This animation revisits four points on a line (0, 2, 8, 10) clustered by Fuzzy C-means instead of hard K-means. Each point gets a two-part membership ring showing its degree of belonging to two centers, computed from inverse-squared distances with fuzzifier m=2. It shows memberships summing to one, the weighted objective function, how centers update using u_ik squared as weights, and why zero distances and equal distances are special cases. Useful for first-year AI students comparing fuzzy clustering to hard K-means and probabilistic models.
Narrated · 16:9 · Preview before teaching · automatic layout checks do not establish subject accuracy
# Fuzzy C-means gives each point a degree of membership Reuse the points 0, 2, 8, and 10, but replace hard cluster colors with two-part membership rings. Start with centers 1 and 9 and fuzzifier m = 2. For the point 2, the distances are 1 and 7, giving memberships 0.98 and 0.02. For point 0, the memberships are 81/82 and 1/82; the remaining two points have symmetric memberships. Show that each point's memberships sum to one. Introduce the objective sum of u_ik^m times squared distance and update each center using u_ik^m as its weight, not u_ik alone. Recalculate memberships after moving the centers. Explain how equal distances give equal memberships and why a zero distance needs special handling. Contrast these fuzzy memberships with probability-model posteriors: they are not automatically the same thing. Use black backgrounds and the same cyan-orange palette as the K-means animation. Audience: 一年级人工智能专业学生