A triplet comparison oracle on a set S takes an object x∈S and for any pair {y,z}⊂S∖{x} declares which of y and z is more similar to x. Partitioned Local Depth (PaLD) supplies a principled non-parametric partitioning of S under such triplet comparisons but needs O(n2logn) oracle calls and O(n3) post-processing steps. We introduce Partitioned Nearest Neighbors Local Depth (PaNNLD), a computationally tractable variant of PaLD leveraging the K-nearest neighbors digraph on S. PaNNLD needs only O(nKlogn) oracle calls, by replacing an oracle call by a coin flip when neither y nor z is adjacent to x in the undirected version of the K-nearest neighbors digraph. By averaging over randomizations, PaNNLD subsequently requires (at best) only O(nK2) post-processing steps. Concentration of measure shows that the probability of randomization-induced error δ in PaNNLD is no more than 2e−δ2K2.
@article{arxiv.2108.08864,
title = {Partitioned K-nearest neighbor local depth for scalable comparison-based learning},
author = {Jacob D. Baron and R. W. R. Darling and J. Laylon Davis and R. Pettit},
journal= {arXiv preprint arXiv:2108.08864},
year = {2021}
}