All strategies require global knowledge on the number of currently existing replicas and the current query rate for each of the replicas.
Thus, we assume the diverse subset for the current query is already evaluated using Greedy heuristic.
Let (Q_c) be the current query and (X_c) be the respective result set of (Q_c).
First, ECAR captures the missing fragments on the current query MF (Line 4).
Comparing thousands of points from the model with the current query photograph is computationally expensive.
The second method is based on the selection of the k most similar queries to the current query.
Let (Q_c) be the current query with result set (X_c) for which a diverse subset (S_c) needs to be generated.
The main idea underlying the adaptive model approach is to utilize the observations from the current query being diversified.
In particular, we show that, with some care, the regressed formula will contain variables relevant to the current query only, rather than all variables in the problem description.
Open image in new window Fig. 6 Reusable Set (S_{R,c}) for current query result set (X_c) generated from diverse overlapping results of previous queries.
