knn

KNN (k-nearest-neighbor) search

Contents

Syntax

Description

nnIndex = knn(mat1, mat2) returns the results of KNN search, where

• mat1: the test vectors (represented by columns)
• mat2: the vectors to be searched (represented by columns)
• nnIndex: the KNN indices into mat2 (each column is the result of each column in mat1)

[nnIndex, nnDistance] = knn(mat1, mat2) returns the KNN indices together with their distances (sorted)

[...] = knn(mat1, mat2, opt) takes extra options for KNN search, with

• opt.method: Method for distance computation
• opt.nnCount: K in KNN
• opt.skipIndex: Skip index into mat2 (Used for cross-validation)

Example

mat1=[2 2; 4 6];	% Columns 3 and 5 of mat2
mat2=[3 4 2 3 2; 2 2 4 5 6];
opt=knn('defaultOpt');
opt.nnCount=3;
opt.skipIndex=[5];
pairwiseDist=distPairwise(mat1, mat2);
[nnIndex, nnDistance]=knn(mat1, mat2, opt);
disp(nnIndex), disp(nnDistance), disp(pairwiseDist)

     3     4
     4     3
     1     1

         0    1.4142
    1.4142    2.0000
    2.2361    4.1231

    2.2361    2.8284         0    1.4142    2.0000
    4.1231    4.4721    2.0000    1.4142         0

Published with MATLAB® R2018a