Implementation:Scikit learn Scikit learn LatentDirichletAllocation

Knowledge Sources	Scikit_learn Scikit-learn Docs
Domains	Topic Modeling, Natural Language Processing
Last Updated	2026-02-08 15:00 GMT

Overview

Concrete tool for topic modeling using Latent Dirichlet Allocation with online variational Bayes provided by scikit-learn.

Description

LatentDirichletAllocation implements the online variational Bayes algorithm for Latent Dirichlet Allocation (LDA), a generative probabilistic model for collections of discrete data such as text corpora. It discovers abstract topics in a set of documents by learning a topic-word distribution and a document-topic distribution. The implementation supports both batch and online learning methods, with the online method being much faster for large datasets.

Usage

Use LatentDirichletAllocation when you need to discover hidden topics in a collection of documents or any discrete count data. It is commonly applied to text mining, document clustering, and content recommendation systems where understanding the thematic structure of a corpus is important.

Code Reference

Source Location

Repository: scikit-learn
File: sklearn/decomposition/_lda.py

Signature

class LatentDirichletAllocation(
    ClassNamePrefixFeaturesOutMixin, TransformerMixin, BaseEstimator
):
    def __init__(
        self,
        n_components=10,
        *,
        doc_topic_prior=None,
        topic_word_prior=None,
        learning_method="batch",
        learning_decay=0.7,
        learning_offset=10.0,
        max_iter=10,
        batch_size=128,
        evaluate_every=-1,
        total_samples=1e6,
        perp_tol=1e-1,
        mean_change_tol=1e-3,
        max_doc_update_iter=100,
        n_jobs=None,
        verbose=0,
        random_state=None,
    ):

Import

from sklearn.decomposition import LatentDirichletAllocation

I/O Contract

Inputs

Name	Type	Required	Description
n_components	int	No	Number of topics (default=10).
doc_topic_prior	float	No	Prior of document topic distribution (alpha). Defaults to 1/n_components.
topic_word_prior	float	No	Prior of topic word distribution (eta). Defaults to 1/n_components.
learning_method	str	No	Method to update components: 'batch' or 'online' (default='batch').
learning_decay	float	No	Controls learning rate in online learning (default=0.7).
learning_offset	float	No	Downweights early iterations in online learning (default=10.0).
max_iter	int	No	Maximum number of passes over the training data (default=10).
batch_size	int	No	Number of documents per EM iteration in online learning (default=128).
evaluate_every	int	No	How often to evaluate perplexity during training (default=-1).
n_jobs	int	No	Number of parallel jobs.
random_state	int or RandomState	No	Random state for reproducibility.

Outputs

Name	Type	Description
components_	ndarray of shape (n_components, n_features)	Variational parameters for topic-word distribution (unnormalized).
exp_dirichlet_component_	ndarray of shape (n_components, n_features)	Exponential of the expectation of log topic-word distribution.
n_batch_iter_	int	Number of mini-batch iterations.
n_iter_	int	Number of passes over the dataset.
bound_	float	Final perplexity score on training set.
n_features_in_	int	Number of features seen during fit.

Usage Examples

Basic Usage

import numpy as np
from sklearn.decomposition import LatentDirichletAllocation

# Simulate a document-term matrix (5 documents, 10 terms)
rng = np.random.RandomState(0)
X = rng.randint(0, 10, size=(5, 10)).astype(float)

lda = LatentDirichletAllocation(n_components=3, random_state=0)
X_topics = lda.fit_transform(X)
print(X_topics.shape)  # (5, 3)
print(lda.components_.shape)  # (3, 10)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment