viz module

This section contains the Python API reference for the uncertainty_toolbox.viz module, which contains code for visualizing uncertainties.

uncertainty_toolbox.viz Module

Visualizations for predictive uncertainties and metrics.

uncertainty_toolbox.viz.filter_subset(input_list, n_subset)

Keep only n_subset random indices from all lists given in input_list.

Parameters

input_list (List[List[Any]]) – list of lists.
n_subset (int) – Number of points to plot after filtering.

Return type

List[List[Any]]

Returns

List of all input lists with sizes reduced to n_subset.

uncertainty_toolbox.viz.plot_adversarial_group_calibration(y_pred, y_std, y_true, n_subset=None, cali_type='mean_abs', curve_label=None, group_size=None, score_mean=None, score_stderr=None, ax=None)

Plot adversarial group calibration plots by varying group size from 0% to 100% of dataset size and recording the worst calibration occurred for each group size.

Parameters

y_pred (ndarray) – 1D array of the predicted means for the held out dataset.
y_std (ndarray) – 1D array of the predicted standard deviations for the held out dataset.
y_true (ndarray) – 1D array of the true labels in the held out dataset.
n_subset (Optional[int]) – Number of points to plot after filtering.
cali_type (str) – Calibration type str.
curve_label (Optional[str]) – legend label str for calibration curve.
group_size (Optional[ndarray]) – 1D array of group size ratios in [0, 1].
score_mean (Optional[ndarray]) – 1D array of metric means for group size ratios in group_size.
score_stderr (Optional[ndarray]) – 1D array of metric standard devations for group size ratios in group_size.
ax (Optional[Axes]) – matplotlib.axes.Axes object.

Return type

Axes

Returns

matplotlib.axes.Axes object with plot added.

uncertainty_toolbox.viz.plot_calibration(y_pred, y_std, y_true, n_subset=None, curve_label=None, show=False, vectorized=True, exp_props=None, obs_props=None, ax=None)

Plot the observed proportion vs prediction proportion of outputs falling into a range of intervals, and display miscalibration area.

Parameters

y_pred (ndarray) – 1D array of the predicted means for the held out dataset.
y_std (ndarray) – 1D array of the predicted standard deviations for the held out dataset.
y_true (ndarray) – 1D array of the true labels in the held out dataset.
n_subset (Optional[int]) – Number of points to plot after filtering.
curve_label (Optional[str]) – legend label str for calibration curve.
vectorized (bool) – plot using get_proportion_lists_vectorized.
exp_props (Optional[ndarray]) – plot using the given expected proportions.
obs_props (Optional[ndarray]) – plot using the given observed proportions.
ax (Optional[Axes]) – matplotlib.axes.Axes object.

Return type

Axes

Returns

matplotlib.axes.Axes object with plot added.

uncertainty_toolbox.viz.plot_intervals(y_pred, y_std, y_true, n_subset=None, ylims=None, num_stds_confidence_bound=2, ax=None)

Plot predictions and predictive intervals versus true values.

Parameters

y_pred (ndarray) – 1D array of the predicted means for the held out dataset.
y_std (ndarray) – 1D array of the predicted standard deviations for the held out dataset.
y_true (ndarray) – 1D array of the true labels in the held out dataset.
n_subset (Optional[int]) – Number of points to plot after filtering.
ylims (Optional[Tuple[float, float]]) – a tuple of y axis plotting bounds, given as (lower, upper).
num_stds_confidence_bound (int) – width of intervals, in terms of number of standard deviations.
ax (Optional[Axes]) – matplotlib.axes.Axes object.

Return type

Axes

Returns

matplotlib.axes.Axes object with plot added.

uncertainty_toolbox.viz.plot_intervals_ordered(y_pred, y_std, y_true, n_subset=None, ylims=None, num_stds_confidence_bound=2, ax=None)

Plot predictions and predictive intervals versus true values, with points ordered by true value along x-axis.

Parameters

y_pred (ndarray) – 1D array of the predicted means for the held out dataset.
y_std (ndarray) – 1D array of the predicted standard deviations for the held out dataset.
y_true (ndarray) – 1D array of the true labels in the held out dataset.
n_subset (Optional[int]) – Number of points to plot after filtering.
ylims (Optional[Tuple[float, float]]) – a tuple of y axis plotting bounds, given as (lower, upper).
num_stds_confidence_bound (int) – width of intervals, in terms of number of standard deviations.
ax (Optional[Axes]) – matplotlib.axes.Axes object.

Return type

Axes

Returns

matplotlib.axes.Axes object with plot added.

uncertainty_toolbox.viz.plot_residuals_vs_stds(y_pred, y_std, y_true, n_subset=None, ax=None)

Plot absolute value of the prediction residuals versus standard deviations of the predictive uncertainties.

Parameters

y_pred (ndarray) – 1D array of the predicted means for the held out dataset.
y_std (ndarray) – 1D array of the predicted standard deviations for the held out dataset.
y_true (ndarray) – 1D array of the true labels in the held out dataset.
n_subset (Optional[int]) – Number of points to plot after filtering.
ax (Optional[Axes]) – matplotlib.axes.Axes object.

Return type

Axes

Returns

matplotlib.axes.Axes object with plot added.

uncertainty_toolbox.viz.plot_sharpness(y_std, n_subset=None, ax=None)

Plot sharpness of the predictive uncertainties.

Parameters

y_std (ndarray) – 1D array of the predicted standard deviations for the held out dataset.
n_subset (Optional[int]) – Number of points to plot after filtering.
ax (Optional[Axes]) – matplotlib.axes.Axes object.

Return type

Axes

Returns

matplotlib.axes.Axes object with plot added.

uncertainty_toolbox.viz.plot_xy(y_pred, y_std, y_true, x, n_subset=None, ylims=None, xlims=None, num_stds_confidence_bound=2, leg_loc=3, ax=None)

Plot one-dimensional inputs with associated predicted values, predictive uncertainties, and true values.

Parameters

y_pred (ndarray) – 1D array of the predicted means for the held out dataset.
y_std (ndarray) – 1D array of the predicted standard deviations for the held out dataset.
y_true (ndarray) – 1D array of the true labels in the held out dataset.
x (ndarray) – 1D array of input values for the held out dataset.
n_subset (Optional[int]) – Number of points to plot after filtering.
ylims (Optional[Tuple[float, float]]) – a tuple of y axis plotting bounds, given as (lower, upper).
xlims (Optional[Tuple[float, float]]) – a tuple of x axis plotting bounds, given as (lower, upper).
num_stds_confidence_bound (int) – width of confidence band, in terms of number of standard deviations.
leg_loc (Union[int, str]) – location of legend as a str or legend code int.
ax (Optional[Axes]) – matplotlib.axes.Axes object.

Return type

Axes

Returns

matplotlib.axes.Axes object with plot added.

uncertainty_toolbox.viz.save_figure(file_name='figure', ext_list=None, white_background=True)

Save matplotlib figure for all extensions in ext_list.

Parameters

file_name (str) – name of saved image file.
ext_list (Union[list, str, None]) – list of strings (or single string) denoting file type.
white_background (bool) – set background of image to white if True.

Return type

NoReturn

uncertainty_toolbox.viz.set_style(style_str='default')

Set the matplotlib plotting style.

Parameters: style_str (str) – string for style file.
Return type: NoReturn

uncertainty_toolbox.viz.update_rc(key_str, value)

Update matplotlibrc parameters.

Parameters

key_str (str) – string for a matplotlibrc parameter.
value (Any) – associated value to set the matplotlibrc parameter.

Return type

NoReturn