def fitRingSizeHalf(t, r, rErr=None):
'''
Fits ring size data to the ringSizeFunc (above)
Returns: minimizer object defined by lmfit class. To access the parameters use res.best_values['r0'].
'''
if rErr==None: rErr = np.ones_like(r)
if r[0]<0.5 or r.size<=3: return None
params = Parameters()
params.add('tau', (t[-1]-t[0])/3., vary=True, min = 0, max=max(t))
params.add('t0', t[np.argmin(np.abs(r-0.5))], vary=True)
params.add('r0', 1.1, vary=False, min=0.)
mod = Model(ringSizeFuncHalf)
ind=np.where(r>0.2)[0]
if ind.size<3: ind = np.where(r>0.)
w = 1./rErr**2
res=mod.fit(r[ind],x=t[ind],weights = w[ind], params=params)
if False:
fig = myFigure()
print(fit_report(res))
fig.plot(t, r, 'r')
delta = (max(t)-min(t))
x=np.arange(min(t)-0.2*delta,max(t)+0.2*delta,delta/100.)
fig.plot(x, ringSizeFuncHalf(x,res.best_values['r0'], res.best_values['tau'],res.best_values['t0']),'k')
fig.show()
return res
def onSaveFit(self, evt=None):
"Save Fit and Compoents to Data File"
print("save fit and components")
nfit = self.current_fit
dgroup = self.datagroup
nfits = int(self.wids['plot_nchoice'].GetStringSelection())
deffile = "%s_LinearFit%i.dat" % (dgroup.filename, nfit + 1)
wcards = 'Data Files (*.dat)|*.dat|All files (*.*)|*.*'
path = FileSave(self,
'Save Fit and Components to File',
default_file=deffile,
wildcard=wcards)
if path is None:
return
form = self.form
label = [' energy ', ' data ', ' best_fit ']
result = dgroup.lcf_result[nfit]
header = [
'Fit #%2.2d' % (nfit + 1),
'Fit arrayname: %s' % form['arrayname'],
'E0: %f ' % form['e0'],
'Energy fit range: [%f, %f]' % (form['elo'], form['ehi']),
'Components: '
]
for key, val in result.weights.items():
header.append(' %s: %f' % (key, val))
report = fit_report(result.result).split('\n')
header.extend(report)
out = [result.xdata, result.ydata, result.yfit]
for compname, compdata in result.ycomps.items():
label.append(' %s' % (compname + ' ' *
(max(1, 15 - len(compname)))))
out.append(compdata)
label = ' '.join(label)
_larch = self.parent.controller.larch
write_ascii(path, header=header, label=label, _larch=_larch, *out)
def onSaveFit(self, evt=None):
"Save Fit and Compoents to Data File"
nfit = self.current_fit
dgroup = self.datagroup
nfits = int(self.wids['plot_nchoice'].GetStringSelection())
deffile = "%s_LinearFit%i.dat" % (dgroup.filename, nfit+1)
wcards = 'Data Files (*.dat)|*.dat|All files (*.*)|*.*'
path = FileSave(self, 'Save Fit and Components to File',
default_file=deffile, wildcard=wcards)
if path is None:
return
form = self.form
label = [' energy ',
' data ',
' best_fit ']
result = dgroup.lcf_result[nfit]
header = ['Larch Linear Fit Result for Fit: #%2.2d' % (nfit+1),
'Dataset filename: %s ' % dgroup.filename,
'Larch group: %s ' % dgroup.groupname,
'Array name: %s' % form['arrayname'],
'E0: %f ' % form['e0'],
'Energy fit range: [%f, %f]' % (form['elo'], form['ehi']),
'Components: ']
for key, val in result.weights.items():
header.append(' %s: %f' % (key, val))
report = fit_report(result.result).split('\n')
header.extend(report)
out = [result.xdata, result.ydata, result.yfit]
for compname, compdata in result.ycomps.items():
label.append(' %s' % (compname + ' '*(max(1, 15-len(compname)))))
out.append(compdata)
label = ' '.join(label)
_larch = self.parent.controller.larch
write_ascii(path, header=header, label=label, _larch=_larch, *out)
def generate_fit_report(self):
report_header = f'\n\n{"="*10}\nFIT REPORT\n{"="*10}\n\n'
return report_header + fit_report(self.result.params)
def main(argv):
parser = ArgumentParser(usage=__doc__.lstrip())
parser.add_argument("global_pilot_file",
help="The name of the global pilot file")
parser.add_argument("--walkers", "-w", type=int, default=100,
help="How many MCMC walkers? Default=100")
parser.add_argument("--steps", "-s", type=int, default=2000,
help="How many MCMC steps? Default=2000")
parser.add_argument("--ntemps", '-T', type=int, default=1,
help="How many parallel tempering temperatures? "
"Default=1")
parser.add_argument("--burn", "-b", type=int, default=500,
help="How many initial MCMC steps do you want to "
"burn? Default=500")
parser.add_argument("--thin", "-t", type=int, default=20,
help="Thins the chain by accepting 1 in every 'thin'. "
"Default=20")
parser.add_argument("--resample", "-r", default=False, action='store_true',
help="Don't do Markov Chain Monte Carlo, do resampling "
"MC instead. All MCMC parameters are ignored if "
"this option is specified.")
parser.add_argument("--qres", "-q", type=float, default=5.0,
help="Constant dq/q resolution. Default=5")
parser.add_argument("--pointqres", "-p", action="store_true",
default=False,
help="Use point by point resolution smearing. "
"Default=False")
parser.add_argument("--chain_input", "-i", type=str,
help="Initialise/restart emcee with this RAW chain. "
"This file is a numpy array (.npy) that would've "
"originally been saved by the --chain_output "
"option.")
parser.add_argument("--chain_output", "-c", default='raw_chain.npy',
type=str,
help="Specify filename for unthinned, unburnt RAW "
"chain. The file is saved as a numpy (.npy) "
"array. You can use this file if you'd like to do "
"the burn/thin procedure yourself. You can also"
" use this file to restart the sampling. The array"
" has shape (walkers, steps, dims), where dims "
"represents the number of parameters you are "
"varying. If ntemps is > 1 then the array has "
"shape (ntemps, walkers, steps, dims). "
"Default=raw_chain.npy")
parser.add_argument("--output", "-o", type=str, default='iterations',
help="Output file for burnt and thinned MCMC chain. "
"This is only written once the sampling has "
"finished")
parser.add_argument("--nprocesses", "-n", type=int, default=1,
help="How many processes for parallelisation? Default=1")
args = parser.parse_args(argv)
# set up global fitting
if args.pointqres:
dqvals = None
else:
dqvals = args.qres
if args.nprocesses < 1:
args.nprocesses = 1
global_fitter = global_fitter_setup(args.global_pilot_file,
dqvals=dqvals)
# do the Monte Carlo
if not args.resample:
# By Markov Chain Monte Carlo
if args.thin < 1:
sys.stdout.write("Can't have thin < 1, setting 'thin' to 1.\n")
args.thin = 1
if (args.burn < 0) or (args.burn > args.steps):
sys.stdout.write("Can't burn < 0 or burn > steps, setting 'burn' to 1.\n")
args.burn = 1
res = _mcmc(args, global_fitter)
else:
# By resampling
res = _resample_mc(args, global_fitter)
# write the iterations.
_write_results(args.output, res)
sys.stdout.write("\nFinished MCMC\n")
sys.stdout.write("-------------\n")
sys.stdout.write(fit_report(res.params))
sys.stdout.write("\n-----------------------------------------------------\n")
t = reflect.Transform('logY').transform
ydata, dydata = t(xdata, ydata, dydata)
kwds = {'dqvals': dxdata, 'transform': t}
# create the fit instance
fitter = reflect.ReflectivityFitFunction(parameters, xdata, ydata, edata=dydata,
kwds=kwds)
#do the fit
if USE_DIFFERENTIAL_EVOLUTION:
fitter.fit(method='differential_evolution')
fitter.fit()
print('-------------------------------------------------------------------')
print(DATASET_NAME)
print(fit_report(fitter))
fig = figure()
ax = fig.add_subplot(2, 1, 1)
ax.scatter(xdata, ydata, label=DATASET_NAME)
ax.plot(xdata, fitter.model(parameters), label='fit')
ylim(min(np.min(ydata), np.min(fitter.model(parameters))),
max(np.max(ydata), np.max(fitter.model(parameters))))
xlim(np.min(xdata), np.max(xdata))
xlabel('Q')
ylabel('logR')
legend()
ax2 = fig.add_subplot(2, 1, 2)
z, rho_z = fitter.sld_profile(parameters)
ax2.plot(z, rho_z)
