SINKING IMPLEMENTATION COSTS WITH ELEVATED LIKELIHOOD IN SEARCHING
Mulugurthi Sarath Kumar, Naresh Sunkara · IJITR International Journal of Innovative Technology and Research - IJITR International Journal of Innovative Technology and Research · 2016
A proper theoretical analysis implies that with high probability, the RCT returns a proper query lead to time that will depend very competitively on the way of measuring the intrinsic dimensionality from the data set. Objects are selected based on their ranks with regards to the query object, allowing much tighter control around the overall execution costs. This paper introduces an information structure for k-NN search, the Rank Cover Tree (RCT), whose pruning tests depend exclusively around the comparison of similarity values other qualities from the underlying space, like the triangular inequality, aren't employed. Additionally they reveal that the RCT is capable of doing meeting or exceeding the amount of performance of condition-of-the-art techniques that utilize metric pruning or any other selection tests involving statistical constraints on distance values. The experimental recent results for the RCT reveal that non-metric pruning techniques for similarity search could be practical even if your representational dimension from the information is very high. Experimental evidence indicating that for practical k-NN search applications, our rank-based technique is very as good as approaches which make explicit utilization of similarity constraints.