def plot_and_label_1d_signal_and_background_with_matplotlib_on_axis(ax: matplotlib.axes.Axes,
jet_hadron: "correlations.Correlations",
apply_correlation_scale_factor: bool = True) -> None:
""" Plot and label the signal and background dominated hists on the given axis.
This is a helper function so that we don't have to repeat code when we need to plot these hists.
It can also be used in other modules.
Args:
ax: Axis on which the histograms should be plotted.
jet_hadron: Correlations object from which the delta_phi hists should be retrieved.
apply_correlation_scale_factor: Whether to scale the histogram by the correlation scale factor.
Returns:
None. The given axis is modified.
"""
# Setup
hists = jet_hadron.correlation_hists_delta_phi
h_signal = histogram.Histogram1D.from_existing_hist(hists.signal_dominated.hist)
if apply_correlation_scale_factor:
h_signal *= jet_hadron.correlation_scale_factor
ax.errorbar(
h_signal.x, h_signal.y, yerr = h_signal.errors,
label = hists.signal_dominated.type.display_str(), marker = "o", linestyle = "",
)
h_background = histogram.Histogram1D.from_existing_hist(hists.background_dominated.hist)
if apply_correlation_scale_factor:
h_background *= jet_hadron.correlation_scale_factor
# Plot with opacity first
background_plot = ax.errorbar(
h_background.x, h_background.y, yerr = h_background.errors,
marker = "o", linestyle = "", alpha = 0.5,
)
# Then restrict range and plot without opacity
near_side = len(h_background.x) // 2
ax.errorbar(
h_background.x[:near_side], h_background.y[:near_side], yerr = h_background.errors[:near_side],
label = hists.background_dominated.type.display_str(),
marker = "o", linestyle = "", color = background_plot[0].get_color()
)
# Set labels.
ax.set_xlabel(labels.make_valid_latex_string(hists.signal_dominated.hist.GetXaxis().GetTitle()))
ax.set_ylabel(labels.make_valid_latex_string(hists.signal_dominated.hist.GetYaxis().GetTitle()))
jet_pt_label = labels.jet_pt_range_string(jet_hadron.jet_pt)
track_pt_label = labels.track_pt_range_string(jet_hadron.track_pt)
ax.set_title(fr"Unsubtracted 1D ${hists.signal_dominated.axis.display_str()}$,"
f" {jet_hadron.reaction_plane_orientation.display_str()} event plane orient.,"
f" {jet_pt_label}, {track_pt_label}")
def err_plot(x, y, yerr=None, xerr=None, axis: mpl.axes.Axes = None, **mpl_args):
"""Plot graph with error bars.
Decided weather to plot points with error bars or lines with shaded areas
depending on the number of plotted points.
Parameters
----------
x, y : array_like
x and y coordinates of the data to plot.
yerr, xerr : array_like, optional
The corresponding error of the data. Has same shape `x` and `y`.
axis : mpl.axes.Axes, optional
`mpl.axes.Axes` object used for plotting.
mpl_args :
Arguments for plotting passed to `axis.errorbar` or `axis.plot`.
"""
axis = plt.gca() if axis is None else axis
x = np.asarray(x)
if x.size > 50: # continuous plot
try:
ecolor = mpl_args.pop('ecolor')
except KeyError: # no color defined -> try color else default
ecolor = mpl_args.get('color', None)
try:
fmt = mpl_args.pop('fmt')
except KeyError:
baseline, = axis.plot(x, y, **mpl_args)
else:
baseline, = axis.plot(x, y, fmt, **mpl_args)
if ecolor is None:
ecolor = baseline.get_color()
if yerr is not None:
axis.fill_between(x, y-yerr, y+yerr, color=ecolor, alpha=.3, zorder=1)
if xerr is not None:
axis.fill_betweenx(y, x-xerr, x+xerr, color=ecolor, alpha=.3, zorder=1)
else:
default_args = {'capsize': 2.5, 'elinewidth': .3}
default_args.update(mpl_args)
axis.errorbar(x, y, yerr=yerr, **default_args)
def _reference_v2_a_data(
reference_data: ReferenceData, ax: matplotlib.axes.Axes,
selected_analysis_options: params.SelectedAnalysisOptions) -> None:
# Determine the centrality key (because they don't map cleanly onto our centrality ranges)
reference_data_centrality_map = {
params.EventActivity.central: "0-5",
params.EventActivity.semi_central: "30-40",
}
centrality_label = reference_data_centrality_map[
selected_analysis_options.event_activity]
# Retrieve the data
hist = reference_data["ptDependent"][centrality_label]
# Plot and label it on the existing axis.
ax.errorbar(
hist.x,
hist.y,
yerr=hist.errors,
marker="o",
linestyle="",
label=fr"ALICE {centrality_label}\% " +
labels.make_valid_latex_string("v_{2}") + r"\{2, $| \Delta\eta |>1$\}",
)
def plot_scatter(ax: mpl.axes.Axes,
accuracy_samples: np.ndarray,
ece_samples: np.ndarray,
limit=5,
plot_kwargs: Dict[str, Any] = {}) -> mpl.axes.Axes:
_plot_kwargs = DEFAULT_PLOT_KWARGS.copy()
_plot_kwargs.update(plot_kwargs)
# plot
x = np.mean(accuracy_samples, axis=1)
y = np.mean(ece_samples, axis=1)
xerr = np.std(accuracy_samples, axis=1)
yerr = np.std(ece_samples, axis=1)
# most accuracy top k
idx = x.argsort()[-limit:][::-1]
ax.errorbar(x[idx], y[idx],
xerr=xerr[idx],
yerr=yerr[idx],
fmt='o', alpha=0.8, color='b', **_plot_kwargs)
# least accuracy top k
idx = x.argsort()[:limit]
ax.errorbar(x[idx], y[idx],
xerr=xerr[idx],
yerr=yerr[idx],
fmt='o', alpha=0.8, color='r', **_plot_kwargs)
# other predicted classes
idx = x.argsort()[limit:-limit]
ax.errorbar(x[idx], y[idx],
xerr=xerr[idx],
yerr=yerr[idx],
fmt='o', alpha=0.2, color='k', **_plot_kwargs)
ax.set_xlabel('Accuracy')
# ax.set_ylabel('ECE')
return ax
def plot_reliability_comparison(ax: mpl.axes.Axes,
N_list: List[int],
bayesian_ece: np.ndarray,
frequentist_ece: np.ndarray,
bayesian_ece_std: np.ndarray,
frequentist_ece_std: np.ndarray,
ece_true: float,
ylim: float,
plot_kwargs: Dict[str, Any] = {},
plot_errorbar: bool = False) -> mpl.axes.Axes:
"""
Plot comparison of calibration estimation error obtained with the Bayesian or frequentist method.
:param ax:
:param N_list: List[int]
:param bayesian_ece: (len(N_list), ) as type float.
Estimation of ECE obtained with the Bayesian method.
:param frequentist_ece: (len(N_list), ) as type float.
Estimation of ECE obtained with the frequentist method.
:param bayesian_ece_std: (len(N_list), ) as type float.
Standard deviation of estimation of ECE obtained with the Bayesian method.
:param frequentist_ece_std: (len(N_list), ) as type float.
Standard deviation of estimation of ECE obtained with the frequentist method.
:param ece_true: float.
Ground truth ECE
:param plot_kwargs: dict.
Keyword arguments passed to the plot.
:return: ax: the generated matplotlib ax
"""
_plot_kwargs = DEFAULT_PLOT_KWARGS.copy()
_plot_kwargs.update(plot_kwargs)
if plot_errorbar:
# todo: debug
ax.errorbar(N_list, (bayesian_ece - ece_true) / ece_true * 100,
bayesian_ece_std / ece_true * 100,
'-*',
**_plot_kwargs,
label='Bayesian',
color='tab:red')
ax.errorbar(N_list, (frequentist_ece - ece_true) / ece_true * 100,
frequentist_ece_std / ece_true * 100,
'-o',
**_plot_kwargs,
label='Frequentist',
color='tab:blue')
else:
ax.plot(N_list, (bayesian_ece - ece_true) / ece_true * 100,
'-*',
**_plot_kwargs,
label='Bayesian',
color='tab:red')
ax.plot(N_list, (frequentist_ece - ece_true) / ece_true * 100,
'-o',
**_plot_kwargs,
label='Frequentist',
color='tab:blue')
ax.set_xscale('log')
ax.set_xlabel('#queries', labelpad=0.2)
ax.xaxis.set_ticks(N_list)
ax.set_ylim(ymin=0, ymax=ylim)
ax.set_yticks((0, ylim / 2, ylim))
ax.yaxis.set_major_formatter(PercentFormatter(decimals=0))
ax.tick_params(pad=0.25, length=1.5)
return ax
