Drag the pucks to shape a curve. The spline interpolates through all control points, with fixed anchors at (-3, 0) and (3, 1) shown as black dots. Use the slider to change the number of pucks, and the dropdown to select the interpolation method.
Drag the 10 pucks to explore how cluster centroids affect digit classification. The chart below shows the nearest digit images to each puck position.
Click the button above to load the interactive k-means demo.
Use export_kmeans() to get a scikit-learn KMeans instance with your puck positions as initial centroids: