Utils

ife_surrogate.utils.train_test_split(X, Y, f=Array([], shape=(0,), dtype=float32), dense=False, *, split=(0.8, 0.2, 0.0), key)

Splits the dataset into training, testing, and validation sets.

Parameters:

  • X (Array) – Input feature array.

  • Y (Array) – Target value array.

  • f (Array) – Additional feature array.

  • dense (bool, optional) – If True, the input features are repeated and concatenated with f. Default is False.

  • split (Tuple[float, float, float], optional) – Ratios for splitting the data into (train, test, validation). Default is (0.8, 0.2, 0.0).

  • key (Key) – Random key for shuffling the data.

Returns:

A tuple of three pairs: (X_train, Y_train), (X_test, Y_test), and (X_val, Y_val).

Return type:

Tuple[Tuple[Array, Array], Tuple[Array, Array], Tuple[Array, Array]]

ife_surrogate.utils.nrms(test_y, prediction)

Calculates the Normalized Root Mean Square Error.

\[\text{NRMS} = \frac{\sqrt{\frac{1}{N}\sum_{i=1}^{N} (y_i - \hat{y}_i)^2}}{\max(y) - \min(y)}\]
Parameters:

  • test_y (ndarray) – The true output values of shape (n, p).

  • prediction (ndarray) – The predicted output values of shape (n, p).

Returns:

The Root Mean Square Percentage Error.

Return type:

float

ife_surrogate.utils.rmse(test_y, prediction)

Calculates the Root Mean Square Error.

\[RMSPE = \sqrt{\frac{1}{n} \sum_{i=1}^{n} \left( \frac{y_i - \hat{y}_i}{y_i} \right)^2}\]
Parameters:

  • test_y (ndarray) – The true output values of shape (n, p).

  • prediction (ndarray) – The predicted output values of shape (n, p).

Returns:

The Root Mean Square Error.

Return type:

Array

class ife_surrogate.utils.Visualizer

Bases: object

Container class for useful plotting functions. Can also export figures to a PDF-File.

save_to_page(figures, file_name='output.pdf')

Function to export matplotlib figures to PDF

Parameters:

  • figures (iterable[matplotlib.pyplot.figure]) – List of matplotlib figures, each figure adds a new page to the output PDF Also allows single figure.

  • file_name (str, optional) – Path and name of the PDF Defaults to “output.pdf”.

plot_bounded_error(x, y, error, function_color='blue', error_color='lightblue', highlight_threshold=0, highlight_color='red')

Default bounded error plot.

Parameters:

  • x (_type_) – _description_

  • y (_type_) – _description_

  • error (_type_) – _description_

Returns:

figure can be adjusted afterwards.

Return type:

matplotlib.pyplot.figure

plot_sns()
plot_testset(frequency, test_y, prediction, var_pred=None, grid=(4, 4), figsize=(15, 15), log_scale=True, error='nrms')

Plot predicted vs true curves for a test set.

Each subplot shows the true and predicted frequency response for one test sample. If predictive variance is given, 95% confidence intervals (±2 std) are shown.

Parameters:

  • frequency (Array) – 1D array of frequency values of shape (p,).

  • test_y (Array) – True output values of shape (n_test, p).

  • prediction (Array) – Predicted output values of shape (n_test, p).

  • var_pred (Array, optional) – Predictive variances of shape (n_test, p). If provided, uncertainty bands are shown.

  • grid (tuple) – Grid size as (rows, cols) to arrange the subplots.

  • figsize (tuple) – Size of the overall figure.

  • log_scale (bool) – If True, x-axis is displayed in logarithmic scale.

  • error (str) – Type of error metric to show in the title of each plot. Currently only supports “rmse” and “nrms.

Returns:

The generated matplotlib figure.

Return type:

matplotlib.figure.Figure

static quickplot(data_dict, fig_height=2, fig_width=7)
Parameters:

data_dict (dict)

ife_surrogate.utils.apply_fsv(Y_test, Y_pred, score_strings=False)

Batch FSV: apply across multiple test/pred pairs. Calculates the FSV parameters as described in [1]

[1] A. P. Duffy, A. J. M. Martin, A. Orlandi, G. Antonini, T. M. Benson, and M. S. Woolfson, “Feature Selective Validation (FSV) for Validation of Computational Electromagnetics (CEM). Part I—The FSV Method,” IEEE Trans. Electromagnetic Compatibility, vol. 48, no. 3, pp. 449-458, Aug. 2006, doi:10.1109/TEMC.2006.879358

Parameters:

  • Y_test (ndarray) – array of shape (M, N) or list of arrays

  • Y_pred (ndarray) – same shape as Y_test

  • score_string – returns the labels instead of the numerical scores

  • score_strings (bool)

Returns:

arrays of length M

Return type:

ADM_list, FDM_list, GDM_list

ife_surrogate.utils.zero_mean_unit_var_axis0(Y, eps=1e-08)

Standardize Y along axis 0 (samples) so each feature/output has zero mean and unit variance.

Parameters:

  • Y (array, shape (N, P)) – Input data (N samples, P outputs).

  • eps (float) – Small constant to avoid division by zero.

Returns:

  • Y_scaled (array, shape (N, P)) – Standardized data.

  • mean (array, shape (1, P)) – Mean of each output across samples.

  • std (array, shape (1, P)) – Std of each output across samples.

class ife_surrogate.utils.MinMaxScaler(feature_range=(0.0, 1.0), eps=1e-08)

Bases: object

fit(Y, axis=None)

Compute min and max for scaling.

transform(Y)

Scale data to the feature range.

inverse_transform(Y_scaled)

Revert back to the original scale.