A scalable solution to the nearest neighbor search problem through local-search methods on neighbor graphs

Nearest neighbor search is a powerful abstraction for data access; however, indexing troublesome even approximate indexes. For intrinsically high-dimensional data, high-quality fast searches demand either indexes with impractically large memory usage or preprocessing time. In this paper, we introduce an algorithm to solve nearest-neighbor query q by minimizing kernel function defined the distance from each object in database. The minimization performed using metaheuristics problem rapidly; when some methods literature use strategy behind scenes, our approach first one it explicitly. We also provide two approaches select edges graph’s construction stage that limit footprint and reduce number of free parameters simultaneously. carry out thorough experimental comparison state-of-the-art through synthetic real-world datasets; found contributions achieve competitive performances regarding speed, accuracy, almost any benchmarks.