def run_fit(self, plot_it=False, burnin_steps=1000, sampling_steps=1000):
self.best_like_fit, chain, self.lnprob =\
gf.fit_group(
self.tb_file, fixed_age=self.fixed_age, burnin_steps=1000,
sampling_steps=1000
)
self.nwalkers, self.nsteps, self.npars = chain.shape
flat_chain = np.reshape(chain, (-1, self.npars))
self.nsamples = self.nwalkers*self.nsteps
logging.info("Fixed fit completed for {}".format(self.fixed_age))
cov_mats = self.calc_covs(flat_chain)
self.calc_pos_radii(cov_mats)
self.calc_mean_fit(cov_mats, flat_chain)
def run_fit(self, plot_it=False, burnin_steps=1000, sampling_steps=1000):
self.best_like_fit, chain, self.lnprob =\
gf.fit_group(
self.tb_file, fixed_age=self.fixed_age, burnin_steps=1000,
sampling_steps=1000
)
self.nwalkers, self.nsteps, self.npars = chain.shape
flat_chain = np.reshape(chain, (-1, self.npars))
self.nsamples = self.nwalkers * self.nsteps
logging.info("Fixed fit completed for {}".format(self.fixed_age))
cov_mats = self.calc_covs(flat_chain)
self.calc_pos_radii(cov_mats)
self.calc_mean_fit(cov_mats, flat_chain)
def run_fit(self, plot_it=False, burnin_steps=1000, sampling_steps=1000):
self.best_like_fit, self.chain, lnprob =\
gf.fit_group(
self.tb_file, plot_it=plot_it, burnin_steps=burnin_steps,
sampling_steps=sampling_steps
)
self.nwalkers, self.nsteps, self.npars = self.chain.shape
flat_chain = np.reshape(self.chain, (-1, self.npars))
self.nsamples = self.nwalkers*self.nsteps
cov_mats = self.calc_covs(flat_chain)
self.calc_pos_radii(cov_mats)
#self.calc_phase_radius(cov_mats)
self.calc_mean_fit(cov_mats, flat_chain)
def run_fit(self, plot_it=False, burnin_steps=1000, sampling_steps=1000):
self.best_like_fit, self.chain, lnprob =\
gf.fit_group(
self.tb_file, plot_it=plot_it, burnin_steps=burnin_steps,
sampling_steps=sampling_steps
)
self.nwalkers, self.nsteps, self.npars = self.chain.shape
flat_chain = np.reshape(self.chain, (-1, self.npars))
self.nsamples = self.nwalkers * self.nsteps
cov_mats = self.calc_covs(flat_chain)
self.calc_pos_radii(cov_mats)
#self.calc_phase_radius(cov_mats)
self.calc_mean_fit(cov_mats, flat_chain)
def maximise(infile,
ngroups,
z=None,
init_conditions=None,
burnin_steps=500,
sampling_steps=1000):
"""Given membership probabilities, maximise the parameters of each group model
Parameters
----------
infile : str
Name of the traceback file being fitted to
ngroups : int
Number of groups to be fitted to the traceback orbits
z : [nstars, ngroups] array
An array designating each star's probability of being a member to each
group. It is populated by floats in the range (0.0, 1.0) such that
each row sums to 1.0, each column sums to the expected size of each
group, and the entire array sums to the number of stars.
init_conditions : [ngroups, npars] array
The initial conditions for the groups, encoded in the 'internal' manner
(that is, 1/dX, 1/dY etc. and no star count)
burnin_steps : int
number of steps during burnin phase
sampling_steps : int
number of steps during sampling phase
Returns
-------
best_fits : [ngroups, npars] array
The best fitting parameters for each group
chains : [ngroups, nwalkers, nsteps, npars] array
The final chain for each group's fit
lnprobs : [ngroups, nwalkers, nsteps] array
The sampler.probability array for each group's fit
TODO: Have some means to enforce fixed ages
"""
NPARS = 14
best_fits = np.zeros((ngroups, NPARS))
chains = None
lnprobs = None
for i in range(ngroups):
best_fit, chain, lnprob = gf.fit_group(infile,
z=ix_snd(z, i),
init_pars=ix_fst(
init_conditions, i),
burnin_steps=burnin_steps,
sampling_steps=sampling_steps,
plot_it=True)
best_fits[i] = best_fit
if chains is None:
dims = np.append(ngroups, chain.shape)
chains = np.zeros(dims)
if lnprobs is None:
dims = np.append(ngroups, lnprob.shape)
lnprobs = np.zeros(dims)
chains[i] = chain
lnprobs[i] = lnprob
return best_fits, chains, lnprobs
def maximise(infile, ngroups, z=None, init_conditions=None, burnin_steps=500,
sampling_steps=1000):
"""Given membership probabilities, maximise the parameters of each group model
Parameters
----------
infile : str
Name of the traceback file being fitted to
ngroups : int
Number of groups to be fitted to the traceback orbits
z : [nstars, ngroups] array
An array designating each star's probability of being a member to each
group. It is populated by floats in the range (0.0, 1.0) such that
each row sums to 1.0, each column sums to the expected size of each
group, and the entire array sums to the number of stars.
init_conditions : [ngroups, npars] array
The initial conditions for the groups, encoded in the 'internal' manner
(that is, 1/dX, 1/dY etc. and no star count)
burnin_steps : int
number of steps during burnin phase
sampling_steps : int
number of steps during sampling phase
Returns
-------
best_fits : [ngroups, npars] array
The best fitting parameters for each group
chains : [ngroups, nwalkers, nsteps, npars] array
The final chain for each group's fit
lnprobs : [ngroups, nwalkers, nsteps] array
The sampler.probability array for each group's fit
TODO: Have some means to enforce fixed ages
"""
NPARS = 14
best_fits = np.zeros((ngroups, NPARS))
chains = None
lnprobs = None
for i in range(ngroups):
best_fit, chain, lnprob = gf.fit_group(
infile, z=ix_snd(z, i), init_pars=ix_fst(init_conditions, i),
burnin_steps=burnin_steps, sampling_steps=sampling_steps,
plot_it=True
)
best_fits[i] = best_fit
if chains is None:
dims = np.append(ngroups, chain.shape)
chains = np.zeros(dims)
if lnprobs is None:
dims = np.append(ngroups, lnprob.shape)
lnprobs = np.zeros(dims)
chains[i] = chain
lnprobs[i] = lnprob
return best_fits, chains, lnprobs