spectrochempy.MSCTransformer

class MSCTransformer(reference=None, dim='y')

Multiplicative Scatter Correction (MSC) transformer.

Learns a reusable reference spectrum during fit(). During transform(), each observation in the dataset being transformed is locally regressed against that reference and corrected with \((x - a) / b\).

inverse_transform() is not supported. The regression coefficients are local to each transformed dataset and are not safe reusable state.

Parameters:

• reference (NDDataset or array-like, optional) – 1-D reference spectrum. If None, the mean spectrum is used.

• dim (str or int, optional, default:’y’) – Dimension that identifies individual observations (spectra).

reference_

Reference spectrum used for fitting.

Type:

ndarray

a_

Intercepts of the per-observation regressions for the dataset passed to fit(). These are diagnostics, not reusable transform state.

Type:

ndarray

b_

Slopes of the per-observation regressions for the dataset passed to fit(). These are diagnostics, not reusable transform state.

Type:

ndarray

Raises:

SpectroChemPyError – If the dataset is not 2-D, the reference or transform spectral geometry is incompatible, a spectrum has too few valid paired points, the effective reference is constant, the local slope is zero, or inverse_transform() is requested.

Examples

>>> scaler = scp.MSCTransformer()
>>> scaler.fit(train)
>>> test_msc = scaler.transform(test)

See also

msc

Procedural MSC function.

Methods Summary

fit(dataset)	Learn parameters from dataset.
fit_transform(dataset)	Fit to dataset, then transform it.
get_params([deep])	Get the constructor parameters of this transformer.
inverse_transform(dataset)	Reverse the learned transformation on dataset.
set_params(**params)	Set constructor parameters on this transformer.
transform(dataset)	Apply the learned transformation to dataset.

Methods Documentation

fit(dataset)

Learn parameters from dataset.

Parameters:

dataset (NDDataset) – Training data.

Returns:

self – The fitted instance.

fit_transform(dataset)

Fit to dataset, then transform it.

Equivalent to self.fit(dataset).transform(dataset) but avoids an intermediate copy when possible.

Parameters:

dataset (NDDataset) – Training data.

Returns:

NDDataset – Transformed dataset.

get_params(deep=True)

Get the constructor parameters of this transformer.

Parameters:

deep (bool, optional, default:True) – Ignored. Present for compatibility with scikit-learn conventions.

Returns:

dict – Mapping of parameter name -> current value.

Examples

>>> scaler = scp.AutoscaleTransformer(dim="y")
>>> scaler.get_params()
{'dim': 'y'}

inverse_transform(dataset)

Reverse the learned transformation on dataset.

Parameters:

dataset (NDDataset) – Data to invert.

Returns:

NDDataset – Dataset in the original space.

Raises:

SpectroChemPyError – If fit() has not been called first.

set_params(**params)

Set constructor parameters on this transformer.

Returns self so that calls can be chained.

Parameters:

**params – Parameter names and values to update.

Returns:

self

Raises:

SpectroChemPyError – If a parameter name does not correspond to a constructor argument.

Examples

>>> scaler = scp.AutoscaleTransformer(dim="y")
>>> scaler.set_params(dim="x")
AutoscaleTransformer(dim='x')

transform(dataset)

Apply the learned transformation to dataset.

Parameters:

dataset (NDDataset) – Data to transform.

Returns:

NDDataset – Transformed dataset.

Raises:

SpectroChemPyError – If fit() has not been called first.

Examples using spectrochempy.MSCTransformer