Module for scaling data before and after interpolation.

class posydon.interpolation.data_scaling.DataScaler[source]

Bases: object

Data Normalization class.

This class provides normalizing tools for float 1D arrays. Features can be standarized or scaled to a range depending on the method chosen when calling the fit ot fit_and_transform functions.

Initialize the data scaler.

No parameters are expected. After instantiation use methods fit or fit_and_transform first to fit a scaling object to a given vector of values.

Example

>>> sc = DataScaler()
fit(x, method='none', lower=- 1.0, upper=1.0)[source]

Fit a transform of 1D numpy array x.

Computes the parameters that define the transform.

Parameters

  • x (numpy.ndarray) – expects a 1D array and finds norm values of columns

  • method (str) – Scaling method. Possible values: ‘min_max’, ‘max_abs’, ‘standarize’ and their log versions ‘log_min_max’, ‘neg_log_min_max’, ‘log_max_abs’, ‘log_standarize’, ‘neg_log_standarize’.

  • lower (float) – lower range value of x_t after (log_)min_max scaling

  • upper (float) – upper range value of x_t after (log_)min_max scaling

fit_and_transform(x, method='none', lower=- 1, upper=1)[source]

Fit and transform the array x according to the chosen scaling.

lower/upper will only be taken into account for (log_)min_max normalization. In this case, the transformed x will have min(x_transf) = lower, max(x_transf) = upper

Parameters

  • x (numpy.ndarray) – expects a 1D array and finds norm values of columns

  • method (str) – scaling method. Possible values: ‘min_max’, ‘max_abs’, ‘standarize’ and the log versions ‘log_min_max’, ‘log_max_abs’, ‘log_standarize’

  • lower (float) – lower range value of x_t after (log_)min_max scaling

  • upper (float) – upper range value of x_t after (log_)min_max scaling

Returns

transformed version of x

Return type

numpy.ndarray

inv_transform(x_t)[source]

Revert the scaling using the stored transform parameters.

Parameters

x_t (numpy.ndarray) – expects a 1D array to unnormalize given the fitted transform.

Returns

denormalized x using the stored parameters.

Return type

numpy.ndarray

transform(x)[source]

Transform x using the already obtained normalization values.

self.fit()` must be called first. lower/upper will only be taken into account for (log_)min_max normalization. In this case, the transformed x will have min(x_transf) = lower, max(x_transf) = upper

Parameters

x (numpy.ndarray) – values to normalize

Returns

transformed version of x

Return type

numpy.ndarray