Improving Clustering-Based Approximate Nearest Neighbor Search through Learning-to-Rank

Learning a ranking function for routing queries to the most relevant partitions can consistently improve the accuracy of clustering-based approximate nearest neighbor search.