Affinity Propagation Clustering Method in ML

Did you know? Affinity Propagation clustering was originally developed for message-passing in brain-like networks, inspired by how neurons communicate. It has become a powerful tool for discovering natural groupings in complex data without needing to predefine the number of clusters.

Affinity Propagation Clustering is a machine learning technique that groups data points based on similarity, without needing to predefine the number of clusters. Unlike k-means, which requires selecting the number of clusters upfront, it automatically determines the optimal number using a similarity matrix.

This method is particularly effective for scenarios where the underlying structure of the data is unknown and can adapt to complex datasets.

In this tutorial, we’ll explore the fundamentals of Affinity Propagation clustering in ML, covering its principles, advantages, and implementation for clustering.

Improve your machine learning skills with our online AI and ML courses — take the next step in your learning journey! 

Understanding Affinity Propagation Clustering in ML

Affinity Propagation is a clustering algorithm that automatically identifies exemplars, the most representative points of a cluster, in a dataset. 

Traditional clustering methods such as k-means require the number of clusters to be predefined. In contrast, Affinity Propagation does not make this assumption. This flexibility makes it particularly useful for exploratory data analysis, where the number of clusters is unknown.

The method was introduced by Frey and Dueck in 2007 and has since been applied in various fields like biology, computer vision, and social network analysis due to its ability to handle complex data structures.

Machine learning professionals skilled in techniques like affinity propagation are in high demand due to their ability to handle complex data. If you're looking to develop skills in AI and ML, here are some top-rated courses to help you get there:

• Masters in Artificial Intelligence and Machine Learning
• Executive Diploma in Machine Learning and AI
• Master’s Degree in Artificial Intelligence and Data Science

Affinity Propagation uses two key matrices: responsibility and availability. These matrices are iteratively updated until the algorithm converges, and the final exemplars are selected.

Responsibility Matrix: This matrix indicates how well-suited each data point is to serve as an exemplar for another point. For example, if point A is similar to point B, the responsibility matrix will indicate that point A can potentially serve as the exemplar for point B.

Availability Matrix: This matrix shows how much each data point wants to choose another point as an exemplar. Essentially, it measures how much each point "prefers" a potential exemplar. If point A is much closer to point B than other points, point A will have a high value in the availability matrix for B

The algorithm alternates between updating these matrices:

• The responsibility matrix is updated based on how well each point serves as an exemplar for others.
• The availability matrix is updated based on how much each point wants to select others as exemplars.

This back-and-forth update process continues until the algorithm reaches a stable state (convergence), where points are assigned to clusters, and exemplars are determined.

Why It’s Unique? Unlike algorithms such as k-means, where the user has to decide the number of clusters upfront, Affinity Propagation automatically determines the optimal number of clusters based on the data's inherent structure. 

The process of identifying exemplars and assigning points to them makes Affinity Propagation well-suited for complex datasets where traditional clustering methods may struggle to identify the true structure of the data.

Also Read: What is Clustering in Machine Learning and Different Types of Clustering Methods 

Now that you know what is affinity propagation clustering, let’s understand how it works with an example.

How Does Affinity Propagation Clustering Work in ML? Step-by-Step Guide

Affinity Propagation is an iterative clustering algorithm that works by identifying exemplars (representative data points) and grouping other points based on their similarity to these exemplars. The algorithm alternates between two key processes: calculating responsibilities and updating availabilities. 

Here’s a step-by-step breakdown of how Affinity Propagation works:

1. Create the Similarity Matrix

The first step in Affinity Propagation is to calculate the similarity matrix, which reflects how similar each pair of data points is.

How it works? We compute a similarity measure (e.g., Euclidean distance or cosine similarity) between all pairs of points. The higher the similarity, the closer the points are in the matrix.

Example: Let's say the similarity between the data points is computed as follows (higher values indicate greater similarity):

    A    B    C    D    E
A   1.0  0.8  0.6  0.2  0.3
B   0.8  1.0  0.7  0.3  0.4
C   0.6  0.7  1.0  0.5  0.2
D   0.2  0.3  0.5  1.0  0.7
E   0.3  0.4  0.2  0.7  1.0

2. Initialize Responsibility and Availability Matrices

Initialize the responsibility and availability matrices. These are initially filled with random values or zeros.

How it works?

• The responsibility matrix indicates how much each data point i should take point k as an exemplar.
• The availability matrix indicates how much point i supports point k as an exemplar.

Example: Initially, both matrices are set to zero or random values:

	Responsibility Matrix (r)   Availability Matrix (a)
   A  B  C  D  E               A  B  C  D  E
A  0  0  0  0  0            A  0  0  0  0  0
B  0  0  0  0  0            B  0  0  0  0  0
C  0  0  0  0  0            C  0  0  0  0  0
D  0  0  0  0  0            D  0  0  0  0  0
E  0  0  0  0  0            E  0  0  0  0  0

3. Update Responsibility Matrix

The responsibility matrix is updated iteratively. It represents how much data point i should assign to point k as an exemplar.

How it works? Each entry in the responsibility matrix is updated using the formula.

This means the responsibility of point i towards k is calculated by subtracting the maximum availability of other points from the similarity score.

Example: After the first iteration, the responsibility matrix might be updated like this:

	Responsibility Matrix (r)
    A    B     C     D    E
A   0    0.1   0.2  -0.1  0.1
B   0    0.2   0.1  -0.2  0.05
C   0.1  0.05  0.3  -0.1  0.1
D  -0.1  0.05 -0.05  0.5  0.3
E   0.2  0.1   0.05  0.2  0.4

4. Update Availability Matrix

The availability matrix is updated based on the responsibility matrix. It shows how much each data point supports other points to become an exemplar.

How it works? The availability of a point i to choose k as an exemplar is calculated using the formula:

This means the availability of a point to select another as an exemplar is updated based on the maximum responsibility values.

Example: After the first update, the availability matrix might look like this:

	Availability Matrix (a)
   A    B   C    D     E
A  0    0.1 0.2  0.05  0.1
B  0.1  0   0.2  0.05  0.1
C  0.2  0.1 0    0.05  0.15
D  0.05 0.1 0.05 0     0.2
E  0.1  0.2 0.1  0.15  0

5. Repeat Responsibility and Availability Updates

The responsibility and availability matrices are updated alternately, iterating several times until the algorithm converges.

How it works? The updates are repeated, refining the responsibility and availability values after each iteration, until the changes between iterations become small, indicating convergence.

Example: After several iterations, the matrices stabilize, and points are assigned to clusters based on the final responsibility and availability values.

6. Select Exemplars and Assign Points to Clusters

Once the algorithm converges, the final exemplars are selected, and the data points are assigned to the closest exemplar.

How it works? The points with the highest responsibility values for themselves are chosen as exemplars, and all other points are assigned to the cluster of the most similar exemplar.

Example: Suppose, after convergence, points A and B are identified as exemplars. Points C, D, and E will then be assigned to either A or B, based on their similarity.

7. Convergence

The algorithm stops when the responsibility and availability matrices no longer change significantly, indicating that the clustering solution has stabilized.

How it works? Once the algorithm converges, each data point belongs to one of the clusters formed by the exemplars, and the final clustering solution is achieved.

Example: In this case, points A and B are final exemplars, and the dataset is successfully clustered into two groups based on their similarities to A and B.

Final Output: After the iterative process of updating the responsibility and availability matrices, Affinity Propagation clusters the data points into groups, each centered around one or more exemplars. 

These exemplars represent the most central and influential points of each cluster. Here's what the final output looks like based on the previous example:

Exemplars: A and B are identified as the final exemplars.

Cluster Assignments:

• Cluster 1: Points A and C (assigned to A based on highest similarity).
• Cluster 2: Points B, D, and E (assigned to B based on highest similarity).

These groups form the final clusters, where each cluster has a representative exemplar that best defines the group, and the other data points are assigned based on their similarity to the exemplars.

This output demonstrates the effectiveness of Affinity Propagation in automatically identifying the optimal number of clusters and their corresponding exemplars. It doesn’t require any user input for the number of clusters, making it ideal for exploratory data analysis where the structure of the data is unknown.

Also Read: Hierarchical Clustering in Python [Concepts and Analysis] 

Now that you know how exactly Affinity Propagation Clustering works, let’s look at how you can implement it using Python.

Implementing Affinity Propagation Clustering in Python

Python is an excellent choice for implementing Affinity Propagation clustering due to its simplicity, the powerful scikit-learn library, and seamless integration with tools like numpy and matplotlib for data manipulation and visualization. These features allow for efficient implementation, easy experimentation, and effective analysis, making Python ideal for both beginners and experts in machine learning.

Let’s go through how you can implement affinity clustering it in Python:

1. Install the Necessary Libraries

Before starting, make sure you have the following libraries installed:

pip install scikit-learn numpy matplotlib

• scikit-learn: The main library for machine learning algorithms, including Affinity Propagation.
• numpy: For handling numerical data and matrices.
• matplotlib: For visualizing the clusters.

2. Import the Libraries

You need to import the necessary modules in your Python script:

import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import AffinityPropagation
from sklearn.metrics import pairwise_distances

You can get a better understanding of Python libraries with upGrad’s Learn Python Libraries: NumPy, Matplotlib & Pandas. Learn how to manipulate data using NumPy, visualize insights with Matplotlib, and analyze datasets with Pandas.

Also Read: Python Modules: Explore 20+ Essential Modules and Best Practices

3. Prepare Your Dataset

For this example, let's create a simple 2D dataset.

# Example: 2D dataset (you can replace this with your own data)
X = np.array([[1, 2], [1, 4], [1, 0], [10, 2], [10, 4], [10, 0]])

If you have a dataset with more complex or higher-dimensional data, you can directly load it using pandas or other data loading methods.

4. Fit the Affinity Propagation Model

Now, we can initialize the AffinityPropagation model and fit it to our data:

# Initialize the Affinity Propagation model
affinity_propagation = AffinityPropagation()

# Fit the model to the data
affinity_propagation.fit(X)

# Get cluster centers (exemplars) and labels
cluster_centers_indices = affinity_propagation.cluster_centers_indices_
labels = affinity_propagation.labels_

5. Visualize the Clusters

To visualize the clusters, let's use matplotlib to plot the data points and the cluster centers (exemplars):

# Plot data points
plt.scatter(X[:, 0], X[:, 1], c=labels, cmap='viridis', marker='o', edgecolors='k')

# Mark the exemplars (cluster centers)
plt.scatter(X[cluster_centers_indices, 0], X[cluster_centers_indices, 1], c='red', marker='x', label='Exemplars')

# Show the plot
plt.title('Affinity Propagation Clustering')
plt.xlabel('Feature 1')
plt.ylabel('Feature 2')
plt.legend()
plt.show()

This code will display a scatter plot of your data points, with different colors representing different clusters. The red 'X' marks indicate the exemplars, which are the center of each cluster.

Here is the output of the Affinity Propagation Clustering visualization. The plot shows the data points in two different clusters, with the exemplars marked in red.

Also Read: Matplotlib in Python: Explained Various Plots with Examples 

6. Tips for Running Affinity Propagation Effectively

Choice of Similarity Measure: Affinity Propagation requires a similarity matrix. By default, scikit-learn uses the negative squared Euclidean distance as the similarity measure. If your data requires a different similarity measure, you can provide a custom one using the affinity parameter in AffinityPropagation().

# Using 'precomputed' affinity if you already have a similarity matrix
similarity_matrix = pairwise_distances(X, metric='euclidean')
affinity_propagation = AffinityPropagation(affinity='precomputed')
affinity_propagation.fit(similarity_matrix)

Preference Parameter: The preference parameter determines how likely points are to become exemplars. Setting a high value for preference may result in fewer clusters (more points chosen as exemplars), while lower values lead to more clusters. Experiment with this parameter for optimal results.

affinity_propagation = AffinityPropagation(preference=-50)

Convergence: If you’re working with large datasets, Affinity Propagation can take a long time to converge. You can speed up the process by adjusting the damping parameter (default is 0.5). A lower damping value will make the algorithm converge faster, but too low a value may reduce stability

affinity_propagation = AffinityPropagation(damping=0.9)

7. Handling Large Datasets

Affinity Propagation can be computationally expensive for large datasets, as it requires the computation of a full similarity matrix, which can be memory-intensive. If you have a very large dataset, consider:

• Using a sampling technique to reduce the number of data points before applying Affinity Propagation.
• Using dimensionality reduction methods like PCA (Principal Component Analysis) to reduce the complexity of the dataset before clustering.

This step-by-step guide provides an easy way to implement Affinity Propagation clustering in Python, leveraging scikit-learn’s built-in functions. 

Feel free to experiment with different datasets and fine-tune the algorithm to get the best results for your specific problem!

Also Read: Top 10 Dimensionality Reduction Techniques for Machine Learning(ML) in 2025

Now that you have a better understanding of how to implement affinity clustering in Python, let’s look at some of its advantages and drawbacks.

Benefits and Limitations of Affinity Propagation Clustering

Affinity Propagation excels in scenarios where the number of clusters is not known beforehand. It automatically identifies the optimal number of clusters based on the data’s inherent structure. 

This makes it particularly useful for exploratory data analysis and complex datasets where traditional methods like k-means might require trial and error. It shines in applications such as image segmentation, gene expression analysis, and customer segmentation. 

However, it falls short when dealing with large datasets, as the algorithm requires a similarity matrix that grows quadratically with the number of data points, leading to high computational costs and memory requirements. 

Here are some of its key benefits and limitations:

To make the most out of Affinity Propagation, here are some best practices:

• The preference parameter is crucial in controlling the number of clusters formed. A higher preference value tends to result in fewer clusters, while a lower value may lead to more clusters. Experiment with different values to find the optimal setting for your dataset.
• If you're working with data that has a complex structure, consider calculating a custom similarity matrix based on domain knowledge, especially if your data isn't easily represented by Euclidean distances.
• For large datasets, the similarity matrix can become too large to store in memory. Consider using dimensionality reduction techniques (e.g., PCA) to reduce the complexity of the data before applying Affinity Propagation. Alternatively, using a subset of the data can help reduce memory usage and speed up computation.
• Affinity Propagation’s convergence time can vary based on the dataset size and complexity. Monitor the convergence during training to ensure it completes in a reasonable amount of time, and consider using early stopping techniques if the process becomes too slow.
• After running Affinity Propagation, visualize the clusters to gain insights into the data. Use scatter plots or other appropriate visualization techniques to examine the quality of the clusters and ensure they align with your expectations.
• Use clustering evaluation metrics, such as Silhouette Score or Adjusted Rand Index, to assess the quality of the clusters formed and adjust parameters if necessary.

By following these best practices, you can ensure more effective use of Affinity Propagation and obtain meaningful insights from your data.

Also Read: What is Cluster Analysis in Data Mining? Methods, Benefits, and More

Next, let’s look at some of the real-life applications of affinity clustering across different industries.

What are the Use Cases of Affinity Propagation Clustering? 5 Real-Life Examples

Affinity Propagation stands out for its ability to automatically determine the number of clusters based on the data's inherent structure, without the need for the user to specify it in advance. This flexibility makes it particularly valuable in industries dealing with complex, high-dimensional, or dynamic datasets, where the optimal number of clusters is not obvious or may change over time. 

These industries benefit from Affinity Propagation’s ability to reveal the true underlying structure of data, providing more accurate, data-driven insights.

Below are five real-life examples where Affinity Propagation has been successfully applied.

1. Image Segmentation in Computer Vision

In fields like medical imaging or satellite imagery, it is crucial to segment images into meaningful regions for detailed analysis. Traditional methods often require users to specify the number of segments, which may not be obvious or consistent.

• Affinity Propagation automatically clusters pixels based on their similarity in color, intensity, or texture, without needing to define the number of segments. 
• The algorithm adapts to the data and finds the optimal number of clusters based on inherent patterns.

Outcome: This allows for more accurate image segmentation, which is essential in applications like detecting tumors in medical scans or differentiating landforms in satellite imagery. The method’s ability to handle complex and large images without predefining clusters makes it especially valuable in these fields.

Also Read: Image Segmentation Techniques [Step By Step Implementation] 

2. Recommendation Systems in E-Commerce

E-commerce platforms aim to recommend products to users based on their preferences, but defining the number of product categories or clusters manually can be difficult and inefficient.

• Affinity Propagation is applied to group products or users based on similarity metrics, such as browsing history, purchasing patterns, or ratings. 
• The algorithm dynamically determines the number of clusters needed to accurately represent user interests and product categories.

Outcome: By automatically determining clusters, Affinity Propagation helps generate highly personalized recommendations. This improves user experience, increases conversion rates, and boosts customer satisfaction. The flexibility in clustering enables platforms to adapt to evolving customer preferences.

Also Read: Simple Guide to Build Recommendation System Machine Learning

3. Customer Segmentation in Marketing

Marketers need to segment their customer base to tailor campaigns effectively. Predefining clusters can be limiting, as customer behaviors and preferences evolve over time.

• Affinity Propagation allows businesses to segment customers dynamically based on a variety of factors, such as demographics, purchase history, and online behavior. 
• It identifies clusters of customers without predefining the number of segments.

Outcome: The method’s ability to automatically identify meaningful customer segments helps marketers craft more targeted and effective campaigns, leading to higher engagement rates, better resource allocation, and improved ROI. Its flexibility makes it ideal for fast-paced industries where customer preferences can shift rapidly.

4. Anomaly Detection in Cybersecurity

Detecting anomalies in network traffic, such as potential cyberattacks, is challenging. Traditional methods require setting predefined thresholds, but the patterns of attack can vary greatly.

• Affinity Propagation clusters normal behavior patterns and identifies outliers that do not belong to any established cluster. 
• These outliers can indicate abnormal or malicious activity, such as cyberattacks or fraud attempts.

Outcome: The ability to automatically adapt and identify abnormal behavior in real-time without predefined limits helps enhance the security of IT systems. The flexibility of Affinity Propagation makes it especially useful in dynamic environments like cybersecurity, where attack patterns constantly evolve.

If you need a better understanding of cybersecurity, upGrad’s free Fundamentals of Cybersecurity course can help you. You will learn key concepts, current challenges, and important terminology to protect systems and data.

Also Read: Anomaly Detection With Machine Learning: What You Need To Know?

5. Genomics and Bioinformatics

In genomics, grouping genes based on their expression patterns or sequence similarity is a complex task, especially when the number of gene groups is unknown.

• Affinity Propagation is used to cluster genes with similar expression profiles, automatically identifying the number of gene groups needed for further analysis. 
• The algorithm is particularly useful in large-scale genomic data analysis where predefined cluster counts can be arbitrary and inefficient.

Outcome: This application leads to more accurate and insightful gene clustering, enabling researchers to identify genes associated with specific diseases or biological processes. The flexibility to discover the optimal number of clusters based on data patterns is invaluable for researchers in genomics and bioinformatics.

This flexibility, combined with its robust performance in varied real-world applications, makes it a popular choice for tackling clustering problems in numerous fields.

Also Read: Explanatory Guide to Clustering in Data Mining - Definition, Applications & Algorithms

To solidify your understanding of the Affinity Propagation Clustering in machine learning, test your knowledge with a quiz. It’ll help reinforce the concepts discussed throughout the tutorial and ensure you're ready to apply them in your projects.

Quiz to Test Your Knowledge on Affinity Propagation Clustering

Assess your understanding of Affinity Propagation, its components, advantages, limitations, and best practices by answering the following multiple-choice questions. Test your knowledge now!

1. What is the key feature of Affinity Propagation clustering?
a) It requires the number of clusters to be specified in advance
b) It automatically determines the number of clusters based on the data
c) It only works for numerical data
d) It cannot handle complex datasets

2. What mechanism does Affinity Propagation use to identify exemplars (representative data points) in a dataset?
a) It randomly selects exemplars from the dataset
b) It uses a similarity matrix to find points that are most representative of each cluster
c) It assigns exemplars based on the highest distance between points
d) It selects points that have the highest value in the similarity matrix

3. Which of the following is an advantage of using Affinity Propagation?
a) It can handle large datasets with low computational cost
b) It does not require a predefined number of clusters
c) It always produces the same number of clusters, regardless of the data
d) It is faster than traditional clustering methods like k-means

4. What is a limitation of Affinity Propagation?
a) It cannot handle continuous data
b) It is computationally expensive, especially for large datasets
c) It cannot handle datasets with more than 100 data points
d) It only works for datasets with linear relationships

5. Which matrix is crucial in Affinity Propagation for determining how well points fit as exemplars?
a) Responsibility matrix
b) Distance matrix
c) Similarity matrix
d) Preference matrix

6. What does the ‘preference’ parameter in Affinity Propagation control?
a) The maximum number of clusters formed
b) The likelihood of a point being selected as an exemplar
c) The number of iterations the algorithm runs
d) The maximum similarity value between points

7. In which scenario would Affinity Propagation be most beneficial?
a) When the number of clusters is known in advance
b) When the data has a fixed, uniform distribution
c) When the number of clusters is unknown and needs to be determined from the data
d) When dealing with low-dimensional datasets only

8. How does Affinity Propagation handle data points that do not fit well into any cluster?
a) It discards points that do not fit well
b) It assigns them to the nearest exemplar, even if the similarity is low
c) It automatically removes outliers from the dataset
d) It assigns them a zero value in the similarity matrix

9. Which of the following is a best practice when implementing Affinity Propagation clustering?
a) Always set the preference parameter to the average similarity value
b) Tune the preference parameter to control the number of clusters
c) Use Affinity Propagation only for small datasets
d) Ignore the similarity matrix and rely on pre-existing labels

10. What type of data is Affinity Propagation particularly useful for?
a) Datasets with a known number of clusters
b) Datasets with high dimensionality and unknown cluster structure
c) Datasets where all points are equally important
d) Datasets where clusters are linearly separable

This quiz will help you evaluate your understanding of the Affinity Propagation clustering method, its strengths, limitations, and how it operates in various applications.

Also Read: 5 Breakthrough Applications of Machine Learning

You can also continue expanding your skills in machine learning with upGrad, which will help you deepen your understanding of advanced ML concepts and real-world applications.

Upskill with upGrad to Stay Ahead of Industry Trends! 

upGrad’s courses provide expert training in machine learning, with a focus on anonymous classes, their practical applications, and best practices. Learn how to optimize your machine learning models for different scenarios.

While the course covered in the tutorial can significantly improve your knowledge, here are some free courses to facilitate your continued learning:

• Hypothesis Testing
• Unsupervised Learning: Clustering
• Logistic Regression for Beginners
• Linear Regression - Step by Step Guide
• Introduction to Natural Language Processing

You can also get personalized career counseling with upGrad to guide your career path, or visit your nearest upGrad center and start hands-on training today! 

Similar Reads:

• Artificial Intelligence Tutorial
• Machine Learning Tutorials
• Deep Learning Tutorial: Master AI & Neural Networks
• Top 10 Machine Learning Applications in 2025 and the Role of Edge Computing
• Q Learning
• Bagging in Machine Learning: Overview, Steps, Benefits & Applications
• Cost Function In Machine Learning
• Image Annotation in Machine Learning 
• Quantum Computing 
• Bootstrap Aggregation 
• Mahalanobis Distance: Formula, Python Code, Applications & Best Practices
• Support Vector Machine Svm For Anomaly Detection 
• Isolation Forest Algorithm for Anomaly Detection
• Exponential Smoothing Method in Forecasting: Techniques and Applications
• Time Series Forecasting with ARIMA Models: Components, Advantages & Steps

Expand your expertise with the best resources available. Browse the programs below to find your ideal fit in Best Machine Learning and AI Courses Online.

Best Machine Learning and AI Courses Online

Discover in-demand Machine Learning skills to expand your expertise. Explore the programs below to find the perfect fit for your goals.

In-demand Machine Learning Skills

Discover popular AI and ML blogs and free courses to deepen your expertise. Explore the programs below to find your perfect fit.

Popular AI and ML Blogs & Free Courses