Utility Functions
Based on: Utilities for developer (opens in a new tab)
Assert
validation.assert_all_finite
: Throw an error if array contains NaNs or Infs._testing.assert_allclose
: quasi equality of arrays, using atol
parameter
Formater / converter
validation.as_float_array
: convert input to array of floatvalidation.check_array
:- check that input is a 2D array. Sparse matrix and other dimensions can be optionally allowed
- call
assert_all_finite
validation.check_X_y
:- check that X and y have consistent lengths
- call
check_array
on X,column_or_1d
on y (usemulti_output=True
for multilabel y)
validation.indexable
: check that input arrays have consistent length and can be sliced or indexed (useful for cross-validation)- do not use
np.asanyarray
ornp.atleast_2d
since it letsnp.matrix
through (different API fromnp.array
)
Random state
validation.check_random_state
: randomness must be handle withnp.random.RandomState
only
Estimators
validation.check_is_fitted
: check that estimator has been fitted before calling predict or transform methodvalidation.has_fit_parameter
: check that the fit method has a given parameterall_estimators
: return a list of all scikit-learn estimatorsmulticlass.type_of_target
: return the type of the target y among continuous, continuous-multioutput, binary, multiclass, multiclass-multioutput, multilabel-indicator or unknownmulticlass.unique_labels
: Extract an ordered array of unique labels, "multiclass-multioutput" input type not allowed.
Warnings & Exceptions
utils.deprecated
: decorator to mark a function or class as deprecatedexceptions.ConvergenceWarning
: Custom warning to catch convergence problems (used in Lasso)