Build the dataset from file

compute l2-squared norms of data stored in row major num_points * dim, needs to be pre-allocated

This method actually restructures the DiskANN index to remove the items that have been marked for deletion.

given training data in train_data of dimensions num_train * dim, generate PQ pivots using k-means algorithm to partition the co-ordinates into num_pq_

k-means algorithm interface

load index from a file stream

* Marks an element with the given label deleted, does NOT really change the current graph. */

Copies data from the source slice to the destination box.

Prunes the neighbors of a given data point based on some criteria and returns a list of pruned ids. # Arguments `location` - The id of the data poin

search for point # Arguments `query` - query vertex `scratch` - in-memory query scratch TODO: use_filter, filteredLindex

assumes memory allocated for pivot_data as new float[num_centers*dim] and select randomly num_centers points as pivots

Sets an artificial capacity of the NeighborPriorityQueue. For benchmarking purposes only.

register for push notifications