Pruning Strategies Sample Clauses

Pruning Strategies. The most important part of our algorithm is the pruning phase which shrinks the search space into cells that might include the answer of the RkNN query (i.e. xxxxx- date cells), or might influence them (i.e. influence cells). To exploit the information provided in DBcnt, we utilize three state-of-the-art uncertain pruning strategies which are able to prune the search space using aggregated counts of data in each cell. These methods include i) Min-Max [10], ii) 60-degree [66], and iii) Half-plane [17] which are applied in the order of their efficiency and the reverse order of their pruning power. We use such ordering because Min-Max is simple and more efficient, however it has less pruning ability compared to 60-degree method, so if Min-Max fails to prune a cell, we apply 60-degree method; similarly, half-plane may substitute 60-degree for the same reasons. In all of these methods, if an object B is pruned by another object A with regard to a query q, it means that B cannot be an answer for q as a result of A. (a) Min-Xxx xxxxxxx (b) Self pruning Min-Xxx Xxxxxxx In Min-Max method [10], for a given query q, an object A can prune another object B if MaxDist(A, B) < MinDist(B, q) Figure 4.5a illustrates an example of Min-Max method. A special type of Min- Max pruning is called Self-pruning in which a cell is pruned by itself as shown in Figure 4.5b. Self-pruning For an uncertain object A and query point q, if the number of points in A is greater than k, and MaxDist(A) < MinDist(A, q) then A is self-pruned [100]. MaxDist denotes the maximum distance between any possible points in A (e.g. diameter in a rectangle) and MinDist denotes the minimum distance between A and the query point q.