This task consists of creating an index, called List of Clusters (LC) using the model of MPI distributed programming.
This index is used to index multimedia elements, in such a way that process queries efficiently by performing element discarding.
-
The first Cluster Center (C1) corresponds to a database element that is chosen randomly. For the purposes of this task, C1 will be the first vector in the list that will be provided. as a database.
-
The K elements closest to C1 are assigned to the first cluster. That is, the first cluster remains formed by the element C1 plus the K elements closest to it and also by the coverage radius of the cluster. This radius is equal to the distance between C1 and the furthest element in the cluster.
-
The next cluster center is the DB element farthest from C1, and the same step 2) is repeated to form the second cluster.
-
From now on, the next cluster center will be the element e that gets the value of the sum elderly:
where Ci is the ith cluster and dist is the distance function between two elements. Step 2) is repeated to form the cluster.
- The algorithm is iterative and stops when there is no more data to store in a cluster.