WEKA classes in map and reduce phases of KMeans Clustering on hadoop

Question

I want to use WEKA's classes inside mapreduce program for performing KMeans Clustering on Instances. I just want an overview for map and reduce classes. How the distance computed by WEKA classes be reviewed in Reducer classes?

Answer 1

Short story is: you can't.

If you want a map-reduce algorithm, it needs to be designed as such.

There is no push-button solution that allows you to parallelize all of Wekas code to map-reduce. Otherwise, we would all be running map-reduce for everything.

It literally would need to be rewritten from scratch to fully run on Mapreduce.