assoc_name + "_xyzuvw.fits"
star_pars = gf.loadXYZUVW(data_file)
z = np.load(final_z_file)
# import pdb; pdb.set_trace()
groups = np.load(final_groups_file)
bg_hists = np.load(bg_hists_file)
bg_ln_ols = em.backgroundLogOverlaps(star_pars['xyzuvw'], bg_hists)
overall_lnlike, z = em.get_overall_lnlikelihood(star_pars,
groups,
bg_ln_ols,
return_memb_probs=True)
print("overall_lnlike with {} comps is: {:.5}".format(
ncomps, overall_lnlike))
bic = calcBIC(star_pars, ncomps, overall_lnlike)
print("BIC is: {}".format(bic))
print("With {:.2} stars accounted for by background"\
.format(np.sum(z[:,-1])))
means = {}
means['fitted_then'] = [g.mean for g in groups]
means['fitted_now'] = [
torb.trace_cartesian_orbit(g.mean, g.age, single_age=True)
for g in groups
]
covs = {}
logging.info("******************************************")
run_dir = rdir + '{}/'.format(ncomps)
mkpath(run_dir)
try:
new_groups = dt.loadGroups(run_dir + 'final/final_groups.npy')
new_meds = np.load(run_dir + 'final/final_med_errs.npy')
new_z = np.load(run_dir + 'final/final_membership.npy')
logging.info("Loaded from previous run")
except IOError:
new_groups, new_meds, new_z = \
em.fit_many_comps(star_pars, ncomps, rdir=run_dir, pool=pool,
bg_ln_ols=bg_ln_ols) # kill component here
new_groups = np.array(new_groups)
new_lnlike = em.get_overall_lnlikelihood(star_pars, new_groups,
bg_ln_ols=bg_ln_ols)
new_lnpost = em.get_overall_lnlikelihood(star_pars, new_groups,
bg_ln_ols=bg_ln_ols,
inc_posterior=True)
new_BIC = em.calc_bic(star_pars, ncomps, new_lnlike)
# handle multiple components
else:
logging.info("******************************************")
logging.info("********* FITTING {} COMPONENTS *********".\
format(ncomps))
logging.info("******************************************")
best_fits = []
lnlikes = []
lnposts = []
BICs = []
all_meds = []
trace_orbit_func=trace_orbit_func,
burnin=config.advanced['burnin_steps'],
sampling_steps=config.advanced['sampling_steps'],
use_background=config.config[
'include_background_distribution'],
init_memb_probs=init_memb_probs,
Component=Component,
store_burnin_chains=store_burnin_chains,
max_iters=MAX_ITERS,
)
# Calculate global score of fit for comparison with future fits with different
# component counts
prev_lnlike = expectmax.get_overall_lnlikelihood(
data_dict,
prev_comps,
# bg_ln_ols=bg_ln_ols,
)
prev_lnpost = expectmax.get_overall_lnlikelihood(
data_dict,
prev_comps,
# bg_ln_ols=bg_ln_ols,
inc_posterior=True)
prev_bic = expectmax.calc_bic(data_dict,
ncomps,
prev_lnlike,
memb_probs=prev_memb_probs,
Component=Component)
ncomps += 1
except IOError:
prev_comps, prev_med_and_spans, prev_memb_probs = \
expectmax.fit_many_comps(data=data_dict, ncomps=ncomps, rdir=run_dir,
trace_orbit_func=trace_orbit_func,
burnin=config.advanced['burnin_steps'],
sampling_steps=config.advanced['sampling_steps'],
use_background=config.config[
'include_background_distribution'],
init_memb_probs=init_memb_probs,
)
# Calculate global score of fit for comparison with future fits with different
# component counts
prev_lnlike = expectmax.get_overall_lnlikelihood(data_dict, prev_comps,
# bg_ln_ols=bg_ln_ols,
)
prev_lnpost = expectmax.get_overall_lnlikelihood(data_dict, prev_comps,
# bg_ln_ols=bg_ln_ols,
inc_posterior=True)
prev_bic = expectmax.calc_bic(data_dict, ncomps, prev_lnlike)
ncomps += 1
# Begin iterative loop, each time trialing the incorporation of a new component
while ncomps < MAX_COMPS:
if ncomps >= MAX_COMPS:
log_message(msg='REACHED MAX COMP LIMIT', symbol='+', surround=True)
break
log_message(msg='FITTING {} COMPONENT'.format(ncomps),