k-medians clustering technique

please, i want matlab code of k-median clustering technique

Answer 1

[NVSL]

Hello @muhammad ismat

I recently looked into whether there's a function for k-median clustering in MATLAB, or if we could tweak the existing "k-means" or "k-medoids" functions to do the job.

I couldn't find anything ready-made, but I figured out that we can use MATLAB's "median" function to create our own "kMedianClustering" function. Hope this helps!

function [centroids, idx] = kMedianClustering(X, k, maxIter)
    % X: data points (n x d matrix)
    % k: number of clusters
    % maxIter: maximum number of iterations

    % Initialize centroids randomly
    [n, d] = size(X);
    centroids = X(randperm(n, k), :);
    idx = zeros(n, 1);

    for iter = 1:maxIter
        % Assign each point to the nearest centroid
        for i = 1:n
            distances = sum(abs(X(i, :) - centroids), 2);
            [~, idx(i)] = min(distances);
        end
        
        % Update centroids
        newCentroids = zeros(k, d);
        for j = 1:k
            clusterPoints = X(idx == j, :);
            if ~isempty(clusterPoints)
                newCentroids(j, :) = median(clusterPoints, 1);
            else
                newCentroids(j, :) = centroids(j, :);
            end
        end

        % Check for convergence
        if all(newCentroids == centroids)
            break;
        end

        centroids = newCentroids;
    end
end

% randomly generated data
data = [randn(50, 2) + 1; randn(50, 2) - 1];

% Number of clusters
k = 2;

% Maximum number of iterations
maxIter = 100;

% Perform k-median clustering
[centroids, idx] = kMedianClustering(data, k, maxIter);

% Plot the results
figure;
hold on;
colors = ['r', 'g', 'b'];
for i = 1:k
    scatter(data(idx == i, 1), data(idx == i, 2), 36, colors(i), 'filled');
end
scatter(centroids(:, 1), centroids(:, 2), 100, 'kx', 'LineWidth', 3);
hold off;
title('K-Median Clustering');

●

[Federico Maddanu]

I think (but I m not sure) kmeans does this when you select 'cityblock' as distance!