Estimating Membership Functions

Theen-Thee Tan (
Mon, 15 Sep 1997 03:28:28 +0200 (MET DST)

I am learning about fuzzy pattern classification. I hope to get some help
from this group as I haven't been able to find references at where I am now.

Given a set of labeled training data I would like to come up with the
membership function for each class.

As far as I have read, commonly there are 3 ways people come up with
membership functions: experts or hueristics, fuzzy clustering (e.g.
fuzzy-c-means, Bezdek), and interpolation from pdf (Keller).

- How exactly clustering being used even when we have labeled training data?

- Other way one gets membership function from a set of training

Please provide some pointers.

Thanks in advance.