Fuzzy C-Means Clustering — Soft Unsupervised Learning

Fuzzy C-Means

Clusters c 3

Fuzzifier m 2.00

Cluster Legend

Algorithm State

Iteration

Objective J_m

—

Avg. crispness

—

ΔCentroid

—

How it works

Fuzzy c-means (FCM) is unsupervised soft clustering: unlike k-means, a point is not forced into one cluster — it gets a membership degree u_ij ∈ [0,1] to every cluster j, with Σ_j u_ij = 1. It minimizes:

J_m = Σ_i Σ_j  u_ij^m · ‖x_i − c_j‖²

centroid update:  c_j = ( Σ_i u_ij^m · x_i ) / ( Σ_i u_ij^m )
membership update: u_ij = 1 / Σ_k ( d_ij / d_ik )^( 2/(m−1) )

c — how many cluster centroids compete for the points.
m (fuzzifier) — m→1 makes membership nearly crisp (k-means-like); larger m spreads membership evenly across clusters, blurring boundaries. This sim blends each point's render color by its own membership vector, so softer m visibly desaturates points near a boundary toward gray/mixed hues.
Step / Run — each step recomputes centroids from the current memberships, then recomputes memberships from the new centroids (alternating optimization), exactly as scikit-learn / MATLAB's fcm do it. J_m decreases monotonically until convergence.
ΔCentroid — total centroid movement since the last step; near zero means the algorithm has converged.

Real-world use: FCM is the standard tool when class boundaries are genuinely ambiguous — image segmentation with partial-volume pixels, gene-expression profiles that belong partly to two pathways, or customer segments that straddle two personas.

Points — colored by soft membership Centroids

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.

Artificial Intelligence Unsupervised Learning Simulation