Implementation:Scikit learn Scikit learn OrthogonalMatchingPursuit

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Machine Learning, Sparse Approximation
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for sparse signal approximation using the Orthogonal Matching Pursuit algorithm provided by scikit-learn.

Description

OrthogonalMatchingPursuit (OMP) implements a greedy algorithm for selecting features in sparse linear regression. At each step, OMP selects the feature most correlated with the current residual, then performs an orthogonal projection to update the residual. The algorithm is parameterized by either the number of non-zero coefficients desired or a tolerance on the squared norm of the residual. The module also provides OrthogonalMatchingPursuitCV for cross-validated selection of the number of non-zero coefficients.

Usage

Use OrthogonalMatchingPursuit when you need a fast, greedy approach to sparse signal recovery, when you know approximately how many features should be active, or when you need sparse approximation of a signal from a dictionary of atoms. It is widely used in compressed sensing and signal processing applications.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/linear_model/_omp.py

Signature

class OrthogonalMatchingPursuit(MultiOutputMixin, RegressorMixin, LinearModel):
    def __init__(
        self,
        *,
        n_nonzero_coefs=None,
        tol=None,
        fit_intercept=True,
        precompute="auto",
    ):

Import

from sklearn.linear_model import OrthogonalMatchingPursuit

I/O Contract

Inputs

Name	Type	Required	Description
n_nonzero_coefs	int	No	Desired number of non-zero entries in the solution (default=None, auto-set to 10% of n_features or 1)
tol	float	No	Maximum squared norm of the residual; overrides n_nonzero_coefs if set (default=None)
fit_intercept	bool	No	Whether to calculate the intercept (default=True)
precompute	bool or 'auto'	No	Whether to use precomputed Gram and Xy matrices (default='auto')

Outputs

Name	Type	Description
coef_	ndarray of shape (n_features,) or (n_targets, n_features)	Parameter vector
intercept_	float or ndarray of shape (n_targets,)	Independent term in decision function
n_iter_	int or array-like	Number of active features across every target
n_nonzero_coefs_	int or None	The resolved number of non-zero coefficients

Usage Examples

Basic Usage

from sklearn.linear_model import OrthogonalMatchingPursuit
from sklearn.datasets import make_regression

X, y = make_regression(n_samples=100, n_features=50, n_informative=5, noise=5, random_state=42)
model = OrthogonalMatchingPursuit(n_nonzero_coefs=5)
model.fit(X, y)
print("Non-zero coefficients:", (model.coef_ != 0).sum())
print("Score:", model.score(X, y))

Related Pages

• Principle:Scikit_learn_Scikit_learn_Linear_Regression