Implementation:Scikit learn Scikit learn DictionaryLearning

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Dimensionality Reduction, Sparse Coding
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for dictionary learning and sparse coding provided by scikit-learn.

Description

DictionaryLearning finds a dictionary (a set of atoms) that performs well at sparsely encoding the fitted data. It solves an optimization problem that minimizes the Frobenius norm of the reconstruction error subject to an L1 penalty on the codes and unit-norm constraints on the dictionary atoms. The module also provides SparseCoder for encoding data against a fixed dictionary, and MiniBatchDictionaryLearning for scalable online dictionary learning.

Usage

Use DictionaryLearning when you need to learn an overcomplete dictionary from data for sparse representation, feature extraction, or signal decomposition tasks. It is particularly useful for image denoising, texture analysis, and compressed sensing applications where sparse representations are desirable.

Code Reference

Source Location

• Repository: scikit-learn
• File: sklearn/decomposition/_dict_learning.py

Signature

class DictionaryLearning(_BaseSparseCoding, BaseEstimator):
    def __init__(
        self,
        n_components=None,
        *,
        alpha=1,
        max_iter=1000,
        tol=1e-8,
        fit_algorithm="lars",
        transform_algorithm="omp",
        transform_n_nonzero_coefs=None,
        transform_alpha=None,
        n_jobs=None,
        code_init=None,
        dict_init=None,
        callback=None,
        verbose=False,
        split_sign=False,
        random_state=None,
        positive_code=False,
        positive_dict=False,
        transform_max_iter=1000,
    ):

Import

from sklearn.decomposition import DictionaryLearning

I/O Contract

Inputs

Name	Type	Required	Description
n_components	int	No	Number of dictionary elements to extract. Defaults to n_features if None.
alpha	float	No	Sparsity controlling parameter (default=1.0).
max_iter	int	No	Maximum number of iterations to perform (default=1000).
tol	float	No	Tolerance for numerical error (default=1e-8).
fit_algorithm	str	No	Algorithm for fitting: 'lars' or 'cd' (default='lars').
transform_algorithm	str	No	Algorithm for transform: 'lasso_lars', 'lasso_cd', 'lars', 'omp', or 'threshold' (default='omp').
transform_n_nonzero_coefs	int	No	Number of nonzero coefficients to target during transform.
transform_alpha	float	No	Regularization parameter for transform. Defaults to alpha.
n_jobs	int	No	Number of parallel jobs to run.
random_state	int or RandomState	No	Random state for reproducibility.

Outputs

Name	Type	Description
components_	ndarray of shape (n_components, n_features)	The dictionary atoms extracted from the data.
error_	array	Vector of errors at each iteration.
n_features_in_	int	Number of features seen during fit.
n_iter_	int	Number of iterations run.

Usage Examples

Basic Usage

import numpy as np
from sklearn.decomposition import DictionaryLearning

X = np.array([[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12]], dtype=float)
dl = DictionaryLearning(n_components=2, transform_algorithm='lasso_lars', random_state=0)
X_transformed = dl.fit_transform(X)
print(X_transformed.shape)  # (3, 2)
print(dl.components_.shape)  # (2, 4)

Related Pages

• Principle:Scikit_learn_Scikit_learn_Dimensionality_Reduction