Score0Colour Apprentice
Iteration0Ready
Inertia—Lower is tighter
Silhouette—Higher is clearer
Puzzle Setup
K = 4
48 colours
Display and Accessibility
Interactive Colour Map
Choose settings, generate a palette, then select or initialise centres.
Click a colour to inspect distances. In player-selected mode, click samples to choose centres. Drag centre stars to test alternatives.
Cluster Centres
K-means alternates between assigning samples and updating each centroid.
Selected Colour Calculation
Select a colour sample to inspect its values and distances.
Plotly Analytics
Run Comparison and Results
| Run | K | Iterations | Inertia | Silhouette | Score | Initialisation | Space |
|---|---|---|---|---|---|---|---|
| Current | 4 | 0 | — | — | 0 | K-means++ | RGB |
How the Puzzle Works
1. Choose K
Select how many colour groups the model should find.
2. Assign Samples
Each colour joins its nearest centre under the chosen distance.
3. Update Centres
Each centroid becomes the mean of its assigned colour samples.
assignment(x) = arg min distance(x, μk) | μk = mean(samples assigned to cluster k) | inertia = Σ ||x - μassignment(x)||²Learning Notes
Important ideas
K-means seeks compact groups around centroids. Initial centres can change the final solution. Different colour spaces can change perceived similarity.
Common edge cases
Empty clusters need replacement centres. Outliers can pull means away from dense groups. Too many clusters may split natural colour families.