Lianrui Zuo
Atlas
1981

Fuzzy c-means

When a sample sits between two classes, how much of it belongs to each?

Primary source J. C. Bezdek, “Pattern Recognition with Fuzzy Objective Function Algorithms,” Plenum Press, New York, 1981. doi:10.1007/978-1-4757-0450-1

Fuzzy c-means on points in two dimensions, one half-step at a time. Every point has one unit of membership to split among the classes. Memberships: each point splits its unit among the rings, more to the nearer ones; its colour is saturated when almost all of it goes to one class and pale grey when it is split evenly. Representatives: each ring moves to an average of all the points, each counted by how much of it belongs to that ring. The plot on the right records the cost after every half-step. Try m = 1.2 and m = 4, and drag the rings to choose your own start.

The full lecture is open to students in ECE 6357. Enter the course password to read it.