def calc_fom_hist(pred, true, weights, bins, range):
"""Calculate relative energy resolution histogram.
Parameters
----------
pred : ndarray, shape (N,)
Array of predicted energies.
true : ndarray, shape (N,)
Array of true energies.
weights : ndarray, shape (N,)
Array of sample weights.
bins : int or ndarray
If int then defined the number of bins in a histogram.
If ndarray then `bins` defines edges of the histogram.
C.f. `np.histogram`
range : (float, float) or None
Range of a histogram (lower, upper). If None, range will be determined
according to the `np.histogram` rules.
Returns
-------
cafplot.RHist1D
`Rhist1D` object containing the relative energy resolution histogram.
"""
# pylint: disable=redefined-builtin
fom = (pred - true) / true
return RHist1D.from_data(fom, bins, weights, range)
def plot_hist_base(
list_of_data_weight_label_color, key, spec, ratio_plot_type, stat_err,
log = False
):
"""Plot multiple energy histograms"""
if ratio_plot_type is not None:
f, ax, axr = make_figure_with_ratio()
else:
f, ax = plt.subplots()
if log:
ax.set_yscale('log')
list_of_rhist_color = []
for (data,weights,label,color) in list_of_data_weight_label_color:
rhist = RHist1D.from_data(data[key], spec.bins_x, weights)
centers = (rhist.bins_x[1:] + rhist.bins_x[:-1]) / 2
mean = np.average(centers, weights = rhist.hist)
plot_rhist1d(
ax, rhist,
histtype = 'step',
marker = None,
linestyle = '-',
linewidth = 2,
label = "%s. MEAN = %.3e" % (label, mean),
color = color,
)
if stat_err:
plot_rhist1d_error(
ax, rhist, err_type = 'bar', color = color, linewidth = 2,
alpha = 0.8
)
list_of_rhist_color.append((rhist, color))
spec.decorate(ax, ratio_plot_type)
if not log:
remove_bottom_margin(ax)
ax.legend()
if ratio_plot_type is not None:
plot_rhist1d_ratios(
axr,
[rhist_color[0] for rhist_color in list_of_rhist_color],
[rhist_color[1] for rhist_color in list_of_rhist_color],
err_kwargs = { 'err_type' : 'bar' if stat_err else None },
)
spec.decorate_ratio(axr, ratio_plot_type)
return f, ax
def _load_rhist1d(rhist_dict):
hist = np.array(rhist_dict['values'])
err_sq = np.array(rhist_dict['err_sq'])
bins = np.array(rhist_dict['bins'])
# Strip overflow/underflow bins
hist = hist[1:-1]
err_sq = err_sq[1:-1]
bins = [
bins,
]
return RHist1D(bins, hist, err_sq)
def get_sgn_bkg_preds(truth, preds, weights, truth_idx, pred_idx, **kwargs):
"""Calculate Signal and Background histograms.
Parameters
----------
truth : ndarray, shape (N_SAMPLES,)
Array of true targets.
preds : ndarray, shape (N_SAMPLES, N_TARGETS)
Array of predicted target scores.
weights : ndarray, shape (N_SAMPLES,)
Sample weights.
truth_idx : int
Value of `truth` target that indicates signal sample.
pred_idx : int
Index of the target to make histogram of.
kwargs : dict
Dictionary of values that will be passed to the `RHist1D` constructor.
Returns
-------
h_sgn : RHist1D
Histogram of signal values.
h_bkg : RHist1D
Histogram of background values.
"""
sgn_mask = (truth == truth_idx)
preds_sgn = preds[sgn_mask, pred_idx]
preds_bkg = preds[~sgn_mask, pred_idx]
w_sgn = weights[sgn_mask]
w_bkg = weights[~sgn_mask]
return (
RHist1D.from_data(preds_sgn, weights=w_sgn, **kwargs),
RHist1D.from_data(preds_bkg, weights=w_bkg, **kwargs),
)
def _load_rhist(path, d):
hist = d.get(path)
ndim = hist.values.ndim
args = ROOTFile._load_hist_internals(hist)
if ndim == 1:
bins, values, err_sq = args
return RHist1D([
bins,
], values, err_sq)
elif ndim == 2:
return RHist2D(*args)
else:
raise NotImplementedError
def plot_counts(counts, labels, plotdir, ext):
# pylint: disable=missing-function-docstring
norm = np.sum(counts)
f, ax = plt.subplots()
rhist = RHist1D(np.arange(len(labels) + 1), counts)
rhist.scale(100 / norm)
plot_rhist1d(ax, rhist, label = "", histtype = 'bar', color = 'tab:blue')
ax.set_xticks([ i + 0.5 for i in range(len(labels))])
ax.set_xticklabels(labels)
ax.set_ylabel("Fraction [%]")
ax.set_title("Distribution of events per category")
save_fig(f, os.path.join(plotdir, "sample_distribution_test"), ext)
def plot_energy(data, weights, name, spec, log_scale=False):
"""Plot single energy distribution."""
f, ax = plt.subplots()
if log_scale:
ax.set_yscale('log')
rhist = RHist1D.from_data(data, spec.bins_x, weights)
plot_rhist1d(
ax,
rhist,
histtype='step',
linewidth=2,
color='C0',
label="True %s. Mean: %.2e" %
(name, np.average(data, weights=weights)),
)
plot_rhist1d_error(ax, rhist, err_type='bar', color='C0', linewidth=2)
spec.decorate(ax)
ax.legend()
remove_bottom_margin(ax)
return f, ax, rhist
def _load_rhist1d(path, d):
return RHist1D(*ROOTFile._load_hist_internals(d.get(path)))