DSAverse

Initializing Sorting Algorithms...

Sorting

Trees

Graphs

Preparing interactive visualizations...

Back to Machine Learning

k-Means Clustering

Assign each point to its nearest centre, move each centre to its points' mean, repeat. Watch the objective fall at every half-step, a bad start get trapped, and why the lowest J cannot choose k.

Builds onk-Nearest Neighbours Norms and Unit Vectors

The engine's k-means, run in your browser · seeded blobs

Act 1 of 6 · The goalstep 1 of 17

X[n, d] =[45, 2]

Speed

45 points and no labels. Clustering asks for k groups; k-means describes each group by a centre and scores a choice of centres by J — the sum of squared distances from every point to its own centre. Low J means tight groups. Here k = 3.

J = \sum_{i=1}^{n} \|x_i - \mu_{c(i)}\|^2, \qquad c(i) = \arg\min_j \|x_i - \mu_j\|^2

k-Means Clustering

Assign each point to its nearest centre, move each centre to its points' mean, repeat. Watch the objective fall at every half-step, a bad start get trapped, and why the lowest J cannot choose k.

Builds onk-Nearest Neighbours Norms and Unit Vectors

The engine's k-means, run in your browser · seeded blobs

Act 1 of 6 · The goalstep 1 of 17

X[n, d] =[45, 2]

Speed

J = \sum_{i=1}^{n} \|x_i - \mu_{c(i)}\|^2, \qquad c(i) = \arg\min_j \|x_i - \mu_j\|^2

DSAverse

k-Means Clustering

See also

k-Means Clustering

See also