Type: Package
Title: An Accurate kNN Implementation with Multiple Distance Measures
Version: 1.0
Date: 2019-10-24
Maintainer: Philipp Angerer <philipp.angerer@helmholtz-muenchen.de>
Description: Similarly to the 'FNN' package, this package allows calculation of the k nearest neighbors (kNN) of a data matrix. The implementation is based on cover trees introduced by Alina Beygelzimer, Sham Kakade, and John Langford (2006) <doi:10.1145/1143844.1143857>.
URL: https://github.com/flying-sheep/knn.covertree
BugReports: https://github.com/flying-sheep/knn.covertree/issues
License: AGPL-3
Imports: Rcpp (≥ 1.0.2), RcppEigen (≥ 0.3.3.5.0), Matrix, methods
Suggests: testthat, FNN
LinkingTo: Rcpp, RcppEigen
SystemRequirements: C++11
NeedsCompilation: yes
Encoding: UTF-8
RoxygenNote: 6.1.1
Packaged: 2019-10-24 08:59:51 UTC; angerer
Author: Philipp Angerer ORCID iD [cre, aut], David Crane [cph, aut]
Repository: CRAN
Date/Publication: 2019-10-28 16:00:02 UTC

A not-too-fast but accurate kNN implementation supporting multiple distance measures

Description

A not-too-fast but accurate kNN implementation supporting multiple distance measures


kNN search

Description

k nearest neighbor search with custom distance function.

Usage

find_knn(data, k, ..., query = NULL, distance = c("euclidean",
  "cosine", "rankcor"), sym = TRUE)

Arguments

data

Data matrix

k

Number of nearest neighbors

...

Unused. All parameters to the right of the ... have to be specified by name (e.g. find_knn(data, k, distance = 'cosine'))

query

Query matrix. In knn and knn_asym, query and data are identical

distance

Distance metric to use. Allowed measures: Euclidean distance (default), cosine distance (1-corr(c_1, c_2)) or rank correlation distance (1-corr(rank(c_1), rank(c_2)))

sym

Return a symmetric matrix (as long as query is NULL)?

Value

A list with the entries:

index

A nrow(data) \times k integer matrix containing the indices of the k nearest neighbors for each cell.

dist

A nrow(data) \times k double matrix containing the distances to the k nearest neighbors for each cell.

dist_mat

A dgCMatrix if sym == TRUE, else a dsCMatrix (nrow(query) \times nrow(data)). Any zero in the matrix (except for the diagonal) indicates that the cells in the corresponding pair are close neighbors.

Examples

# The default: symmetricised pairwise distances between all rows
pairwise <- find_knn(mtcars, 5L)
image(as.matrix(pairwise$dist_mat))

# Nearest neighbors of a subset within all
mercedeses <- grepl('Merc', rownames(mtcars))
merc_vs_all <- find_knn(mtcars, 5L, query = mtcars[mercedeses, ])
# Replace row index matrix with row name matrix
matrix(
  rownames(mtcars)[merc_vs_all$index],
  nrow(merc_vs_all$index),
  dimnames = list(rownames(merc_vs_all$index), NULL)
)[, -1]  # 1st nearest neighbor is always the same row