""" The :mod:`sklearn.utils.validation` module includes functions to validate input and parameters within scikit-learn estimators. """ # Authors: Olivier Grisel # Gael Varoquaux # Andreas Mueller # Lars Buitinck # Alexandre Gramfort # Nicolas Tresegnie # Sylvain Marie # License: BSD 3 clause import numbers import operator import sys import warnings from contextlib import suppress from functools import reduce, wraps from inspect import Parameter, isclass, signature import joblib import numpy as np import scipy.sparse as sp from .. import get_config as _get_config from ..exceptions import DataConversionWarning, NotFittedError, PositiveSpectrumWarning from ..utils._array_api import _asarray_with_order, _is_numpy_namespace, get_namespace from ..utils.fixes import ComplexWarning, _preserve_dia_indices_dtype from ._isfinite import FiniteStatus, cy_isfinite from .fixes import _object_dtype_isnan FLOAT_DTYPES = (np.float64, np.float32, np.float16) # This function is not used anymore at this moment in the code base but we keep it in # case that we merge a new public function without kwarg only by mistake, which would # require a deprecation cycle to fix. def _deprecate_positional_args(func=None, *, version="1.3"): """Decorator for methods that issues warnings for positional arguments. Using the keyword-only argument syntax in pep 3102, arguments after the * will issue a warning when passed as a positional argument. Parameters ---------- func : callable, default=None Function to check arguments on. version : callable, default="1.3" The version when positional arguments will result in error. """ def _inner_deprecate_positional_args(f): sig = signature(f) kwonly_args = [] all_args = [] for name, param in sig.parameters.items(): if param.kind == Parameter.POSITIONAL_OR_KEYWORD: all_args.append(name) elif param.kind == Parameter.KEYWORD_ONLY: kwonly_args.append(name) @wraps(f) def inner_f(*args, **kwargs): extra_args = len(args) - len(all_args) if extra_args <= 0: return f(*args, **kwargs) # extra_args > 0 args_msg = [ "{}={}".format(name, arg) for name, arg in zip(kwonly_args[:extra_args], args[-extra_args:]) ] args_msg = ", ".join(args_msg) warnings.warn( ( f"Pass {args_msg} as keyword args. From version " f"{version} passing these as positional arguments " "will result in an error" ), FutureWarning, ) kwargs.update(zip(sig.parameters, args)) return f(**kwargs) return inner_f if func is not None: return _inner_deprecate_positional_args(func) return _inner_deprecate_positional_args def _assert_all_finite( X, allow_nan=False, msg_dtype=None, estimator_name=None, input_name="" ): """Like assert_all_finite, but only for ndarray.""" xp, _ = get_namespace(X) if _get_config()["assume_finite"]: return X = xp.asarray(X) # for object dtype data, we only check for NaNs (GH-13254) if X.dtype == np.dtype("object") and not allow_nan: if _object_dtype_isnan(X).any(): raise ValueError("Input contains NaN") # We need only consider float arrays, hence can early return for all else. if not xp.isdtype(X.dtype, ("real floating", "complex floating")): return # First try an O(n) time, O(1) space solution for the common case that # everything is finite; fall back to O(n) space `np.isinf/isnan` or custom # Cython implementation to prevent false positives and provide a detailed # error message. with np.errstate(over="ignore"): first_pass_isfinite = xp.isfinite(xp.sum(X)) if first_pass_isfinite: return _assert_all_finite_element_wise( X, xp=xp, allow_nan=allow_nan, msg_dtype=msg_dtype, estimator_name=estimator_name, input_name=input_name, ) def _assert_all_finite_element_wise( X, *, xp, allow_nan, msg_dtype=None, estimator_name=None, input_name="" ): # Cython implementation doesn't support FP16 or complex numbers use_cython = ( xp is np and X.data.contiguous and X.dtype.type in {np.float32, np.float64} ) if use_cython: out = cy_isfinite(X.reshape(-1), allow_nan=allow_nan) has_nan_error = False if allow_nan else out == FiniteStatus.has_nan has_inf = out == FiniteStatus.has_infinite else: has_inf = xp.any(xp.isinf(X)) has_nan_error = False if allow_nan else xp.any(xp.isnan(X)) if has_inf or has_nan_error: if has_nan_error: type_err = "NaN" else: msg_dtype = msg_dtype if msg_dtype is not None else X.dtype type_err = f"infinity or a value too large for {msg_dtype!r}" padded_input_name = input_name + " " if input_name else "" msg_err = f"Input {padded_input_name}contains {type_err}." if estimator_name and input_name == "X" and has_nan_error: # Improve the error message on how to handle missing values in # scikit-learn. msg_err += ( f"\n{estimator_name} does not accept missing values" " encoded as NaN natively. For supervised learning, you might want" " to consider sklearn.ensemble.HistGradientBoostingClassifier and" " Regressor which accept missing values encoded as NaNs natively." " Alternatively, it is possible to preprocess the data, for" " instance by using an imputer transformer in a pipeline or drop" " samples with missing values. See" " https://scikit-learn.org/stable/modules/impute.html" " You can find a list of all estimators that handle NaN values" " at the following page:" " https://scikit-learn.org/stable/modules/impute.html" "#estimators-that-handle-nan-values" ) raise ValueError(msg_err) def assert_all_finite( X, *, allow_nan=False, estimator_name=None, input_name="", ): """Throw a ValueError if X contains NaN or infinity. Parameters ---------- X : {ndarray, sparse matrix} The input data. allow_nan : bool, default=False If True, do not throw error when `X` contains NaN. estimator_name : str, default=None The estimator name, used to construct the error message. input_name : str, default="" The data name used to construct the error message. In particular if `input_name` is "X" and the data has NaN values and allow_nan is False, the error message will link to the imputer documentation. Examples -------- >>> from sklearn.utils import assert_all_finite >>> import numpy as np >>> array = np.array([1, np.inf, np.nan, 4]) >>> try: ... assert_all_finite(array) ... print("Test passed: Array contains only finite values.") ... except ValueError: ... print("Test failed: Array contains non-finite values.") Test failed: Array contains non-finite values. """ _assert_all_finite( X.data if sp.issparse(X) else X, allow_nan=allow_nan, estimator_name=estimator_name, input_name=input_name, ) def as_float_array(X, *, copy=True, force_all_finite=True): """Convert an array-like to an array of floats. The new dtype will be np.float32 or np.float64, depending on the original type. The function can create a copy or modify the argument depending on the argument copy. Parameters ---------- X : {array-like, sparse matrix} The input data. copy : bool, default=True If True, a copy of X will be created. If False, a copy may still be returned if X's dtype is not a floating point type. force_all_finite : bool or 'allow-nan', default=True Whether to raise an error on np.inf, np.nan, pd.NA in X. The possibilities are: - True: Force all values of X to be finite. - False: accepts np.inf, np.nan, pd.NA in X. - 'allow-nan': accepts only np.nan and pd.NA values in X. Values cannot be infinite. .. versionadded:: 0.20 ``force_all_finite`` accepts the string ``'allow-nan'``. .. versionchanged:: 0.23 Accepts `pd.NA` and converts it into `np.nan` Returns ------- XT : {ndarray, sparse matrix} An array of type float. Examples -------- >>> from sklearn.utils import as_float_array >>> import numpy as np >>> array = np.array([0, 0, 1, 2, 2], dtype=np.int64) >>> as_float_array(array) array([0., 0., 1., 2., 2.]) """ if isinstance(X, np.matrix) or ( not isinstance(X, np.ndarray) and not sp.issparse(X) ): return check_array( X, accept_sparse=["csr", "csc", "coo"], dtype=np.float64, copy=copy, force_all_finite=force_all_finite, ensure_2d=False, ) elif sp.issparse(X) and X.dtype in [np.float32, np.float64]: return X.copy() if copy else X elif X.dtype in [np.float32, np.float64]: # is numpy array return X.copy("F" if X.flags["F_CONTIGUOUS"] else "C") if copy else X else: if X.dtype.kind in "uib" and X.dtype.itemsize <= 4: return_dtype = np.float32 else: return_dtype = np.float64 return X.astype(return_dtype) def _is_arraylike(x): """Returns whether the input is array-like.""" if sp.issparse(x): return False return hasattr(x, "__len__") or hasattr(x, "shape") or hasattr(x, "__array__") def _is_arraylike_not_scalar(array): """Return True if array is array-like and not a scalar""" return _is_arraylike(array) and not np.isscalar(array) def _use_interchange_protocol(X): """Use interchange protocol for non-pandas dataframes that follow the protocol. Note: at this point we chose not to use the interchange API on pandas dataframe to ensure strict behavioral backward compatibility with older versions of scikit-learn. """ return not _is_pandas_df(X) and hasattr(X, "__dataframe__") def _num_features(X): """Return the number of features in an array-like X. This helper function tries hard to avoid to materialize an array version of X unless necessary. For instance, if X is a list of lists, this function will return the length of the first element, assuming that subsequent elements are all lists of the same length without checking. Parameters ---------- X : array-like array-like to get the number of features. Returns ------- features : int Number of features """ type_ = type(X) if type_.__module__ == "builtins": type_name = type_.__qualname__ else: type_name = f"{type_.__module__}.{type_.__qualname__}" message = f"Unable to find the number of features from X of type {type_name}" if not hasattr(X, "__len__") and not hasattr(X, "shape"): if not hasattr(X, "__array__"): raise TypeError(message) # Only convert X to a numpy array if there is no cheaper, heuristic # option. X = np.asarray(X) if hasattr(X, "shape"): if not hasattr(X.shape, "__len__") or len(X.shape) <= 1: message += f" with shape {X.shape}" raise TypeError(message) return X.shape[1] first_sample = X[0] # Do not consider an array-like of strings or dicts to be a 2D array if isinstance(first_sample, (str, bytes, dict)): message += f" where the samples are of type {type(first_sample).__qualname__}" raise TypeError(message) try: # If X is a list of lists, for instance, we assume that all nested # lists have the same length without checking or converting to # a numpy array to keep this function call as cheap as possible. return len(first_sample) except Exception as err: raise TypeError(message) from err def _num_samples(x): """Return number of samples in array-like x.""" message = "Expected sequence or array-like, got %s" % type(x) if hasattr(x, "fit") and callable(x.fit): # Don't get num_samples from an ensembles length! raise TypeError(message) if _use_interchange_protocol(x): return x.__dataframe__().num_rows() if not hasattr(x, "__len__") and not hasattr(x, "shape"): if hasattr(x, "__array__"): x = np.asarray(x) else: raise TypeError(message) if hasattr(x, "shape") and x.shape is not None: if len(x.shape) == 0: raise TypeError( "Singleton array %r cannot be considered a valid collection." % x ) # Check that shape is returning an integer or default to len # Dask dataframes may not return numeric shape[0] value if isinstance(x.shape[0], numbers.Integral): return x.shape[0] try: return len(x) except TypeError as type_error: raise TypeError(message) from type_error def check_memory(memory): """Check that ``memory`` is joblib.Memory-like. joblib.Memory-like means that ``memory`` can be converted into a joblib.Memory instance (typically a str denoting the ``location``) or has the same interface (has a ``cache`` method). Parameters ---------- memory : None, str or object with the joblib.Memory interface - If string, the location where to create the `joblib.Memory` interface. - If None, no caching is done and the Memory object is completely transparent. Returns ------- memory : object with the joblib.Memory interface A correct joblib.Memory object. Raises ------ ValueError If ``memory`` is not joblib.Memory-like. Examples -------- >>> from sklearn.utils.validation import check_memory >>> check_memory("caching_dir") Memory(location=caching_dir/joblib) """ if memory is None or isinstance(memory, str): memory = joblib.Memory(location=memory, verbose=0) elif not hasattr(memory, "cache"): raise ValueError( "'memory' should be None, a string or have the same" " interface as joblib.Memory." " Got memory='{}' instead.".format(memory) ) return memory def check_consistent_length(*arrays): """Check that all arrays have consistent first dimensions. Checks whether all objects in arrays have the same shape or length. Parameters ---------- *arrays : list or tuple of input objects. Objects that will be checked for consistent length. Examples -------- >>> from sklearn.utils.validation import check_consistent_length >>> a = [1, 2, 3] >>> b = [2, 3, 4] >>> check_consistent_length(a, b) """ lengths = [_num_samples(X) for X in arrays if X is not None] uniques = np.unique(lengths) if len(uniques) > 1: raise ValueError( "Found input variables with inconsistent numbers of samples: %r" % [int(l) for l in lengths] ) def _make_indexable(iterable): """Ensure iterable supports indexing or convert to an indexable variant. Convert sparse matrices to csr and other non-indexable iterable to arrays. Let `None` and indexable objects (e.g. pandas dataframes) pass unchanged. Parameters ---------- iterable : {list, dataframe, ndarray, sparse matrix} or None Object to be converted to an indexable iterable. """ if sp.issparse(iterable): return iterable.tocsr() elif hasattr(iterable, "__getitem__") or hasattr(iterable, "iloc"): return iterable elif iterable is None: return iterable return np.array(iterable) def indexable(*iterables): """Make arrays indexable for cross-validation. Checks consistent length, passes through None, and ensures that everything can be indexed by converting sparse matrices to csr and converting non-interable objects to arrays. Parameters ---------- *iterables : {lists, dataframes, ndarrays, sparse matrices} List of objects to ensure sliceability. Returns ------- result : list of {ndarray, sparse matrix, dataframe} or None Returns a list containing indexable arrays (i.e. NumPy array, sparse matrix, or dataframe) or `None`. Examples -------- >>> from sklearn.utils import indexable >>> from scipy.sparse import csr_matrix >>> import numpy as np >>> iterables = [ ... [1, 2, 3], np.array([2, 3, 4]), None, csr_matrix([[5], [6], [7]]) ... ] >>> indexable(*iterables) [[1, 2, 3], array([2, 3, 4]), None, <3x1 sparse matrix ...>] """ result = [_make_indexable(X) for X in iterables] check_consistent_length(*result) return result def _ensure_sparse_format( sparse_container, accept_sparse, dtype, copy, force_all_finite, accept_large_sparse, estimator_name=None, input_name="", ): """Convert a sparse container to a given format. Checks the sparse format of `sparse_container` and converts if necessary. Parameters ---------- sparse_container : sparse matrix or array Input to validate and convert. accept_sparse : str, bool or list/tuple of str String[s] representing allowed sparse matrix formats ('csc', 'csr', 'coo', 'dok', 'bsr', 'lil', 'dia'). If the input is sparse but not in the allowed format, it will be converted to the first listed format. True allows the input to be any format. False means that a sparse matrix input will raise an error. dtype : str, type or None Data type of result. If None, the dtype of the input is preserved. copy : bool Whether a forced copy will be triggered. If copy=False, a copy might be triggered by a conversion. force_all_finite : bool or 'allow-nan' Whether to raise an error on np.inf, np.nan, pd.NA in X. The possibilities are: - True: Force all values of X to be finite. - False: accepts np.inf, np.nan, pd.NA in X. - 'allow-nan': accepts only np.nan and pd.NA values in X. Values cannot be infinite. .. versionadded:: 0.20 ``force_all_finite`` accepts the string ``'allow-nan'``. .. versionchanged:: 0.23 Accepts `pd.NA` and converts it into `np.nan` estimator_name : str, default=None The estimator name, used to construct the error message. input_name : str, default="" The data name used to construct the error message. In particular if `input_name` is "X" and the data has NaN values and allow_nan is False, the error message will link to the imputer documentation. Returns ------- sparse_container_converted : sparse matrix or array Sparse container (matrix/array) that is ensured to have an allowed type. """ if dtype is None: dtype = sparse_container.dtype changed_format = False sparse_container_type_name = type(sparse_container).__name__ if isinstance(accept_sparse, str): accept_sparse = [accept_sparse] # Indices dtype validation _check_large_sparse(sparse_container, accept_large_sparse) if accept_sparse is False: padded_input = " for " + input_name if input_name else "" raise TypeError( f"Sparse data was passed{padded_input}, but dense data is required. " "Use '.toarray()' to convert to a dense numpy array." ) elif isinstance(accept_sparse, (list, tuple)): if len(accept_sparse) == 0: raise ValueError( "When providing 'accept_sparse' as a tuple or list, it must contain at " "least one string value." ) # ensure correct sparse format if sparse_container.format not in accept_sparse: # create new with correct sparse sparse_container = sparse_container.asformat(accept_sparse[0]) changed_format = True elif accept_sparse is not True: # any other type raise ValueError( "Parameter 'accept_sparse' should be a string, boolean or list of strings." f" You provided 'accept_sparse={accept_sparse}'." ) if dtype != sparse_container.dtype: # convert dtype sparse_container = sparse_container.astype(dtype) elif copy and not changed_format: # force copy sparse_container = sparse_container.copy() if force_all_finite: if not hasattr(sparse_container, "data"): warnings.warn( f"Can't check {sparse_container.format} sparse matrix for nan or inf.", stacklevel=2, ) else: _assert_all_finite( sparse_container.data, allow_nan=force_all_finite == "allow-nan", estimator_name=estimator_name, input_name=input_name, ) # TODO: Remove when the minimum version of SciPy supported is 1.12 # With SciPy sparse arrays, conversion from DIA format to COO, CSR, or BSR # triggers the use of `np.int64` indices even if the data is such that it could # be more efficiently represented with `np.int32` indices. # https://github.com/scipy/scipy/issues/19245 Since not all scikit-learn # algorithms support large indices, the following code downcasts to `np.int32` # indices when it's safe to do so. if changed_format: # accept_sparse is specified to a specific format and a conversion occurred requested_sparse_format = accept_sparse[0] _preserve_dia_indices_dtype( sparse_container, sparse_container_type_name, requested_sparse_format ) return sparse_container def _ensure_no_complex_data(array): if ( hasattr(array, "dtype") and array.dtype is not None and hasattr(array.dtype, "kind") and array.dtype.kind == "c" ): raise ValueError("Complex data not supported\n{}\n".format(array)) def _check_estimator_name(estimator): if estimator is not None: if isinstance(estimator, str): return estimator else: return estimator.__class__.__name__ return None def _pandas_dtype_needs_early_conversion(pd_dtype): """Return True if pandas extension pd_dtype need to be converted early.""" # Check these early for pandas versions without extension dtypes from pandas import SparseDtype from pandas.api.types import ( is_bool_dtype, is_float_dtype, is_integer_dtype, ) if is_bool_dtype(pd_dtype): # bool and extension booleans need early conversion because __array__ # converts mixed dtype dataframes into object dtypes return True if isinstance(pd_dtype, SparseDtype): # Sparse arrays will be converted later in `check_array` return False try: from pandas.api.types import is_extension_array_dtype except ImportError: return False if isinstance(pd_dtype, SparseDtype) or not is_extension_array_dtype(pd_dtype): # Sparse arrays will be converted later in `check_array` # Only handle extension arrays for integer and floats return False elif is_float_dtype(pd_dtype): # Float ndarrays can normally support nans. They need to be converted # first to map pd.NA to np.nan return True elif is_integer_dtype(pd_dtype): # XXX: Warn when converting from a high integer to a float return True return False def _is_extension_array_dtype(array): # Pandas extension arrays have a dtype with an na_value return hasattr(array, "dtype") and hasattr(array.dtype, "na_value") def check_array( array, accept_sparse=False, *, accept_large_sparse=True, dtype="numeric", order=None, copy=False, force_all_finite=True, ensure_2d=True, allow_nd=False, ensure_min_samples=1, ensure_min_features=1, estimator=None, input_name="", ): """Input validation on an array, list, sparse matrix or similar. By default, the input is checked to be a non-empty 2D array containing only finite values. If the dtype of the array is object, attempt converting to float, raising on failure. Parameters ---------- array : object Input object to check / convert. accept_sparse : str, bool or list/tuple of str, default=False String[s] representing allowed sparse matrix formats, such as 'csc', 'csr', etc. If the input is sparse but not in the allowed format, it will be converted to the first listed format. True allows the input to be any format. False means that a sparse matrix input will raise an error. accept_large_sparse : bool, default=True If a CSR, CSC, COO or BSR sparse matrix is supplied and accepted by accept_sparse, accept_large_sparse=False will cause it to be accepted only if its indices are stored with a 32-bit dtype. .. versionadded:: 0.20 dtype : 'numeric', type, list of type or None, default='numeric' Data type of result. If None, the dtype of the input is preserved. If "numeric", dtype is preserved unless array.dtype is object. If dtype is a list of types, conversion on the first type is only performed if the dtype of the input is not in the list. order : {'F', 'C'} or None, default=None Whether an array will be forced to be fortran or c-style. When order is None (default), then if copy=False, nothing is ensured about the memory layout of the output array; otherwise (copy=True) the memory layout of the returned array is kept as close as possible to the original array. copy : bool, default=False Whether a forced copy will be triggered. If copy=False, a copy might be triggered by a conversion. force_all_finite : bool or 'allow-nan', default=True Whether to raise an error on np.inf, np.nan, pd.NA in array. The possibilities are: - True: Force all values of array to be finite. - False: accepts np.inf, np.nan, pd.NA in array. - 'allow-nan': accepts only np.nan and pd.NA values in array. Values cannot be infinite. .. versionadded:: 0.20 ``force_all_finite`` accepts the string ``'allow-nan'``. .. versionchanged:: 0.23 Accepts `pd.NA` and converts it into `np.nan` ensure_2d : bool, default=True Whether to raise a value error if array is not 2D. allow_nd : bool, default=False Whether to allow array.ndim > 2. ensure_min_samples : int, default=1 Make sure that the array has a minimum number of samples in its first axis (rows for a 2D array). Setting to 0 disables this check. ensure_min_features : int, default=1 Make sure that the 2D array has some minimum number of features (columns). The default value of 1 rejects empty datasets. This check is only enforced when the input data has effectively 2 dimensions or is originally 1D and ``ensure_2d`` is True. Setting to 0 disables this check. estimator : str or estimator instance, default=None If passed, include the name of the estimator in warning messages. input_name : str, default="" The data name used to construct the error message. In particular if `input_name` is "X" and the data has NaN values and allow_nan is False, the error message will link to the imputer documentation. .. versionadded:: 1.1.0 Returns ------- array_converted : object The converted and validated array. Examples -------- >>> from sklearn.utils.validation import check_array >>> X = [[1, 2, 3], [4, 5, 6]] >>> X_checked = check_array(X) >>> X_checked array([[1, 2, 3], [4, 5, 6]]) """ if isinstance(array, np.matrix): raise TypeError( "np.matrix is not supported. Please convert to a numpy array with " "np.asarray. For more information see: " "https://numpy.org/doc/stable/reference/generated/numpy.matrix.html" ) xp, is_array_api_compliant = get_namespace(array) # store reference to original array to check if copy is needed when # function returns array_orig = array # store whether originally we wanted numeric dtype dtype_numeric = isinstance(dtype, str) and dtype == "numeric" dtype_orig = getattr(array, "dtype", None) if not is_array_api_compliant and not hasattr(dtype_orig, "kind"): # not a data type (e.g. a column named dtype in a pandas DataFrame) dtype_orig = None # check if the object contains several dtypes (typically a pandas # DataFrame), and store them. If not, store None. dtypes_orig = None pandas_requires_conversion = False # track if we have a Series-like object to raise a better error message type_if_series = None if hasattr(array, "dtypes") and hasattr(array.dtypes, "__array__"): # throw warning if columns are sparse. If all columns are sparse, then # array.sparse exists and sparsity will be preserved (later). with suppress(ImportError): from pandas import SparseDtype def is_sparse(dtype): return isinstance(dtype, SparseDtype) if not hasattr(array, "sparse") and array.dtypes.apply(is_sparse).any(): warnings.warn( "pandas.DataFrame with sparse columns found." "It will be converted to a dense numpy array." ) dtypes_orig = list(array.dtypes) pandas_requires_conversion = any( _pandas_dtype_needs_early_conversion(i) for i in dtypes_orig ) if all(isinstance(dtype_iter, np.dtype) for dtype_iter in dtypes_orig): dtype_orig = np.result_type(*dtypes_orig) elif pandas_requires_conversion and any(d == object for d in dtypes_orig): # Force object if any of the dtypes is an object dtype_orig = object elif (_is_extension_array_dtype(array) or hasattr(array, "iloc")) and hasattr( array, "dtype" ): # array is a pandas series type_if_series = type(array) pandas_requires_conversion = _pandas_dtype_needs_early_conversion(array.dtype) if isinstance(array.dtype, np.dtype): dtype_orig = array.dtype else: # Set to None to let array.astype work out the best dtype dtype_orig = None if dtype_numeric: if ( dtype_orig is not None and hasattr(dtype_orig, "kind") and dtype_orig.kind == "O" ): # if input is object, convert to float. dtype = xp.float64 else: dtype = None if isinstance(dtype, (list, tuple)): if dtype_orig is not None and dtype_orig in dtype: # no dtype conversion required dtype = None else: # dtype conversion required. Let's select the first element of the # list of accepted types. dtype = dtype[0] if pandas_requires_conversion: # pandas dataframe requires conversion earlier to handle extension dtypes with # nans # Use the original dtype for conversion if dtype is None new_dtype = dtype_orig if dtype is None else dtype array = array.astype(new_dtype) # Since we converted here, we do not need to convert again later dtype = None if force_all_finite not in (True, False, "allow-nan"): raise ValueError( 'force_all_finite should be a bool or "allow-nan". Got {!r} instead'.format( force_all_finite ) ) if dtype is not None and _is_numpy_namespace(xp): # convert to dtype object to conform to Array API to be use `xp.isdtype` later dtype = np.dtype(dtype) estimator_name = _check_estimator_name(estimator) context = " by %s" % estimator_name if estimator is not None else "" # When all dataframe columns are sparse, convert to a sparse array if hasattr(array, "sparse") and array.ndim > 1: with suppress(ImportError): from pandas import SparseDtype # noqa: F811 def is_sparse(dtype): return isinstance(dtype, SparseDtype) if array.dtypes.apply(is_sparse).all(): # DataFrame.sparse only supports `to_coo` array = array.sparse.to_coo() if array.dtype == np.dtype("object"): unique_dtypes = set([dt.subtype.name for dt in array_orig.dtypes]) if len(unique_dtypes) > 1: raise ValueError( "Pandas DataFrame with mixed sparse extension arrays " "generated a sparse matrix with object dtype which " "can not be converted to a scipy sparse matrix." "Sparse extension arrays should all have the same " "numeric type." ) if sp.issparse(array): _ensure_no_complex_data(array) array = _ensure_sparse_format( array, accept_sparse=accept_sparse, dtype=dtype, copy=copy, force_all_finite=force_all_finite, accept_large_sparse=accept_large_sparse, estimator_name=estimator_name, input_name=input_name, ) if ensure_2d and array.ndim < 2: raise ValueError( f"Expected 2D input, got input with shape {array.shape}.\n" "Reshape your data either using array.reshape(-1, 1) if " "your data has a single feature or array.reshape(1, -1) " "if it contains a single sample." ) else: # If np.array(..) gives ComplexWarning, then we convert the warning # to an error. This is needed because specifying a non complex # dtype to the function converts complex to real dtype, # thereby passing the test made in the lines following the scope # of warnings context manager. with warnings.catch_warnings(): try: warnings.simplefilter("error", ComplexWarning) if dtype is not None and xp.isdtype(dtype, "integral"): # Conversion float -> int should not contain NaN or # inf (numpy#14412). We cannot use casting='safe' because # then conversion float -> int would be disallowed. array = _asarray_with_order(array, order=order, xp=xp) if xp.isdtype(array.dtype, ("real floating", "complex floating")): _assert_all_finite( array, allow_nan=False, msg_dtype=dtype, estimator_name=estimator_name, input_name=input_name, ) array = xp.astype(array, dtype, copy=False) else: array = _asarray_with_order(array, order=order, dtype=dtype, xp=xp) except ComplexWarning as complex_warning: raise ValueError( "Complex data not supported\n{}\n".format(array) ) from complex_warning # It is possible that the np.array(..) gave no warning. This happens # when no dtype conversion happened, for example dtype = None. The # result is that np.array(..) produces an array of complex dtype # and we need to catch and raise exception for such cases. _ensure_no_complex_data(array) if ensure_2d: # If input is scalar raise error if array.ndim == 0: raise ValueError( "Expected 2D array, got scalar array instead:\narray={}.\n" "Reshape your data either using array.reshape(-1, 1) if " "your data has a single feature or array.reshape(1, -1) " "if it contains a single sample.".format(array) ) # If input is 1D raise error if array.ndim == 1: # If input is a Series-like object (eg. pandas Series or polars Series) if type_if_series is not None: msg = ( f"Expected a 2-dimensional container but got {type_if_series} " "instead. Pass a DataFrame containing a single row (i.e. " "single sample) or a single column (i.e. single feature) " "instead." ) else: msg = ( f"Expected 2D array, got 1D array instead:\narray={array}.\n" "Reshape your data either using array.reshape(-1, 1) if " "your data has a single feature or array.reshape(1, -1) " "if it contains a single sample." ) raise ValueError(msg) if dtype_numeric and hasattr(array.dtype, "kind") and array.dtype.kind in "USV": raise ValueError( "dtype='numeric' is not compatible with arrays of bytes/strings." "Convert your data to numeric values explicitly instead." ) if not allow_nd and array.ndim >= 3: raise ValueError( "Found array with dim %d. %s expected <= 2." % (array.ndim, estimator_name) ) if force_all_finite: _assert_all_finite( array, input_name=input_name, estimator_name=estimator_name, allow_nan=force_all_finite == "allow-nan", ) if copy: if _is_numpy_namespace(xp): # only make a copy if `array` and `array_orig` may share memory` if np.may_share_memory(array, array_orig): array = _asarray_with_order( array, dtype=dtype, order=order, copy=True, xp=xp ) else: # always make a copy for non-numpy arrays array = _asarray_with_order( array, dtype=dtype, order=order, copy=True, xp=xp ) if ensure_min_samples > 0: n_samples = _num_samples(array) if n_samples < ensure_min_samples: raise ValueError( "Found array with %d sample(s) (shape=%s) while a" " minimum of %d is required%s." % (n_samples, array.shape, ensure_min_samples, context) ) if ensure_min_features > 0 and array.ndim == 2: n_features = array.shape[1] if n_features < ensure_min_features: raise ValueError( "Found array with %d feature(s) (shape=%s) while" " a minimum of %d is required%s." % (n_features, array.shape, ensure_min_features, context) ) # With an input pandas dataframe or series, we know we can always make the # resulting array writeable: # - if copy=True, we have already made a copy so it is fine to make the # array writeable # - if copy=False, the caller is telling us explicitly that we can do # in-place modifications # See https://pandas.pydata.org/docs/dev/user_guide/copy_on_write.html#read-only-numpy-arrays # for more details about pandas copy-on-write mechanism, that is enabled by # default in pandas 3.0.0.dev. if _is_pandas_df_or_series(array_orig) and hasattr(array, "flags"): array.flags.writeable = True return array def _check_large_sparse(X, accept_large_sparse=False): """Raise a ValueError if X has 64bit indices and accept_large_sparse=False""" if not accept_large_sparse: supported_indices = ["int32"] if X.format == "coo": index_keys = ["col", "row"] elif X.format in ["csr", "csc", "bsr"]: index_keys = ["indices", "indptr"] else: return for key in index_keys: indices_datatype = getattr(X, key).dtype if indices_datatype not in supported_indices: raise ValueError( "Only sparse matrices with 32-bit integer indices are accepted." f" Got {indices_datatype} indices. Please do report a minimal" " reproducer on scikit-learn issue tracker so that support for" " your use-case can be studied by maintainers. See:" " https://scikit-learn.org/dev/developers/minimal_reproducer.html" ) def check_X_y( X, y, accept_sparse=False, *, accept_large_sparse=True, dtype="numeric", order=None, copy=False, force_all_finite=True, ensure_2d=True, allow_nd=False, multi_output=False, ensure_min_samples=1, ensure_min_features=1, y_numeric=False, estimator=None, ): """Input validation for standard estimators. Checks X and y for consistent length, enforces X to be 2D and y 1D. By default, X is checked to be non-empty and containing only finite values. Standard input checks are also applied to y, such as checking that y does not have np.nan or np.inf targets. For multi-label y, set multi_output=True to allow 2D and sparse y. If the dtype of X is object, attempt converting to float, raising on failure. Parameters ---------- X : {ndarray, list, sparse matrix} Input data. y : {ndarray, list, sparse matrix} Labels. accept_sparse : str, bool or list of str, default=False String[s] representing allowed sparse matrix formats, such as 'csc', 'csr', etc. If the input is sparse but not in the allowed format, it will be converted to the first listed format. True allows the input to be any format. False means that a sparse matrix input will raise an error. accept_large_sparse : bool, default=True If a CSR, CSC, COO or BSR sparse matrix is supplied and accepted by accept_sparse, accept_large_sparse will cause it to be accepted only if its indices are stored with a 32-bit dtype. .. versionadded:: 0.20 dtype : 'numeric', type, list of type or None, default='numeric' Data type of result. If None, the dtype of the input is preserved. If "numeric", dtype is preserved unless array.dtype is object. If dtype is a list of types, conversion on the first type is only performed if the dtype of the input is not in the list. order : {'F', 'C'}, default=None Whether an array will be forced to be fortran or c-style. If `None`, then the input data's order is preserved when possible. copy : bool, default=False Whether a forced copy will be triggered. If copy=False, a copy might be triggered by a conversion. force_all_finite : bool or 'allow-nan', default=True Whether to raise an error on np.inf, np.nan, pd.NA in X. This parameter does not influence whether y can have np.inf, np.nan, pd.NA values. The possibilities are: - True: Force all values of X to be finite. - False: accepts np.inf, np.nan, pd.NA in X. - 'allow-nan': accepts only np.nan or pd.NA values in X. Values cannot be infinite. .. versionadded:: 0.20 ``force_all_finite`` accepts the string ``'allow-nan'``. .. versionchanged:: 0.23 Accepts `pd.NA` and converts it into `np.nan` ensure_2d : bool, default=True Whether to raise a value error if X is not 2D. allow_nd : bool, default=False Whether to allow X.ndim > 2. multi_output : bool, default=False Whether to allow 2D y (array or sparse matrix). If false, y will be validated as a vector. y cannot have np.nan or np.inf values if multi_output=True. ensure_min_samples : int, default=1 Make sure that X has a minimum number of samples in its first axis (rows for a 2D array). ensure_min_features : int, default=1 Make sure that the 2D array has some minimum number of features (columns). The default value of 1 rejects empty datasets. This check is only enforced when X has effectively 2 dimensions or is originally 1D and ``ensure_2d`` is True. Setting to 0 disables this check. y_numeric : bool, default=False Whether to ensure that y has a numeric type. If dtype of y is object, it is converted to float64. Should only be used for regression algorithms. estimator : str or estimator instance, default=None If passed, include the name of the estimator in warning messages. Returns ------- X_converted : object The converted and validated X. y_converted : object The converted and validated y. Examples -------- >>> from sklearn.utils.validation import check_X_y >>> X = [[1, 2], [3, 4], [5, 6]] >>> y = [1, 2, 3] >>> X, y = check_X_y(X, y) >>> X array([[1, 2], [3, 4], [5, 6]]) >>> y array([1, 2, 3]) """ if y is None: if estimator is None: estimator_name = "estimator" else: estimator_name = _check_estimator_name(estimator) raise ValueError( f"{estimator_name} requires y to be passed, but the target y is None" ) X = check_array( X, accept_sparse=accept_sparse, accept_large_sparse=accept_large_sparse, dtype=dtype, order=order, copy=copy, force_all_finite=force_all_finite, ensure_2d=ensure_2d, allow_nd=allow_nd, ensure_min_samples=ensure_min_samples, ensure_min_features=ensure_min_features, estimator=estimator, input_name="X", ) y = _check_y(y, multi_output=multi_output, y_numeric=y_numeric, estimator=estimator) check_consistent_length(X, y) return X, y def _check_y(y, multi_output=False, y_numeric=False, estimator=None): """Isolated part of check_X_y dedicated to y validation""" if multi_output: y = check_array( y, accept_sparse="csr", force_all_finite=True, ensure_2d=False, dtype=None, input_name="y", estimator=estimator, ) else: estimator_name = _check_estimator_name(estimator) y = column_or_1d(y, warn=True) _assert_all_finite(y, input_name="y", estimator_name=estimator_name) _ensure_no_complex_data(y) if y_numeric and hasattr(y.dtype, "kind") and y.dtype.kind == "O": y = y.astype(np.float64) return y def column_or_1d(y, *, dtype=None, warn=False): """Ravel column or 1d numpy array, else raises an error. Parameters ---------- y : array-like Input data. dtype : data-type, default=None Data type for `y`. .. versionadded:: 1.2 warn : bool, default=False To control display of warnings. Returns ------- y : ndarray Output data. Raises ------ ValueError If `y` is not a 1D array or a 2D array with a single row or column. Examples -------- >>> from sklearn.utils.validation import column_or_1d >>> column_or_1d([1, 1]) array([1, 1]) """ xp, _ = get_namespace(y) y = check_array( y, ensure_2d=False, dtype=dtype, input_name="y", force_all_finite=False, ensure_min_samples=0, ) shape = y.shape if len(shape) == 1: return _asarray_with_order(xp.reshape(y, (-1,)), order="C", xp=xp) if len(shape) == 2 and shape[1] == 1: if warn: warnings.warn( ( "A column-vector y was passed when a 1d array was" " expected. Please change the shape of y to " "(n_samples, ), for example using ravel()." ), DataConversionWarning, stacklevel=2, ) return _asarray_with_order(xp.reshape(y, (-1,)), order="C", xp=xp) raise ValueError( "y should be a 1d array, got an array of shape {} instead.".format(shape) ) def check_random_state(seed): """Turn seed into a np.random.RandomState instance. Parameters ---------- seed : None, int or instance of RandomState If seed is None, return the RandomState singleton used by np.random. If seed is an int, return a new RandomState instance seeded with seed. If seed is already a RandomState instance, return it. Otherwise raise ValueError. Returns ------- :class:`numpy:numpy.random.RandomState` The random state object based on `seed` parameter. Examples -------- >>> from sklearn.utils.validation import check_random_state >>> check_random_state(42) RandomState(MT19937) at 0x... """ if seed is None or seed is np.random: return np.random.mtrand._rand if isinstance(seed, numbers.Integral): return np.random.RandomState(seed) if isinstance(seed, np.random.RandomState): return seed raise ValueError( "%r cannot be used to seed a numpy.random.RandomState instance" % seed ) def has_fit_parameter(estimator, parameter): """Check whether the estimator's fit method supports the given parameter. Parameters ---------- estimator : object An estimator to inspect. parameter : str The searched parameter. Returns ------- is_parameter : bool Whether the parameter was found to be a named parameter of the estimator's fit method. Examples -------- >>> from sklearn.svm import SVC >>> from sklearn.utils.validation import has_fit_parameter >>> has_fit_parameter(SVC(), "sample_weight") True """ return parameter in signature(estimator.fit).parameters def check_symmetric(array, *, tol=1e-10, raise_warning=True, raise_exception=False): """Make sure that array is 2D, square and symmetric. If the array is not symmetric, then a symmetrized version is returned. Optionally, a warning or exception is raised if the matrix is not symmetric. Parameters ---------- array : {ndarray, sparse matrix} Input object to check / convert. Must be two-dimensional and square, otherwise a ValueError will be raised. tol : float, default=1e-10 Absolute tolerance for equivalence of arrays. Default = 1E-10. raise_warning : bool, default=True If True then raise a warning if conversion is required. raise_exception : bool, default=False If True then raise an exception if array is not symmetric. Returns ------- array_sym : {ndarray, sparse matrix} Symmetrized version of the input array, i.e. the average of array and array.transpose(). If sparse, then duplicate entries are first summed and zeros are eliminated. Examples -------- >>> import numpy as np >>> from sklearn.utils.validation import check_symmetric >>> symmetric_array = np.array([[0, 1, 2], [1, 0, 1], [2, 1, 0]]) >>> check_symmetric(symmetric_array) array([[0, 1, 2], [1, 0, 1], [2, 1, 0]]) >>> from scipy.sparse import csr_matrix >>> sparse_symmetric_array = csr_matrix(symmetric_array) >>> check_symmetric(sparse_symmetric_array) <3x3 sparse matrix of type '' with 6 stored elements in Compressed Sparse Row format> """ if (array.ndim != 2) or (array.shape[0] != array.shape[1]): raise ValueError( "array must be 2-dimensional and square. shape = {0}".format(array.shape) ) if sp.issparse(array): diff = array - array.T # only csr, csc, and coo have `data` attribute if diff.format not in ["csr", "csc", "coo"]: diff = diff.tocsr() symmetric = np.all(abs(diff.data) < tol) else: symmetric = np.allclose(array, array.T, atol=tol) if not symmetric: if raise_exception: raise ValueError("Array must be symmetric") if raise_warning: warnings.warn( ( "Array is not symmetric, and will be converted " "to symmetric by average with its transpose." ), stacklevel=2, ) if sp.issparse(array): conversion = "to" + array.format array = getattr(0.5 * (array + array.T), conversion)() else: array = 0.5 * (array + array.T) return array def _is_fitted(estimator, attributes=None, all_or_any=all): """Determine if an estimator is fitted Parameters ---------- estimator : estimator instance Estimator instance for which the check is performed. attributes : str, list or tuple of str, default=None Attribute name(s) given as string or a list/tuple of strings Eg.: ``["coef_", "estimator_", ...], "coef_"`` If `None`, `estimator` is considered fitted if there exist an attribute that ends with a underscore and does not start with double underscore. all_or_any : callable, {all, any}, default=all Specify whether all or any of the given attributes must exist. Returns ------- fitted : bool Whether the estimator is fitted. """ if attributes is not None: if not isinstance(attributes, (list, tuple)): attributes = [attributes] return all_or_any([hasattr(estimator, attr) for attr in attributes]) if hasattr(estimator, "__sklearn_is_fitted__"): return estimator.__sklearn_is_fitted__() fitted_attrs = [ v for v in vars(estimator) if v.endswith("_") and not v.startswith("__") ] return len(fitted_attrs) > 0 def check_is_fitted(estimator, attributes=None, *, msg=None, all_or_any=all): """Perform is_fitted validation for estimator. Checks if the estimator is fitted by verifying the presence of fitted attributes (ending with a trailing underscore) and otherwise raises a NotFittedError with the given message. If an estimator does not set any attributes with a trailing underscore, it can define a ``__sklearn_is_fitted__`` method returning a boolean to specify if the estimator is fitted or not. See :ref:`sphx_glr_auto_examples_developing_estimators_sklearn_is_fitted.py` for an example on how to use the API. Parameters ---------- estimator : estimator instance Estimator instance for which the check is performed. attributes : str, list or tuple of str, default=None Attribute name(s) given as string or a list/tuple of strings Eg.: ``["coef_", "estimator_", ...], "coef_"`` If `None`, `estimator` is considered fitted if there exist an attribute that ends with a underscore and does not start with double underscore. msg : str, default=None The default error message is, "This %(name)s instance is not fitted yet. Call 'fit' with appropriate arguments before using this estimator." For custom messages if "%(name)s" is present in the message string, it is substituted for the estimator name. Eg. : "Estimator, %(name)s, must be fitted before sparsifying". all_or_any : callable, {all, any}, default=all Specify whether all or any of the given attributes must exist. Raises ------ TypeError If the estimator is a class or not an estimator instance NotFittedError If the attributes are not found. Examples -------- >>> from sklearn.linear_model import LogisticRegression >>> from sklearn.utils.validation import check_is_fitted >>> from sklearn.exceptions import NotFittedError >>> lr = LogisticRegression() >>> try: ... check_is_fitted(lr) ... except NotFittedError as exc: ... print(f"Model is not fitted yet.") Model is not fitted yet. >>> lr.fit([[1, 2], [1, 3]], [1, 0]) LogisticRegression() >>> check_is_fitted(lr) """ if isclass(estimator): raise TypeError("{} is a class, not an instance.".format(estimator)) if msg is None: msg = ( "This %(name)s instance is not fitted yet. Call 'fit' with " "appropriate arguments before using this estimator." ) if not hasattr(estimator, "fit"): raise TypeError("%s is not an estimator instance." % (estimator)) if not _is_fitted(estimator, attributes, all_or_any): raise NotFittedError(msg % {"name": type(estimator).__name__}) def check_non_negative(X, whom): """ Check if there is any negative value in an array. Parameters ---------- X : {array-like, sparse matrix} Input data. whom : str Who passed X to this function. """ xp, _ = get_namespace(X) # avoid X.min() on sparse matrix since it also sorts the indices if sp.issparse(X): if X.format in ["lil", "dok"]: X = X.tocsr() if X.data.size == 0: X_min = 0 else: X_min = X.data.min() else: X_min = xp.min(X) if X_min < 0: raise ValueError("Negative values in data passed to %s" % whom) def check_scalar( x, name, target_type, *, min_val=None, max_val=None, include_boundaries="both", ): """Validate scalar parameters type and value. Parameters ---------- x : object The scalar parameter to validate. name : str The name of the parameter to be printed in error messages. target_type : type or tuple Acceptable data types for the parameter. min_val : float or int, default=None The minimum valid value the parameter can take. If None (default) it is implied that the parameter does not have a lower bound. max_val : float or int, default=None The maximum valid value the parameter can take. If None (default) it is implied that the parameter does not have an upper bound. include_boundaries : {"left", "right", "both", "neither"}, default="both" Whether the interval defined by `min_val` and `max_val` should include the boundaries. Possible choices are: - `"left"`: only `min_val` is included in the valid interval. It is equivalent to the interval `[ min_val, max_val )`. - `"right"`: only `max_val` is included in the valid interval. It is equivalent to the interval `( min_val, max_val ]`. - `"both"`: `min_val` and `max_val` are included in the valid interval. It is equivalent to the interval `[ min_val, max_val ]`. - `"neither"`: neither `min_val` nor `max_val` are included in the valid interval. It is equivalent to the interval `( min_val, max_val )`. Returns ------- x : numbers.Number The validated number. Raises ------ TypeError If the parameter's type does not match the desired type. ValueError If the parameter's value violates the given bounds. If `min_val`, `max_val` and `include_boundaries` are inconsistent. Examples -------- >>> from sklearn.utils.validation import check_scalar >>> check_scalar(10, "x", int, min_val=1, max_val=20) 10 """ def type_name(t): """Convert type into humman readable string.""" module = t.__module__ qualname = t.__qualname__ if module == "builtins": return qualname elif t == numbers.Real: return "float" elif t == numbers.Integral: return "int" return f"{module}.{qualname}" if not isinstance(x, target_type): if isinstance(target_type, tuple): types_str = ", ".join(type_name(t) for t in target_type) target_type_str = f"{{{types_str}}}" else: target_type_str = type_name(target_type) raise TypeError( f"{name} must be an instance of {target_type_str}, not" f" {type(x).__qualname__}." ) expected_include_boundaries = ("left", "right", "both", "neither") if include_boundaries not in expected_include_boundaries: raise ValueError( f"Unknown value for `include_boundaries`: {repr(include_boundaries)}. " f"Possible values are: {expected_include_boundaries}." ) if max_val is None and include_boundaries == "right": raise ValueError( "`include_boundaries`='right' without specifying explicitly `max_val` " "is inconsistent." ) if min_val is None and include_boundaries == "left": raise ValueError( "`include_boundaries`='left' without specifying explicitly `min_val` " "is inconsistent." ) comparison_operator = ( operator.lt if include_boundaries in ("left", "both") else operator.le ) if min_val is not None and comparison_operator(x, min_val): raise ValueError( f"{name} == {x}, must be" f" {'>=' if include_boundaries in ('left', 'both') else '>'} {min_val}." ) comparison_operator = ( operator.gt if include_boundaries in ("right", "both") else operator.ge ) if max_val is not None and comparison_operator(x, max_val): raise ValueError( f"{name} == {x}, must be" f" {'<=' if include_boundaries in ('right', 'both') else '<'} {max_val}." ) return x def _check_psd_eigenvalues(lambdas, enable_warnings=False): """Check the eigenvalues of a positive semidefinite (PSD) matrix. Checks the provided array of PSD matrix eigenvalues for numerical or conditioning issues and returns a fixed validated version. This method should typically be used if the PSD matrix is user-provided (e.g. a Gram matrix) or computed using a user-provided dissimilarity metric (e.g. kernel function), or if the decomposition process uses approximation methods (randomized SVD, etc.). It checks for three things: - that there are no significant imaginary parts in eigenvalues (more than 1e-5 times the maximum real part). If this check fails, it raises a ``ValueError``. Otherwise all non-significant imaginary parts that may remain are set to zero. This operation is traced with a ``PositiveSpectrumWarning`` when ``enable_warnings=True``. - that eigenvalues are not all negative. If this check fails, it raises a ``ValueError`` - that there are no significant negative eigenvalues with absolute value more than 1e-10 (1e-6) and more than 1e-5 (5e-3) times the largest positive eigenvalue in double (simple) precision. If this check fails, it raises a ``ValueError``. Otherwise all negative eigenvalues that may remain are set to zero. This operation is traced with a ``PositiveSpectrumWarning`` when ``enable_warnings=True``. Finally, all the positive eigenvalues that are too small (with a value smaller than the maximum eigenvalue multiplied by 1e-12 (2e-7)) are set to zero. This operation is traced with a ``PositiveSpectrumWarning`` when ``enable_warnings=True``. Parameters ---------- lambdas : array-like of shape (n_eigenvalues,) Array of eigenvalues to check / fix. enable_warnings : bool, default=False When this is set to ``True``, a ``PositiveSpectrumWarning`` will be raised when there are imaginary parts, negative eigenvalues, or extremely small non-zero eigenvalues. Otherwise no warning will be raised. In both cases, imaginary parts, negative eigenvalues, and extremely small non-zero eigenvalues will be set to zero. Returns ------- lambdas_fixed : ndarray of shape (n_eigenvalues,) A fixed validated copy of the array of eigenvalues. Examples -------- >>> from sklearn.utils.validation import _check_psd_eigenvalues >>> _check_psd_eigenvalues([1, 2]) # nominal case array([1, 2]) >>> _check_psd_eigenvalues([5, 5j]) # significant imag part Traceback (most recent call last): ... ValueError: There are significant imaginary parts in eigenvalues (1 of the maximum real part). Either the matrix is not PSD, or there was an issue while computing the eigendecomposition of the matrix. >>> _check_psd_eigenvalues([5, 5e-5j]) # insignificant imag part array([5., 0.]) >>> _check_psd_eigenvalues([-5, -1]) # all negative Traceback (most recent call last): ... ValueError: All eigenvalues are negative (maximum is -1). Either the matrix is not PSD, or there was an issue while computing the eigendecomposition of the matrix. >>> _check_psd_eigenvalues([5, -1]) # significant negative Traceback (most recent call last): ... ValueError: There are significant negative eigenvalues (0.2 of the maximum positive). Either the matrix is not PSD, or there was an issue while computing the eigendecomposition of the matrix. >>> _check_psd_eigenvalues([5, -5e-5]) # insignificant negative array([5., 0.]) >>> _check_psd_eigenvalues([5, 4e-12]) # bad conditioning (too small) array([5., 0.]) """ lambdas = np.array(lambdas) is_double_precision = lambdas.dtype == np.float64 # note: the minimum value available is # - single-precision: np.finfo('float32').eps = 1.2e-07 # - double-precision: np.finfo('float64').eps = 2.2e-16 # the various thresholds used for validation # we may wish to change the value according to precision. significant_imag_ratio = 1e-5 significant_neg_ratio = 1e-5 if is_double_precision else 5e-3 significant_neg_value = 1e-10 if is_double_precision else 1e-6 small_pos_ratio = 1e-12 if is_double_precision else 2e-7 # Check that there are no significant imaginary parts if not np.isreal(lambdas).all(): max_imag_abs = np.abs(np.imag(lambdas)).max() max_real_abs = np.abs(np.real(lambdas)).max() if max_imag_abs > significant_imag_ratio * max_real_abs: raise ValueError( "There are significant imaginary parts in eigenvalues (%g " "of the maximum real part). Either the matrix is not PSD, or " "there was an issue while computing the eigendecomposition " "of the matrix." % (max_imag_abs / max_real_abs) ) # warn about imaginary parts being removed if enable_warnings: warnings.warn( "There are imaginary parts in eigenvalues (%g " "of the maximum real part). Either the matrix is not" " PSD, or there was an issue while computing the " "eigendecomposition of the matrix. Only the real " "parts will be kept." % (max_imag_abs / max_real_abs), PositiveSpectrumWarning, ) # Remove all imaginary parts (even if zero) lambdas = np.real(lambdas) # Check that there are no significant negative eigenvalues max_eig = lambdas.max() if max_eig < 0: raise ValueError( "All eigenvalues are negative (maximum is %g). " "Either the matrix is not PSD, or there was an " "issue while computing the eigendecomposition of " "the matrix." % max_eig ) else: min_eig = lambdas.min() if ( min_eig < -significant_neg_ratio * max_eig and min_eig < -significant_neg_value ): raise ValueError( "There are significant negative eigenvalues (%g" " of the maximum positive). Either the matrix is " "not PSD, or there was an issue while computing " "the eigendecomposition of the matrix." % (-min_eig / max_eig) ) elif min_eig < 0: # Remove all negative values and warn about it if enable_warnings: warnings.warn( "There are negative eigenvalues (%g of the " "maximum positive). Either the matrix is not " "PSD, or there was an issue while computing the" " eigendecomposition of the matrix. Negative " "eigenvalues will be replaced with 0." % (-min_eig / max_eig), PositiveSpectrumWarning, ) lambdas[lambdas < 0] = 0 # Check for conditioning (small positive non-zeros) too_small_lambdas = (0 < lambdas) & (lambdas < small_pos_ratio * max_eig) if too_small_lambdas.any(): if enable_warnings: warnings.warn( "Badly conditioned PSD matrix spectrum: the largest " "eigenvalue is more than %g times the smallest. " "Small eigenvalues will be replaced with 0." "" % (1 / small_pos_ratio), PositiveSpectrumWarning, ) lambdas[too_small_lambdas] = 0 return lambdas def _check_sample_weight( sample_weight, X, dtype=None, copy=False, only_non_negative=False ): """Validate sample weights. Note that passing sample_weight=None will output an array of ones. Therefore, in some cases, you may want to protect the call with: if sample_weight is not None: sample_weight = _check_sample_weight(...) Parameters ---------- sample_weight : {ndarray, Number or None}, shape (n_samples,) Input sample weights. X : {ndarray, list, sparse matrix} Input data. only_non_negative : bool, default=False, Whether or not the weights are expected to be non-negative. .. versionadded:: 1.0 dtype : dtype, default=None dtype of the validated `sample_weight`. If None, and the input `sample_weight` is an array, the dtype of the input is preserved; otherwise an array with the default numpy dtype is be allocated. If `dtype` is not one of `float32`, `float64`, `None`, the output will be of dtype `float64`. copy : bool, default=False If True, a copy of sample_weight will be created. Returns ------- sample_weight : ndarray of shape (n_samples,) Validated sample weight. It is guaranteed to be "C" contiguous. """ n_samples = _num_samples(X) if dtype is not None and dtype not in [np.float32, np.float64]: dtype = np.float64 if sample_weight is None: sample_weight = np.ones(n_samples, dtype=dtype) elif isinstance(sample_weight, numbers.Number): sample_weight = np.full(n_samples, sample_weight, dtype=dtype) else: if dtype is None: dtype = [np.float64, np.float32] sample_weight = check_array( sample_weight, accept_sparse=False, ensure_2d=False, dtype=dtype, order="C", copy=copy, input_name="sample_weight", ) if sample_weight.ndim != 1: raise ValueError("Sample weights must be 1D array or scalar") if sample_weight.shape != (n_samples,): raise ValueError( "sample_weight.shape == {}, expected {}!".format( sample_weight.shape, (n_samples,) ) ) if only_non_negative: check_non_negative(sample_weight, "`sample_weight`") return sample_weight def _allclose_dense_sparse(x, y, rtol=1e-7, atol=1e-9): """Check allclose for sparse and dense data. Both x and y need to be either sparse or dense, they can't be mixed. Parameters ---------- x : {array-like, sparse matrix} First array to compare. y : {array-like, sparse matrix} Second array to compare. rtol : float, default=1e-7 Relative tolerance; see numpy.allclose. atol : float, default=1e-9 absolute tolerance; see numpy.allclose. Note that the default here is more tolerant than the default for numpy.testing.assert_allclose, where atol=0. """ if sp.issparse(x) and sp.issparse(y): x = x.tocsr() y = y.tocsr() x.sum_duplicates() y.sum_duplicates() return ( np.array_equal(x.indices, y.indices) and np.array_equal(x.indptr, y.indptr) and np.allclose(x.data, y.data, rtol=rtol, atol=atol) ) elif not sp.issparse(x) and not sp.issparse(y): return np.allclose(x, y, rtol=rtol, atol=atol) raise ValueError( "Can only compare two sparse matrices, not a sparse matrix and an array" ) def _check_response_method(estimator, response_method): """Check if `response_method` is available in estimator and return it. .. versionadded:: 1.3 Parameters ---------- estimator : estimator instance Classifier or regressor to check. response_method : {"predict_proba", "predict_log_proba", "decision_function", "predict"} or list of such str Specifies the response method to use get prediction from an estimator (i.e. :term:`predict_proba`, :term:`predict_log_proba`, :term:`decision_function` or :term:`predict`). Possible choices are: - if `str`, it corresponds to the name to the method to return; - if a list of `str`, it provides the method names in order of preference. The method returned corresponds to the first method in the list and which is implemented by `estimator`. Returns ------- prediction_method : callable Prediction method of estimator. Raises ------ AttributeError If `response_method` is not available in `estimator`. """ if isinstance(response_method, str): list_methods = [response_method] else: list_methods = response_method prediction_method = [getattr(estimator, method, None) for method in list_methods] prediction_method = reduce(lambda x, y: x or y, prediction_method) if prediction_method is None: raise AttributeError( f"{estimator.__class__.__name__} has none of the following attributes: " f"{', '.join(list_methods)}." ) return prediction_method def _check_method_params(X, params, indices=None): """Check and validate the parameters passed to a specific method like `fit`. Parameters ---------- X : array-like of shape (n_samples, n_features) Data array. params : dict Dictionary containing the parameters passed to the method. indices : array-like of shape (n_samples,), default=None Indices to be selected if the parameter has the same size as `X`. Returns ------- method_params_validated : dict Validated parameters. We ensure that the values support indexing. """ from . import _safe_indexing method_params_validated = {} for param_key, param_value in params.items(): if ( not _is_arraylike(param_value) and not sp.issparse(param_value) or _num_samples(param_value) != _num_samples(X) ): # Non-indexable pass-through (for now for backward-compatibility). # https://github.com/scikit-learn/scikit-learn/issues/15805 method_params_validated[param_key] = param_value else: # Any other method_params should support indexing # (e.g. for cross-validation). method_params_validated[param_key] = _make_indexable(param_value) method_params_validated[param_key] = _safe_indexing( method_params_validated[param_key], indices ) return method_params_validated def _is_pandas_df_or_series(X): """Return True if the X is a pandas dataframe or series.""" try: pd = sys.modules["pandas"] except KeyError: return False return isinstance(X, (pd.DataFrame, pd.Series)) def _is_pandas_df(X): """Return True if the X is a pandas dataframe.""" try: pd = sys.modules["pandas"] except KeyError: return False return isinstance(X, pd.DataFrame) def _is_polars_df_or_series(X): """Return True if the X is a polars dataframe or series.""" try: pl = sys.modules["polars"] except KeyError: return False return isinstance(X, (pl.DataFrame, pl.Series)) def _is_polars_df(X): """Return True if the X is a polars dataframe.""" try: pl = sys.modules["polars"] except KeyError: return False return isinstance(X, pl.DataFrame) def _get_feature_names(X): """Get feature names from X. Support for other array containers should place its implementation here. Parameters ---------- X : {ndarray, dataframe} of shape (n_samples, n_features) Array container to extract feature names. - pandas dataframe : The columns will be considered to be feature names. If the dataframe contains non-string feature names, `None` is returned. - All other array containers will return `None`. Returns ------- names: ndarray or None Feature names of `X`. Unrecognized array containers will return `None`. """ feature_names = None # extract feature names for support array containers if _is_pandas_df(X): # Make sure we can inspect columns names from pandas, even with # versions too old to expose a working implementation of # __dataframe__.column_names() and avoid introducing any # additional copy. # TODO: remove the pandas-specific branch once the minimum supported # version of pandas has a working implementation of # __dataframe__.column_names() that is guaranteed to not introduce any # additional copy of the data without having to impose allow_copy=False # that could fail with other libraries. Note: in the longer term, we # could decide to instead rely on the __dataframe_namespace__ API once # adopted by our minimally supported pandas version. feature_names = np.asarray(X.columns, dtype=object) elif hasattr(X, "__dataframe__"): df_protocol = X.__dataframe__() feature_names = np.asarray(list(df_protocol.column_names()), dtype=object) if feature_names is None or len(feature_names) == 0: return types = sorted(t.__qualname__ for t in set(type(v) for v in feature_names)) # mixed type of string and non-string is not supported if len(types) > 1 and "str" in types: raise TypeError( "Feature names are only supported if all input features have string names, " f"but your input has {types} as feature name / column name types. " "If you want feature names to be stored and validated, you must convert " "them all to strings, by using X.columns = X.columns.astype(str) for " "example. Otherwise you can remove feature / column names from your input " "data, or convert them all to a non-string data type." ) # Only feature names of all strings are supported if len(types) == 1 and types[0] == "str": return feature_names def _check_feature_names_in(estimator, input_features=None, *, generate_names=True): """Check `input_features` and generate names if needed. Commonly used in :term:`get_feature_names_out`. Parameters ---------- input_features : array-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. generate_names : bool, default=True Whether to generate names when `input_features` is `None` and `estimator.feature_names_in_` is not defined. This is useful for transformers that validates `input_features` but do not require them in :term:`get_feature_names_out` e.g. `PCA`. Returns ------- feature_names_in : ndarray of str or `None` Feature names in. """ feature_names_in_ = getattr(estimator, "feature_names_in_", None) n_features_in_ = getattr(estimator, "n_features_in_", None) if input_features is not None: input_features = np.asarray(input_features, dtype=object) if feature_names_in_ is not None and not np.array_equal( feature_names_in_, input_features ): raise ValueError("input_features is not equal to feature_names_in_") if n_features_in_ is not None and len(input_features) != n_features_in_: raise ValueError( "input_features should have length equal to number of " f"features ({n_features_in_}), got {len(input_features)}" ) return input_features if feature_names_in_ is not None: return feature_names_in_ if not generate_names: return # Generates feature names if `n_features_in_` is defined if n_features_in_ is None: raise ValueError("Unable to generate feature names without n_features_in_") return np.asarray([f"x{i}" for i in range(n_features_in_)], dtype=object) def _generate_get_feature_names_out(estimator, n_features_out, input_features=None): """Generate feature names out for estimator using the estimator name as the prefix. The input_feature names are validated but not used. This function is useful for estimators that generate their own names based on `n_features_out`, i.e. PCA. Parameters ---------- estimator : estimator instance Estimator producing output feature names. n_feature_out : int Number of feature names out. input_features : array-like of str or None, default=None Only used to validate feature names with `estimator.feature_names_in_`. Returns ------- feature_names_in : ndarray of str or `None` Feature names in. """ _check_feature_names_in(estimator, input_features, generate_names=False) estimator_name = estimator.__class__.__name__.lower() return np.asarray( [f"{estimator_name}{i}" for i in range(n_features_out)], dtype=object ) def _check_monotonic_cst(estimator, monotonic_cst=None): """Check the monotonic constraints and return the corresponding array. This helper function should be used in the `fit` method of an estimator that supports monotonic constraints and called after the estimator has introspected input data to set the `n_features_in_` and optionally the `feature_names_in_` attributes. .. versionadded:: 1.2 Parameters ---------- estimator : estimator instance monotonic_cst : array-like of int, dict of str or None, default=None Monotonic constraints for the features. - If array-like, then it should contain only -1, 0 or 1. Each value will be checked to be in [-1, 0, 1]. If a value is -1, then the corresponding feature is required to be monotonically decreasing. - If dict, then it the keys should be the feature names occurring in `estimator.feature_names_in_` and the values should be -1, 0 or 1. - If None, then an array of 0s will be allocated. Returns ------- monotonic_cst : ndarray of int Monotonic constraints for each feature. """ original_monotonic_cst = monotonic_cst if monotonic_cst is None or isinstance(monotonic_cst, dict): monotonic_cst = np.full( shape=estimator.n_features_in_, fill_value=0, dtype=np.int8, ) if isinstance(original_monotonic_cst, dict): if not hasattr(estimator, "feature_names_in_"): raise ValueError( f"{estimator.__class__.__name__} was not fitted on data " "with feature names. Pass monotonic_cst as an integer " "array instead." ) unexpected_feature_names = list( set(original_monotonic_cst) - set(estimator.feature_names_in_) ) unexpected_feature_names.sort() # deterministic error message n_unexpeced = len(unexpected_feature_names) if unexpected_feature_names: if len(unexpected_feature_names) > 5: unexpected_feature_names = unexpected_feature_names[:5] unexpected_feature_names.append("...") raise ValueError( f"monotonic_cst contains {n_unexpeced} unexpected feature " f"names: {unexpected_feature_names}." ) for feature_idx, feature_name in enumerate(estimator.feature_names_in_): if feature_name in original_monotonic_cst: cst = original_monotonic_cst[feature_name] if cst not in [-1, 0, 1]: raise ValueError( f"monotonic_cst['{feature_name}'] must be either " f"-1, 0 or 1. Got {cst!r}." ) monotonic_cst[feature_idx] = cst else: unexpected_cst = np.setdiff1d(monotonic_cst, [-1, 0, 1]) if unexpected_cst.shape[0]: raise ValueError( "monotonic_cst must be an array-like of -1, 0 or 1. Observed " f"values: {unexpected_cst.tolist()}." ) monotonic_cst = np.asarray(monotonic_cst, dtype=np.int8) if monotonic_cst.shape[0] != estimator.n_features_in_: raise ValueError( f"monotonic_cst has shape {monotonic_cst.shape} but the input data " f"X has {estimator.n_features_in_} features." ) return monotonic_cst def _check_pos_label_consistency(pos_label, y_true): """Check if `pos_label` need to be specified or not. In binary classification, we fix `pos_label=1` if the labels are in the set {-1, 1} or {0, 1}. Otherwise, we raise an error asking to specify the `pos_label` parameters. Parameters ---------- pos_label : int, float, bool, str or None The positive label. y_true : ndarray of shape (n_samples,) The target vector. Returns ------- pos_label : int, float, bool or str If `pos_label` can be inferred, it will be returned. Raises ------ ValueError In the case that `y_true` does not have label in {-1, 1} or {0, 1}, it will raise a `ValueError`. """ # ensure binary classification if pos_label is not specified # classes.dtype.kind in ('O', 'U', 'S') is required to avoid # triggering a FutureWarning by calling np.array_equal(a, b) # when elements in the two arrays are not comparable. if pos_label is None: # Compute classes only if pos_label is not specified: classes = np.unique(y_true) if classes.dtype.kind in "OUS" or not ( np.array_equal(classes, [0, 1]) or np.array_equal(classes, [-1, 1]) or np.array_equal(classes, [0]) or np.array_equal(classes, [-1]) or np.array_equal(classes, [1]) ): classes_repr = ", ".join([repr(c) for c in classes.tolist()]) raise ValueError( f"y_true takes value in {{{classes_repr}}} and pos_label is not " "specified: either make y_true take value in {0, 1} or " "{-1, 1} or pass pos_label explicitly." ) pos_label = 1 return pos_label def _to_object_array(sequence): """Convert sequence to a 1-D NumPy array of object dtype. numpy.array constructor has a similar use but it's output is ambiguous. It can be 1-D NumPy array of object dtype if the input is a ragged array, but if the input is a list of equal length arrays, then the output is a 2D numpy.array. _to_object_array solves this ambiguity by guarantying that the output is a 1-D NumPy array of objects for any input. Parameters ---------- sequence : array-like of shape (n_elements,) The sequence to be converted. Returns ------- out : ndarray of shape (n_elements,), dtype=object The converted sequence into a 1-D NumPy array of object dtype. Examples -------- >>> import numpy as np >>> from sklearn.utils.validation import _to_object_array >>> _to_object_array([np.array([0]), np.array([1])]) array([array([0]), array([1])], dtype=object) >>> _to_object_array([np.array([0]), np.array([1, 2])]) array([array([0]), array([1, 2])], dtype=object) >>> _to_object_array([np.array([0]), np.array([1, 2])]) array([array([0]), array([1, 2])], dtype=object) """ out = np.empty(len(sequence), dtype=object) out[:] = sequence return out