Indyk and Naor introduced embeddings that preserve this algorithm’s output for sets with a low doubling constant or low aspect ratio. A “condensed” version of this algorithm that uses prototypes to reduce the size of the dataset was developed by Peter Hart. (r1, r2, p1, p2)-sensitive families of functions were originally introduced to perform this algorithm using locality-sensitive hashing. Cover and Hart showed that the simplest version of this algorithm has error bounded by two times the Bayes (-5[1])error rate. Usage of a (*) k-d tree allows single queries in this algorithm to be computed in O(log (-5[1])n) time. When used for classification, this algorithm’s namesake parameter is often chosen to be odd to avoid ties. For 10 points, distance-based or simple majority voting can be used in what classification algorithm that examines (10[1])close samples to an input? ■END■ (10[2])