bayesvalidrox.post_processing.post_processing.PostProcessing

class bayesvalidrox.post_processing.post_processing.PostProcessing(engine, name='calib', out_dir='', out_format='pdf')

Bases: object

This class provides post-processing functions for the trained metamodels.

Parameters

engineobj

Trained Engine object, is expected to contain a trained MetaModel object.

namestring

Name of the PostProcessing object to be used for saving the generated files. The default is ‘calib’.

out_dirstring

Output directory in which the PostProcessing results are placed. The results are contained in a subfolder ‘/Outputs_PostProcessing_name’ The default is ‘’.

out_formatstring

Format of the saved plots. Supports ‘png’ and ‘pdf’. The default is ‘pdf’.

Raises

AttributeError

engine must be trained.

__init__(engine, name='calib', out_dir='', out_format='pdf')

Methods

__init__(engine[, name, out_dir, out_format])
plot_correl(y_model, y_mm[, r_2, name])	Plot the correlation between the model and metamodel outputs.
plot_expdesign([n_mc, show_samples])	Visualizes training samples over their given distributions as a pairplot.
plot_moments([plot_type])	Plots the moments in a user defined output format (standard is pdf) in the directory Outputs_PostProcessing.
plot_residual_hist(y, y_mm)	Checks the quality of the metamodel residuals via visualization and Normality (Shapiro-Wilk) test.
plot_seq_design_diagnostics([plot_single, ...])	Plots the validation metrics calculated in the engine.
plot_sobol(sobol_values, par_names, outputs)	Generalized function to plot Sobol' indices from PCE models or SALib.
plot_validation_outputs(model_out, out_mean)	Plots outputs for visual comparison of metamodel outputs with that of the (full) multioutput original model
sobol_indices([plot_type, save, plot])	Visualizes and writes out Sobol' and Total Sobol' indices of the trained metamodel.
validate_metamodel([n_samples, sampling_method])	Evaluates all available validation metrics for the engine on validation samples and visualizes the results.

plot_correl(y_model, y_mm, r_2=None, name='valid') → None

Plot the correlation between the model and metamodel outputs.

Parameters

y_modeldict

Model evaluations.

y_mmdict

MetaModel evaluations.

r_2dict, optional

R2 score for each output key (as a list per key). The default is None.

namestring, optional

Name of the file. The default is ‘valid’.

plot_expdesign(n_mc=10000, show_samples=False)

Visualizes training samples over their given distributions as a pairplot.

Parameters

n_mcint, optional

Number of samples from the priors to use for plotting. The default is 10000.

show_samplesbool, optional

If set to True the training samples are also visualized. The default is False.

Returns

None.

plot_moments(plot_type: str = 'line')

Plots the moments in a user defined output format (standard is pdf) in the directory Outputs_PostProcessing.

Parameters

plot_typestr, optional

Supports ‘bar’ for barplots and ‘line’ for lineplots The default is line.

Raises

AttributeError

Plot type must be ‘bar’ or ‘line’.

Returns

means: dict

Mean of the model outputs.

means: dict

Standard deviation of the model outputs.

plot_residual_hist(y, y_mm) → None

Checks the quality of the metamodel residuals via visualization and Normality (Shapiro-Wilk) test.

Parameters

ydict

Model evaluations.

y_mmdict

Corresponding MetaModel evaluations

plot_seq_design_diagnostics(plot_single=True, metrics=None, name='engine') → None

Plots the validation metrics calculated in the engine.

Parameters

plot_singlebool

If set to True, generates a single file with all the results. If set to False, generates a file for each validation type.

metricslist

List of the metrics dictionaries, if generated in validation and not as part of the engine.

namestr, optional

Name of the plot. Will be ‘post’ if used from self.validate_metamodel. The default is ‘engine’.

plot_sobol(sobol_values: dict, par_names: list, outputs: list, sobol_type: str = 'sobol', i_order: int = 1, x: ndarray | None = None, xlabel: str = 'Time', plot_type: str = 'line', out_dir: str = './', out_format: str = 'png', quantiles: tuple | None = None)

Generalized function to plot Sobol’ indices from PCE models or SALib.

Parameters

sobol_valuesdict

Dictionary of Sobol’ indices for each output.

par_nameslist

List of parameter names or interaction tuples.

outputslist

List of output variable names.

sobol_typestr

‘sobol’ or ‘totalsobol’.

i_orderint

Order of Sobol’ index (only for sobol_type=’sobol’).

xnp.ndarray, optional

X-axis values (e.g., time). If None, defaults to index range.

xlabelstr

Label for x-axis.

plot_typestr

‘line’ or ‘bar’.

out_dirstr

Output directory for plots.

out_formatstr

File format for saving plots.

quantilestuple, optional

Tuple of (q5, q975) quantile arrays for confidence intervals.

plot_validation_outputs(model_out, out_mean, out_std=None)

Plots outputs for visual comparison of metamodel outputs with that of the (full) multioutput original model

Parameters

model_outdict

Model outputs.

out_meandict

MetaModel mean outputs.

out_stddict

MetaModel stdev outputs.

Raises

AttributeError: This evaluation only support PCE-type models!

Returns

None

sobol_indices(plot_type: str = 'line', save: bool = True, plot: bool = True)

Visualizes and writes out Sobol’ and Total Sobol’ indices of the trained metamodel. One file is created for each index and output key.

Parameters

plot_typestr, optional

Plot type, supports ‘line’ for lineplots and ‘bar’ for barplots. The default is line. Bar chart can be selected by bar.

savebool, optional

Write out the inidces as csv files if set to True. The default is True.

Raises

AttributeError

MetaModel in given Engine needs to be of type ‘pce’ or ‘apce’.

AttributeError

Plot-type must be ‘line’ or ‘bar’.

Returns

sobol_alldict

All possible Sobol’ indices for the given metamodel.

total_sobol_alldict

All Total Sobol’ indices for the given metamodel.

validate_metamodel(n_samples=None, sampling_method='random')

Evaluates all available validation metrics for the engine on validation samples and visualizes the results.

Parameters

n_samplesint, optional

Number of validation samples to generate. If this is set to None, then the validation samples in the exp_design are used. The default is None.

sampling_methodstr, optional

Sampling method. The default is ‘random’.

Returns

valid, bayesdict, dict

Evaluation of the applicable metrics.