Classification engineering using regional locality-sensitive hashing (LSH) searches

A non-transitory computer readable medium having instructions stored thereon executable by a processor to:

• identify a search point contained in geo-located data, the search point being nearest in distance to a query location;
• determine a search region associated with the search point, the search region comprising a plurality of neighbor points contained in the geo-located data that are nearest k points by distance to the search point;
• assign the plurality of neighbor points to corresponding Product Quantization (PQ) pipelines based on PQ point classes including corresponding sets of subspace centroids;
• compare the PQ point classes in the search region with other regions contained in the geo-located data to determine similar regions using a Locality Sensitivity Hashing LSH) regional search heuristic;
• determine a plurality of counts corresponding to a number of subspace centroid matches, by performing vectorization with a vectorized loop in parallel to provide fast performance, the vectorization being provided at a hardware level;
• order the plurality of counts in descending order, corresponding to listing the other regions having the highest similarity first;
• and
• generate an ordered list of the other regions based on their respective similarities to the search region.