def run(args, cutoff, max_n_terms, six_term=False, params=None,
plots_dir="kissel_fits_plots", verbose=0):
if (params is None):
params = cctbx.eltbx.gaussian_fit.fit_parameters(
max_n_terms=max_n_terms)
chunk_n = 1
chunk_i = 0
if (len(args) > 0 and len(args[0].split(",")) == 2):
chunk_n, chunk_i = [int(i) for i in args[0].split(",")]
args = args[1:]
if (not six_term):
if (not os.path.isdir(plots_dir)):
print "No plots because target directory does not exist (mkdir %s)." % \
plots_dir
plots_dir = None
if (chunk_n > 1):
assert plots_dir is not None
i_chunk = 0
for file_name in args:
flag = i_chunk % chunk_n == chunk_i
i_chunk += 1
if (not flag):
continue
results = {}
results["fit_parameters"] = params
tab = kissel_io.read_table(file_name)
more_selection = tab.itvc_sampling_selection()
fit_selection = more_selection & (tab.x <= cutoff + 1.e-6)
null_fit = scitbx.math.gaussian.fit(
tab.x.select(fit_selection),
tab.y.select(fit_selection),
tab.sigmas.select(fit_selection),
xray_scattering.gaussian(0, False))
null_fit_more = scitbx.math.gaussian.fit(
tab.x.select(more_selection),
tab.y.select(more_selection),
tab.sigmas.select(more_selection),
xray_scattering.gaussian(0, False))
if (not six_term):
results[tab.element] = cctbx.eltbx.gaussian_fit.incremental_fits(
label=tab.element,
null_fit=null_fit,
params=params,
plots_dir=plots_dir,
verbose=verbose)
else:
best_min = scitbx.math.gaussian_fit.fit_with_golay_starts(
label=tab.element,
null_fit=null_fit,
null_fit_more=null_fit_more,
params=params)
g = best_min.final_gaussian_fit
results[tab.element] = [xray_scattering.fitted_gaussian(
stol=g.table_x()[-1], gaussian_sum=g)]
sys.stdout.flush()
pickle_file_name = "%s_fits.pickle" % identifier(tab.element)
easy_pickle.dump(pickle_file_name, results)
def zig_zag_fits(label, null_fit, null_fit_more, params):
six_term_best_min = scitbx.math.gaussian_fit.fit_with_golay_starts(
label=label,
null_fit=null_fit,
null_fit_more=null_fit_more,
params=params)
results = []
n_term_best_min = six_term_best_min
have_all_points_in_previous = True
while 1:
while 1:
if (n_term_best_min.final_gaussian_fit.n_terms() == 1):
existing_gaussian = null_fit
else:
decr_best_min = scitbx.math.gaussian_fit.decremental_fit(
existing_gaussian=n_term_best_min.final_gaussian_fit,
params=params)
assert decr_best_min is not None
print "Decremental:",
decr_best_min.show_summary()
existing_gaussian = decr_best_min.final_gaussian_fit
if (n_term_best_min.max_error <= params.negligible_max_error
and have_all_points_in_previous):
break
incr_best_min = find_max_x_multi(
null_fit=null_fit,
existing_gaussian=existing_gaussian,
target_powers=params.target_powers,
minimize_using_sigmas=params.minimize_using_sigmas,
n_repeats_minimization=params.n_repeats_minimization,
shift_sqrt_b_mod_n=params.shift_sqrt_b_mod_n,
b_min=params.b_min,
max_max_error=params.max_max_error,
n_start_fractions=params.n_start_fractions)
assert incr_best_min is not None
print "Incremental:",
incr_best_min.show_summary()
if (existing_gaussian is null_fit):
break
if (not incr_best_min.is_better_than(n_term_best_min)):
break
n_term_best_min = incr_best_min
print " Settled on:",
n_term_best_min.show_summary()
results.append(
xray_scattering.fitted_gaussian(
stol=n_term_best_min.final_gaussian_fit.table_x()[-1],
gaussian_sum=n_term_best_min.final_gaussian_fit,
max_error=n_term_best_min.max_error))
if (existing_gaussian is null_fit):
break
have_all_points_in_previous = (
n_term_best_min.final_gaussian_fit.table_x().size() ==
decr_best_min.final_gaussian_fit.table_x().size())
n_term_best_min = decr_best_min
print
return results
def zig_zag_fits(label, null_fit, null_fit_more, params):
six_term_best_min = scitbx.math.gaussian_fit.fit_with_golay_starts(
label=label,
null_fit=null_fit,
null_fit_more=null_fit_more,
params=params)
results = []
n_term_best_min = six_term_best_min
have_all_points_in_previous = True
while 1:
while 1:
if (n_term_best_min.final_gaussian_fit.n_terms() == 1):
existing_gaussian = null_fit
else:
decr_best_min = scitbx.math.gaussian_fit.decremental_fit(
existing_gaussian=n_term_best_min.final_gaussian_fit,
params=params)
assert decr_best_min is not None
print "Decremental:",
decr_best_min.show_summary()
existing_gaussian = decr_best_min.final_gaussian_fit
if (n_term_best_min.max_error <= params.negligible_max_error
and have_all_points_in_previous):
break
incr_best_min = find_max_x_multi(
null_fit=null_fit,
existing_gaussian=existing_gaussian,
target_powers=params.target_powers,
minimize_using_sigmas=params.minimize_using_sigmas,
n_repeats_minimization=params.n_repeats_minimization,
shift_sqrt_b_mod_n=params.shift_sqrt_b_mod_n,
b_min=params.b_min,
max_max_error=params.max_max_error,
n_start_fractions=params.n_start_fractions)
assert incr_best_min is not None
print "Incremental:",
incr_best_min.show_summary()
if (existing_gaussian is null_fit):
break
if (not incr_best_min.is_better_than(n_term_best_min)):
break
n_term_best_min = incr_best_min
print " Settled on:",
n_term_best_min.show_summary()
results.append(xray_scattering.fitted_gaussian(
stol=n_term_best_min.final_gaussian_fit.table_x()[-1],
gaussian_sum=n_term_best_min.final_gaussian_fit,
max_error=n_term_best_min.max_error))
if (existing_gaussian is null_fit):
break
have_all_points_in_previous = (
n_term_best_min.final_gaussian_fit.table_x().size()
== decr_best_min.final_gaussian_fit.table_x().size())
n_term_best_min = decr_best_min
print
return results
def incremental_fits(label, null_fit, params=None, plots_dir=None, verbose=0):
if (params is None): params = fit_parameters()
f0 = null_fit.table_y()[0]
results = []
previous_n_points = 0
existing_gaussian = xray_scattering.gaussian([], [])
while (existing_gaussian.n_terms() < params.max_n_terms):
if (previous_n_points == null_fit.table_x().size()):
print "%s: Full fit with %d terms. Search stopped." % (
label, existing_gaussian.n_terms())
print
break
n_terms = existing_gaussian.n_terms() + 1
best_min = find_max_x_multi(
null_fit=null_fit,
existing_gaussian=existing_gaussian,
target_powers=params.target_powers,
minimize_using_sigmas=params.minimize_using_sigmas,
n_repeats_minimization=params.n_repeats_minimization,
shift_sqrt_b_mod_n=params.shift_sqrt_b_mod_n,
b_min=params.b_min,
max_max_error=params.max_max_error,
n_start_fractions=params.n_start_fractions)
if (best_min is None):
print "Warning: No fit: %s n_terms=%d" % (label, n_terms)
print
break
if (previous_n_points > best_min.final_gaussian_fit.table_x().size()):
print "Warning: previous fit included more sampling points."
previous_n_points = best_min.final_gaussian_fit.table_x().size()
show_fit_summary("Best fit", label, best_min.final_gaussian_fit,
best_min.max_error)
show_literature_fits(
label=label,
n_terms=n_terms,
null_fit=null_fit,
n_points=best_min.final_gaussian_fit.table_x().size(),
e_other=best_min.max_error)
best_min.final_gaussian_fit.show()
best_min.show_minimization_parameters()
existing_gaussian = best_min.final_gaussian_fit
print
show_minimize_multi_histogram()
sys.stdout.flush()
if (plots_dir):
write_plots(plots_dir=plots_dir,
label=label + "_%d" % n_terms,
gaussian_fit=best_min.final_gaussian_fit)
g = best_min.final_gaussian_fit
results.append(
xray_scattering.fitted_gaussian(stol=g.table_x()[-1],
gaussian_sum=g))
return results
def incremental_fits(label, null_fit, params=None, plots_dir=None, verbose=0):
if (params is None): params = fit_parameters()
f0 = null_fit.table_y()[0]
results = []
previous_n_points = 0
existing_gaussian = xray_scattering.gaussian([],[])
while (existing_gaussian.n_terms() < params.max_n_terms):
if (previous_n_points == null_fit.table_x().size()):
print "%s: Full fit with %d terms. Search stopped." % (
label, existing_gaussian.n_terms())
print
break
n_terms = existing_gaussian.n_terms() + 1
best_min = find_max_x_multi(
null_fit=null_fit,
existing_gaussian=existing_gaussian,
target_powers=params.target_powers,
minimize_using_sigmas=params.minimize_using_sigmas,
n_repeats_minimization=params.n_repeats_minimization,
shift_sqrt_b_mod_n=params.shift_sqrt_b_mod_n,
b_min=params.b_min,
max_max_error=params.max_max_error,
n_start_fractions=params.n_start_fractions)
if (best_min is None):
print "Warning: No fit: %s n_terms=%d" % (label, n_terms)
print
break
if (previous_n_points > best_min.final_gaussian_fit.table_x().size()):
print "Warning: previous fit included more sampling points."
previous_n_points = best_min.final_gaussian_fit.table_x().size()
show_fit_summary(
"Best fit", label, best_min.final_gaussian_fit, best_min.max_error)
show_literature_fits(
label=label,
n_terms=n_terms,
null_fit=null_fit,
n_points=best_min.final_gaussian_fit.table_x().size(),
e_other=best_min.max_error)
best_min.final_gaussian_fit.show()
best_min.show_minimization_parameters()
existing_gaussian = best_min.final_gaussian_fit
print
show_minimize_multi_histogram()
sys.stdout.flush()
if (plots_dir):
write_plots(
plots_dir=plots_dir,
label=label+"_%d"%n_terms,
gaussian_fit=best_min.final_gaussian_fit)
g = best_min.final_gaussian_fit
results.append(xray_scattering.fitted_gaussian(
stol=g.table_x()[-1], gaussian_sum=g))
return results
def decremental_fits(label,
null_fit,
full_fit=None,
params=None,
plots_dir=None,
verbose=0):
if (params is None): params = fit_parameters()
results = []
last_fit = scitbx.math.gaussian.fit(null_fit.table_x(), null_fit.table_y(),
null_fit.table_sigmas(), full_fit)
while (last_fit.n_terms() > 1):
good_min = scitbx.math.gaussian_fit.decremental_fit(
existing_gaussian=last_fit, params=params)
if (good_min is None):
print "%s n_terms=%d: No successful minimization. Aborting." % (
label, last_fit.n_terms() - 1)
break
show_fit_summary("Best fit", label, good_min.final_gaussian_fit,
good_min.max_error)
show_literature_fits(
label=label,
n_terms=good_min.final_gaussian_fit.n_terms(),
null_fit=null_fit,
n_points=good_min.final_gaussian_fit.table_x().size(),
e_other=good_min.max_error)
good_min.final_gaussian_fit.show()
good_min.show_minimization_parameters()
last_fit = good_min.final_gaussian_fit
print
show_minimize_multi_histogram()
sys.stdout.flush()
if (plots_dir):
write_plots(plots_dir=plots_dir,
label=label +
"_%d" % good_min.final_gaussian_fit.n_terms(),
gaussian_fit=good_min.final_gaussian_fit)
g = good_min.final_gaussian_fit
results.append(
xray_scattering.fitted_gaussian(stol=g.table_x()[-1],
gaussian_sum=g))
return results
def decremental_fits(label, null_fit, full_fit=None, params=None,
plots_dir=None, verbose=0):
if (params is None): params = fit_parameters()
results = []
last_fit = scitbx.math.gaussian.fit(
null_fit.table_x(),
null_fit.table_y(),
null_fit.table_sigmas(),
full_fit)
while (last_fit.n_terms() > 1):
good_min = scitbx.math.gaussian_fit.decremental_fit(
existing_gaussian=last_fit,
params=params)
if (good_min is None):
print "%s n_terms=%d: No successful minimization. Aborting." % (
label, last_fit.n_terms()-1)
break
show_fit_summary(
"Best fit", label, good_min.final_gaussian_fit, good_min.max_error)
show_literature_fits(
label=label,
n_terms=good_min.final_gaussian_fit.n_terms(),
null_fit=null_fit,
n_points=good_min.final_gaussian_fit.table_x().size(),
e_other=good_min.max_error)
good_min.final_gaussian_fit.show()
good_min.show_minimization_parameters()
last_fit = good_min.final_gaussian_fit
print
show_minimize_multi_histogram()
sys.stdout.flush()
if (plots_dir):
write_plots(
plots_dir=plots_dir,
label=label+"_%d"%good_min.final_gaussian_fit.n_terms(),
gaussian_fit=good_min.final_gaussian_fit)
g = good_min.final_gaussian_fit
results.append(xray_scattering.fitted_gaussian(
stol=g.table_x()[-1], gaussian_sum=g))
return results
def run(file_name,
args,
cutoff,
params,
zig_zag=False,
six_term=False,
full_fits=None,
plots_dir="itvc_fits_plots",
verbose=0):
tab = itvc_section61_io.read_table6111(file_name)
chunk_n = 1
chunk_i = 0
if (len(args) > 0 and len(args[0].split(",")) == 2):
chunk_n, chunk_i = [int(i) for i in args[0].split(",")]
args = args[1:]
if (not six_term and not zig_zag):
if (not os.path.isdir(plots_dir)):
print("No plots because target directory does not exist (mkdir %s)." % \
plots_dir)
plots_dir = None
if (chunk_n > 1):
assert plots_dir is not None
stols_more = cctbx.eltbx.gaussian_fit.international_tables_stols
sel = stols_more <= cutoff + 1.e-6
stols = stols_more.select(sel)
i_chunk = 0
for element in tab.elements + ["O2-", "SDS"]:
if (len(args) > 0 and element not in args): continue
flag = i_chunk % chunk_n == chunk_i
i_chunk += 1
if (not flag):
continue
results = {}
results["fit_parameters"] = params
if (element == "SDS"):
wrk_lbl = element
from cctbx.eltbx.development.hydrogen_plots import fit_input
fi = fit_input()
sel = fi.stols <= cutoff + 1.e-6
null_fit = scitbx.math.gaussian.fit(
fi.stols.select(sel), fi.data.select(sel),
fi.sigmas.select(sel), xray_scattering.gaussian(0, False))
null_fit_more = scitbx.math.gaussian.fit(
fi.stols, fi.data, fi.sigmas,
xray_scattering.gaussian(0, False))
else:
wrk_lbl = xray_scattering.wk1995(element, True)
if (element != "O2-"):
entry = tab.entries[element]
null_fit = scitbx.math.gaussian.fit(
stols, entry.table_y[:stols.size()],
entry.table_sigmas[:stols.size()],
xray_scattering.gaussian(0, False))
null_fit_more = scitbx.math.gaussian.fit(
stols_more, entry.table_y[:stols_more.size()],
entry.table_sigmas[:stols_more.size()],
xray_scattering.gaussian(0, False))
else:
rrg_stols_more = rez_rez_grant.table_2_stol
sel = rrg_stols_more <= cutoff + 1.e-6
rrg_stols = rrg_stols_more.select(sel)
null_fit = scitbx.math.gaussian.fit(
rrg_stols,
rez_rez_grant.table_2_o2minus[:rrg_stols.size()],
rez_rez_grant.table_2_sigmas[:rrg_stols.size()],
xray_scattering.gaussian(0, False))
null_fit_more = scitbx.math.gaussian.fit(
rrg_stols_more,
rez_rez_grant.table_2_o2minus[:rrg_stols_more.size()],
rez_rez_grant.table_2_sigmas[:rrg_stols_more.size()],
xray_scattering.gaussian(0, False))
if (zig_zag):
results[wrk_lbl] = cctbx.eltbx.gaussian_fit.zig_zag_fits(
label=wrk_lbl,
null_fit=null_fit,
null_fit_more=null_fit_more,
params=params)
elif (full_fits is not None):
assert len(full_fits.all[wrk_lbl]) == 1
results[wrk_lbl] = cctbx.eltbx.gaussian_fit.decremental_fits(
label=wrk_lbl,
null_fit=null_fit,
full_fit=full_fits.all[wrk_lbl][0],
params=params,
plots_dir=plots_dir,
verbose=verbose)
elif (not six_term):
results[wrk_lbl] = cctbx.eltbx.gaussian_fit.incremental_fits(
label=wrk_lbl,
null_fit=null_fit,
params=params,
plots_dir=plots_dir,
verbose=verbose)
else:
best_min = scitbx.math.gaussian_fit.fit_with_golay_starts(
label=wrk_lbl,
null_fit=null_fit,
null_fit_more=null_fit_more,
params=params)
g = best_min.final_gaussian_fit
results[wrk_lbl] = [
xray_scattering.fitted_gaussian(stol=g.table_x()[-1],
gaussian_sum=g)
]
sys.stdout.flush()
pickle_file_name = "%s_fits.pickle" % identifier(wrk_lbl)
easy_pickle.dump(pickle_file_name, results)
def run(file_name, args, cutoff, params,
zig_zag=False, six_term=False, full_fits=None,
plots_dir="itvc_fits_plots", verbose=0):
tab = itvc_section61_io.read_table6111(file_name)
chunk_n = 1
chunk_i = 0
if (len(args) > 0 and len(args[0].split(",")) == 2):
chunk_n, chunk_i = [int(i) for i in args[0].split(",")]
args = args[1:]
if (not six_term and not zig_zag):
if (not os.path.isdir(plots_dir)):
print "No plots because target directory does not exist (mkdir %s)." % \
plots_dir
plots_dir = None
if (chunk_n > 1):
assert plots_dir is not None
stols_more = cctbx.eltbx.gaussian_fit.international_tables_stols
sel = stols_more <= cutoff + 1.e-6
stols = stols_more.select(sel)
i_chunk = 0
for element in tab.elements + ["O2-", "SDS"]:
if (len(args) > 0 and element not in args): continue
flag = i_chunk % chunk_n == chunk_i
i_chunk += 1
if (not flag):
continue
results = {}
results["fit_parameters"] = params
if (element == "SDS"):
wrk_lbl = element
from cctbx.eltbx.development.hydrogen_plots import fit_input
fi = fit_input()
sel = fi.stols <= cutoff + 1.e-6
null_fit = scitbx.math.gaussian.fit(
fi.stols.select(sel),
fi.data.select(sel),
fi.sigmas.select(sel),
xray_scattering.gaussian(0, False))
null_fit_more = scitbx.math.gaussian.fit(
fi.stols,
fi.data,
fi.sigmas,
xray_scattering.gaussian(0, False))
else:
wrk_lbl = xray_scattering.wk1995(element, True)
if (element != "O2-"):
entry = tab.entries[element]
null_fit = scitbx.math.gaussian.fit(
stols,
entry.table_y[:stols.size()],
entry.table_sigmas[:stols.size()],
xray_scattering.gaussian(0, False))
null_fit_more = scitbx.math.gaussian.fit(
stols_more,
entry.table_y[:stols_more.size()],
entry.table_sigmas[:stols_more.size()],
xray_scattering.gaussian(0, False))
else:
rrg_stols_more = rez_rez_grant.table_2_stol
sel = rrg_stols_more <= cutoff + 1.e-6
rrg_stols = rrg_stols_more.select(sel)
null_fit = scitbx.math.gaussian.fit(
rrg_stols,
rez_rez_grant.table_2_o2minus[:rrg_stols.size()],
rez_rez_grant.table_2_sigmas[:rrg_stols.size()],
xray_scattering.gaussian(0, False))
null_fit_more = scitbx.math.gaussian.fit(
rrg_stols_more,
rez_rez_grant.table_2_o2minus[:rrg_stols_more.size()],
rez_rez_grant.table_2_sigmas[:rrg_stols_more.size()],
xray_scattering.gaussian(0, False))
if (zig_zag):
results[wrk_lbl] = cctbx.eltbx.gaussian_fit.zig_zag_fits(
label=wrk_lbl,
null_fit=null_fit,
null_fit_more=null_fit_more,
params=params)
elif (full_fits is not None):
assert len(full_fits.all[wrk_lbl]) == 1
results[wrk_lbl] = cctbx.eltbx.gaussian_fit.decremental_fits(
label=wrk_lbl,
null_fit=null_fit,
full_fit=full_fits.all[wrk_lbl][0],
params=params,
plots_dir=plots_dir,
verbose=verbose)
elif (not six_term):
results[wrk_lbl] = cctbx.eltbx.gaussian_fit.incremental_fits(
label=wrk_lbl,
null_fit=null_fit,
params=params,
plots_dir=plots_dir,
verbose=verbose)
else:
best_min = scitbx.math.gaussian_fit.fit_with_golay_starts(
label=wrk_lbl,
null_fit=null_fit,
null_fit_more=null_fit_more,
params=params)
g = best_min.final_gaussian_fit
results[wrk_lbl] = [xray_scattering.fitted_gaussian(
stol=g.table_x()[-1], gaussian_sum=g)]
sys.stdout.flush()
pickle_file_name = "%s_fits.pickle" % identifier(wrk_lbl)
easy_pickle.dump(pickle_file_name, results)
