auswahl.IntervalRandomFrog

class auswahl.IntervalRandomFrog(n_intervals_to_select: int = 1, interval_width: Optional[Union[int, float]] = None, n_iterations: int = 10000, n_initial_intervals: Union[int, float] = 0.1, variance_factor: float = 0.3, subset_expansion_factor: float = 3, acceptance_factor: float = 0.1, n_cv_folds: int = 5, n_jobs: int = 1, pls: Optional[PLSRegression] = None, model_hyperparams: Optional[Union[Dict, List[Dict]]] = None, random_state: Optional[Union[int, RandomState]] = None)

Feature selection with the Interval Random Frog (iRF) method.

The selection frequencies are computed according to Yun et al. [1].

Read more in the User Guide.

Parameters

n_intervals_to_selectint, default=1

Number of intervals to select.

interval_width: int or float, default=None

Size of the selected intervals. If None, the intervals are n_features/2 long. If integer, the parameter directly defines the number of consecutive features that form an interval. If float between 0 and 1, the intervals are n_features*n_intervals_to_select long.

n_iterationsint, default=10000

Number of variable selection iterations. This variable is called N in the original publication.

n_initial_intervalsint or float, default=0.1

Number of intervals in the initial interval subset. If None, 10 % of the intervals are used. If integer, the parameter is the size of the initial subset. If float between 0 and 1, it is the fraction of intervals to use. This variable is called Q in the original publication.

variance_factorfloat, default=0.3

Variance of the normal distribution which samples determine the amount of intervals that are added or removed to the candidate set in each iteration. This variable is called θ in the original publication.

subset_expansion_factorfloat, default=3

Multiple of the number of intervals that are explored if the candidate subset is expanded. If the current interval subset is n and m new intervals have to be added to the new interval subset, m*subset_expansion_factors intervals are added to a candidate set. After fitting a PLS model, only the n+m intervals with the highest coefficients are kept. This variable is called ω in the original publication.

acceptance_factorfloat, default=0.1

The factor is used to calculate the probability that an interval subset is selected even though it leads to a worse cross-validation performance of a fitted PLS model. The probability is computed by multiplying the acceptance_factor with the relative decrease of the cross-validated performance score. This variable is called η in the original publication.

n_cv_foldsint, default=5

Number of cross validation folds used to evaluate the features.

n_jobsint, default=1

Number of parallel processes used to fit the PLS models on the cross-validation splits.

plsPLSRegression, default=None

Estimator instance of the PLSRegression class. Use this to adjust the hyperparameters of the PLS method.

random_stateint or numpy.random.RandomState, default=None

Seed for the random subset sampling. Pass an int for reproducible output across function calls.

Attributes

frequencies_ndarray of shape (n_features,)

Number of times each interval has been selected after all iterations.

support_ndarray of shape (n_features,)

Mask of selected intervals.

References

1

Yong-Huan Yun and Hong-Dong Li and Leslie R. E. Wood and Wei Fan and Jia-Jun Wang and Dong-Sheng Cao and Qing-Song Xu and Yi-Zeng Liang, ‘An efficient method of wavelength interval selection based on random frog for multivariate spectral calibration’, Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 111, 31-36, 2013

Examples

>>> import numpy as np
>>> from auswahl import IntervalRandomFrog
>>> np.random.seed(1337)
>>> X = np.random.randn(100, 10)
>>> y = 5 * X[:, 0] - 3 * X[:, 1] + 2 * X[:, 5] - 3 * X[:, 6]  # y only depends on two intervals
>>> selector = IntervalRandomFrog(n_intervals_to_select=2, interval_width=2, n_iterations=1000, random_state=7331)
>>> selector.fit(X, y).get_support()
array([ True, True, False, False, False, True, True, False, False, False])

__init__(n_intervals_to_select: int = 1, interval_width: Optional[Union[int, float]] = None, n_iterations: int = 10000, n_initial_intervals: Union[int, float] = 0.1, variance_factor: float = 0.3, subset_expansion_factor: float = 3, acceptance_factor: float = 0.1, n_cv_folds: int = 5, n_jobs: int = 1, pls: Optional[PLSRegression] = None, model_hyperparams: Optional[Union[Dict, List[Dict]]] = None, random_state: Optional[Union[int, RandomState]] = None)

evaluate(X, y, model, do_cv=True, *args)

Conduct a cross validationand hyperparameter optimization of the underlying estimator model.

Parameters

X: array-like, shape (n_samples, n_features)

Spectral data to be fitted

y: array-like, shape (n_samples,)

Regression targets

model: BaseEstimator

Regression model

do_cv: bool, default=True

If True, the model is fitted to the data and a cross validation score is provided

*args: arbitrary payload

Arbitrary payload returned with the evaluation result. Used for instance for identification of threads, if multiple models are evaluated in parallel

Returns

tuple: float, BaseEstimator

cross validation score if requested (otherwise None) and fitted estimator

fit(X, y, mask=None)

Run the feature selection process.

Parameters

Xarray-like of shape (n_samples, n_features)

The input samples.

yarray-like of shape (n_samples,)

The target values.

mask: array-like of shape (n_features,)

Mask indicating (values == 0), which features are not to be taken into account during the feature selection

Returns

SpectralSelectorself

Returns the instance itself.

fit_transform(X, y=None, **fit_params)

Fit to data, then transform it.

Fits transformer to X and y with optional parameters fit_params and returns a transformed version of X.

Parameters

Xarray-like of shape (n_samples, n_features)

Input samples.

yarray-like of shape (n_samples,) or (n_samples, n_outputs), default=None

Target values (None for unsupervised transformations).

**fit_paramsdict

Additional fit parameters.

Returns

X_newndarray array of shape (n_samples, n_features_new)

Transformed array.

get_best_estimator() → BaseEstimator

Retrieve the best estimator model fitted on the selected features

Returns

best model fitted on selected features: sklearn.base.BaseEstimator

get_feature_names_out(input_features=None)

Mask feature names according to selected features.

Parameters

input_featuresarray-like of str or None, default=None

Input features.

• If input_features is None, then feature_names_in_ is used as feature names in. If feature_names_in_ is not defined, then the following input feature names are generated: [“x0”, “x1”, …, “x(n_features_in_ - 1)”].

• If input_features is an array-like, then input_features must match feature_names_in_ if feature_names_in_ is defined.

Returns

feature_names_outndarray of str objects

Transformed feature names.

get_params(deep=True)

Get parameters for this estimator.

Parameters

deepbool, default=True

If True, will return the parameters for this estimator and contained subobjects that are estimators.

Returns

paramsdict

Parameter names mapped to their values.

get_support(indices=False)

Get a mask, or integer index, of the features selected.

Parameters

indicesbool, default=False

If True, the return value will be an array of integers, rather than a boolean mask.

Returns

supportarray

An index that selects the retained features from a feature vector. If indices is False, this is a boolean array of shape [# input features], in which an element is True iff its corresponding feature is selected for retention. If indices is True, this is an integer array of shape [# output features] whose values are indices into the input feature vector.

inverse_transform(X)

Reverse the transformation operation.

Parameters

Xarray of shape [n_samples, n_selected_features]

The input samples.

Returns

X_rarray of shape [n_samples, n_original_features]

X with columns of zeros inserted where features would have been removed by transform().

reseed(seed: Union[int, RandomState])

Random state updating interface for benchmarking. Selector methods with more complex internal structure (such as methods wrapping other methods) are required to override this function accordingly.

rethread(n_jobs: int)

n_jobs updating interface for benchmarking. Selector methods with more complex internal structure (such as methods wrapping other methods) are required to override this function accordingly.

set_params(**params)

Set the parameters of this estimator.

The method works on simple estimators as well as on nested objects (such as Pipeline). The latter have parameters of the form <component>__<parameter> so that it’s possible to update each component of a nested object.

Parameters

**paramsdict

Estimator parameters.

Returns

selfestimator instance

Estimator instance.

transform(X)

Reduce X to the selected features.

Parameters

Xarray of shape [n_samples, n_features]

The input samples.

Returns

X_rarray of shape [n_samples, n_selected_features]

The input samples with only the selected features.

Examples using auswahl.IntervalRandomFrog

IntervalRandomFrog - Basic example

IntervalRandomFrog - Basic example