Can I use kmeans matlab function to perform k-medoids algorithm ?

Hi I am kind of new to the clustering algorithm so apologize for the bad questions first. I notice that matlab has kmeans built-in function and it can be specified to find component-wise centroid instead of mean by using kmeasn(X,clusterNum,"distance" "city"). I am wondering how this differs from k-medoids algorithm. Does this mean it performs a k-medoids algorithm ? Or it is still a kmeans algorithm ? Is there some way to configure the kmeans function into k-medoids or there is another function for k-medioids ?

Thank you very much,

Answer 1

[the cyclist]

No, but there is a k-medoids function in the File Exchange:

http://www.mathworks.com/matlabcentral/fileexchange/28898

[Cassie]

Thank you very much for your help and I will try the k-medoids function you showed. Then, may I also ask what does it mean when we use kmeans function with "Distance" "city" setting ? What is a component-wise kmeans comcpared with k-medoids ? They seem really similar ?

[the cyclist]

I am not an expert in this, but my understanding is that a k-medoids are required to be members of the original set, whereas the k-means are not. The "distance" parameter just specifies which mathematical definition of distance will be used.

[Also, please "accept" this solution if you found it helpful, for the sake of future seekers of knowledge.]

[Cassie]

Thank you very much,

[Farrukh Hassan]

I am trying to calculate different clustering validation indices such as daviesbouldin, gap, silhouette and CalinskiHarabasz for K-means, K-medoids, and heirarichal clustering.i am trying to calculate daviesbouldin and gap index value for k-mediods but i didnt find any solution. if anyone can help.