Implementation:Scikit learn Scikit learn ArrayApiDelegation

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Utilities, Array API
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for delegating array operations to backend-specific implementations via the Array API standard provided by scikit-learn.

Description

This module provides delegation functions that route array operations to the appropriate backend-specific implementations based on the input array's namespace (NumPy, CuPy, Dask, JAX, PyTorch, or PyData Sparse). It implements Public API functions like isclose, nan_to_num, one_hot, and pad that work transparently across different array backends. The module detects which array library is being used and dispatches to optimized implementations accordingly.

Usage

Use this module when writing scikit-learn code that needs to work with multiple array backends (NumPy, CuPy, JAX, etc.) through the Python Array API standard, ensuring consistent behavior across backends.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/externals/array_api_extra/_delegation.py

Signature

def isclose(
    a: Array | complex,
    b: Array | complex,
    *,
    rtol: float = 1e-05,
    atol: float = 1e-08,
    equal_nan: bool = False,
    xp: ModuleType | None = None,
) -> Array

def nan_to_num(...)
def one_hot(...)
def pad(...)

Import

from sklearn.externals.array_api_extra._delegation import isclose, nan_to_num, one_hot, pad

I/O Contract

Inputs

Name	Type	Required	Description
a	Array or complex	Yes	First input array or scalar for comparison
b	Array or complex	Yes	Second input array or scalar for comparison
rtol	float	No	Relative tolerance parameter (default: 1e-05)
atol	float	No	Absolute tolerance parameter (default: 1e-08)
equal_nan	bool	No	Whether to treat NaN values as equal (default: False)
xp	ModuleType or None	No	Array namespace to use; inferred if not provided

Outputs

Name	Type	Description
result	Array	Boolean array where elements are True where inputs are close

Usage Examples

Basic Usage

import numpy as np
from sklearn.externals.array_api_extra._delegation import isclose

a = np.array([1.0, 2.0, 3.0])
b = np.array([1.0, 2.00001, 3.1])
result = isclose(a, b, atol=1e-4)
print(result)  # [ True  True False]

Related Pages

• Principle:Scikit_learn_Scikit_learn_Feature_Transformation