Souti K-Means

#Solution
def K_means_clustering(X, n_clusters=2, seed=1, num_iterations=10):
    # Initialitize centroids based on a random selection of #n_clusters samples of X
    rng = np.random.RandomState(seed)
    i = rng.permutation(X.shape[0])[:n_clusters]
    centroids = X[i]
    labels = []

    #Repeat the process during num_iterations or convergence achieved
    for num in range(0, num_iterations):
        distancesMatrix = np.matrix([calculate_distance(X, centroid) for centroid in centroids]).T

    #For each iteration, calculate the shortest distance of each point of X to centroids
    #Labels are based on the index in the centroids array
        labels = [distancesMatrix[j].argmin() for j in range(len(distancesMatrix))]

    #Calculate the new centroids based on the means of each point assigned to each cluster
        new_centroids= [np.array([0 for _ in range(X.shape[1])]) for n in range(n_clusters)]
        for j in range(len(X)):
            new_centroids[labels[j]] = new_centroids[labels[j]] + X[j]
        for j in range(len(new_centroids)):
            new_centroids[j] = new_centroids[j]/labels.count(j)

    # Evaluate convergence: if new_centroids=centroids, stop iterations
        if np.array_equal(centroids, new_centroids):
            print('Convergence achieved with:',num+1, 'iterations')
            break
        else:
            if (num+1)%10 == 0 and num != 0:
                print('No convergence yet after', num+1, 'iterations')
        centroids = new_centroids
    # print(labels.count(0), labels.count(1), labels.count(2))
    return centroids, labels