def __init__ (self,
i_obs,
crystal_symmetry=None,
n_bins=None,
i_over_sigma_min=2.0,
r_merge_max=0.5,
r_meas_max=0.5,
completeness_min_conservative=0.9,
completeness_min_permissive=0.5,
cc_one_half_min=0.5) :
self.n_bins = n_bins
self.i_over_sigma_min = i_over_sigma_min
self.r_merge_max = r_merge_max
self.r_meas_max = r_meas_max
self.completeness_min_conservative = completeness_min_conservative
self.completeness_min_permissive = completeness_min_permissive
self.cc_one_half_min = cc_one_half_min
for attr in self.cutoffs_attr :
setattr(self, attr, None)
# Decide n_bins
if n_bins is None:
n_bins = min(200, int(i_obs.complete_set().indices().size()/500.+.5)) # not well tested.
#print "n_bins=",n_bins
i_obs.setup_binner(n_bins=n_bins)
bins = []
for bin in i_obs.binner().range_used():
unmerged = i_obs.select(i_obs.binner().selection(bin))
try:
bin_stats = merging_statistics.merging_stats(unmerged,
anomalous=False)
bins.append(bin_stats)
except RuntimeError: # complains that no reflections left after sigma-filtering.
continue
self.d_min_overall = bins[-1].d_min
d_min_last = float("inf")
for bin in bins :
if (self.i_over_sigma_cut is None) :
if (bin.i_over_sigma_mean < self.i_over_sigma_min) :
self.i_over_sigma_cut = d_min_last
if (self.cc_one_half_cut is None) :
if (bin.cc_one_half < self.cc_one_half_min) :
self.cc_one_half_cut = d_min_last
if (self.r_merge_cut is None) :
if (bin.r_merge > self.r_merge_max) :
self.r_merge_cut = d_min_last
if (self.r_meas_cut is None) :
if (bin.r_meas > self.r_meas_max) :
self.r_meas_cut = d_min_last
if (self.completeness_cut_conservative is None) :
if (bin.completeness < completeness_min_conservative) :
self.completeness_cut_conservative = d_min_last
if (self.completeness_cut_permissive is None) :
if (bin.completeness < completeness_min_permissive) :
self.completeness_cut_permissive = d_min_last
d_min_last = bin.d_min
def initial_est_byfit(self):
import scipy.optimize
# Up to 200 bins. If few reflections, 50 reflections per bin. At least 9 shells.
n_bins = max(min(int(self.i_obs.size() / 50. + .5), 200), 9)
self.log_out.write("Using %d bins for initial estimate\n" % n_bins)
self.i_obs.setup_binner(n_bins=n_bins)
bins = []
s_list, cc_list = [], []
for bin in self.i_obs.binner().range_used():
unmerged = self.i_obs.select(self.i_obs.binner().selection(bin))
try:
bin_stats = merging_statistics.merging_stats(
unmerged, anomalous=self.anomalous_flag)
s_list.append(1. / bin_stats.d_min**2)
cc_list.append(bin_stats.cc_one_half)
except RuntimeError: # complains that no reflections left after sigma-filtering.
continue
# Fit curve
def fun(x, s, cc):
d0, r = x
return self.fun_ed_aimless(s, d0, r) - cc
# fun()
x0 = [0.5 * min(s_list), 1.] #Initial d0, r
self.log_out.write(" Initial d0, r = %s\n" % x0)
lsq = scipy.optimize.least_squares(fun,
x0,
args=(s_list, cc_list),
loss="soft_l1",
f_scale=.1)
#lsq = fit_ed(s_list, cc_list)
self.shells_and_fit = (s_list, cc_list, lsq.x)
self.log_out.write("Least-square result:\n")
self.log_out.write(str(lsq))
self.log_out.write("\n")
d0, r = lsq.x
self.log_out.write(" Final d0, r = %s\n" % lsq.x)
d_min = 1. / numpy.sqrt(
(numpy.arctanh(1. - 2. * self.cc_half_min) * r + d0))
return d_min
def initial_est_byfit(self):
import scipy.optimize
# Up to 200 bins. If few reflections, 50 reflections per bin. At least 9 shells.
n_bins = max(min(int(self.i_obs.size()/50. + .5), 200), 9)
self.log_out.write("Using %d bins for initial estimate\n" % n_bins)
self.i_obs.setup_binner(n_bins=n_bins)
bins = []
s_list, cc_list = [], []
for bin in self.i_obs.binner().range_used():
unmerged = self.i_obs.select(self.i_obs.binner().selection(bin))
try:
bin_stats = merging_statistics.merging_stats(unmerged,
anomalous=self.anomalous_flag)
s_list.append(1./bin_stats.d_min**2)
cc_list.append(bin_stats.cc_one_half)
except RuntimeError: # complains that no reflections left after sigma-filtering.
continue
# Fit curve
def fun(x, s, cc):
d0, r = x
return self.fun_ed_aimless(s,d0,r) - cc
# fun()
x0 = [0.5*min(s_list), 1.] #Initial d0, r
self.log_out.write(" Initial d0, r = %s\n" % x0)
lsq = scipy.optimize.least_squares(fun, x0, args=(s_list, cc_list), loss="soft_l1", f_scale=.1)
#lsq = fit_ed(s_list, cc_list)
self.shells_and_fit = (s_list, cc_list, lsq.x)
self.log_out.write("Least-square result:\n")
self.log_out.write(str(lsq))
self.log_out.write("\n")
d0, r = lsq.x
self.log_out.write(" Final d0, r = %s\n" % lsq.x)
d_min = 1./numpy.sqrt((numpy.arctanh(1.-2.*self.cc_half_min)*r + d0))
return d_min
def initial_estimate_byfit_cchalf(i_obs, cc_half_min, anomalous_flag, log_out):
# Up to 200 bins. If few reflections, 50 reflections per bin. At least 9 shells.
n_bins = max(min(int(i_obs.size() / 50. + .5), 200), 9)
log_out.write("Using %d bins for initial estimate\n" % n_bins)
i_obs.setup_binner(n_bins=n_bins)
bins = []
s_list, cc_list = [], []
for bin in i_obs.binner().range_used():
unmerged = i_obs.select(i_obs.binner().selection(bin))
try:
bin_stats = merging_statistics.merging_stats(
unmerged, anomalous=anomalous_flag)
s_list.append(1. / bin_stats.d_min**2)
cc_list.append(bin_stats.cc_one_half)
except RuntimeError: # complains that no reflections left after sigma-filtering.
continue
d0, r = fit_curve_for_cchalf(s_list, cc_list, log_out)
shells_and_fit = (s_list, cc_list, (d0, r))
d_min = resolution_fitted(d0, r, cc_half_min)
return d_min, shells_and_fit
def cc_outer_shell(self, d_min): binner = self.i_obs.setup_binner(d_min=d_min, n_bins=self.n_bins) bin_stats = merging_statistics.merging_stats(self.i_obs.select(binner.selection(binner.range_used()[-1])), anomalous=False) return bin_stats.cc_one_half
def __init__(self,
i_obs,
crystal_symmetry=None,
n_bins=None,
i_over_sigma_min=2.0,
r_merge_max=0.5,
r_meas_max=0.5,
completeness_min_conservative=0.9,
completeness_min_permissive=0.5,
cc_one_half_min=0.5):
self.n_bins = n_bins
self.i_over_sigma_min = i_over_sigma_min
self.r_merge_max = r_merge_max
self.r_meas_max = r_meas_max
self.completeness_min_conservative = completeness_min_conservative
self.completeness_min_permissive = completeness_min_permissive
self.cc_one_half_min = cc_one_half_min
for attr in self.cutoffs_attr:
setattr(self, attr, None)
# Decide n_bins
if n_bins is None:
n_bins = min(200,
int(i_obs.complete_set().indices().size() / 500. +
.5)) # not well tested.
print "n_bins=", n_bins
i_obs.setup_binner(n_bins=n_bins)
bins = []
for bin in i_obs.binner().range_used():
unmerged = i_obs.select(i_obs.binner().selection(bin))
try:
bin_stats = merging_statistics.merging_stats(unmerged,
anomalous=False)
bins.append(bin_stats)
except RuntimeError: # complains that no reflections left after sigma-filtering.
continue
self.d_min_overall = bins[-1].d_min
d_min_last = float("inf")
for bin in bins:
if (self.i_over_sigma_cut is None):
if (bin.i_over_sigma_mean < self.i_over_sigma_min):
self.i_over_sigma_cut = d_min_last
if (self.cc_one_half_cut is None):
if (bin.cc_one_half < self.cc_one_half_min):
self.cc_one_half_cut = d_min_last
if (self.r_merge_cut is None):
if (bin.r_merge > self.r_merge_max):
self.r_merge_cut = d_min_last
if (self.r_meas_cut is None):
if (bin.r_meas > self.r_meas_max):
self.r_meas_cut = d_min_last
if (self.completeness_cut_conservative is None):
if (bin.completeness < completeness_min_conservative):
self.completeness_cut_conservative = d_min_last
if (self.completeness_cut_permissive is None):
if (bin.completeness < completeness_min_permissive):
self.completeness_cut_permissive = d_min_last
d_min_last = bin.d_min
def cc_outer_shell(self, d_min):
binner = self.i_obs.setup_binner(d_min=d_min, n_bins=self.n_bins)
bin_stats = merging_statistics.merging_stats(self.i_obs.select(
binner.selection(binner.range_used()[-1])),
anomalous=False)
return bin_stats.cc_one_half
