octave-ann

Octave-ann is a software package equipped with powerful bindings that provide users with access to the ANN Library, making it easier to find approximate nearest neighbors in a high-dimensional point set. The ANN library comes with state-of-the-art algorithms and data structures that enable users to perform exact or approximate nearest neighbor queries. Thanks to the bindings, users can now execute queries such as "octave:1> ann octave:2> pts=[-0.297462,0.176102; 0.565538,-0.361496; 0.909313,-0.182785; 0.920712,0.478408; 0.167682,0.0499836; 0.305223,-0.0805835; 0.114973,0.882453; 0.742916,0.16376; 0.0724605,-0.826775; 0.690960,-0.559284; 0.188485,-0.643934; 0.749427,-0.942415; -0.970662,-0.223466; 0.916110,0.879597; 0.927417,-0.382593; -0.711327,0.278713; -0.519172,0.986146; 0.135338,0.924588; -0.0837537,0.61687; 0.0520465,0.896306]; octave:3> kd=ANNkd_tree(pts); octave:4> [nn_idx,dd]=kd.annkPriSearch([0.826225,-0.30962],5)" directly from within Octave.

The returned result is stored in 'nn_idx' which contains the row indices of the 5 (approximate) nearest neighbors with their distances stored in 'dd'. These data are computed in O(k log(n)) or less, depending on the value of 'k', which varies according to the obtained approximation's quality. The software has no restrictions on the number of points with hundreds of dimensions involved, except for the available memory. Once a data set is constructed, it can be repeatedly used to perform multiple lookups. Multiple data structures, algorithms, and parameters can be selected from using this software.

Users seeking more information on these features can visit the official ANN homepage, while attempting the test examples in the bindings/tests/octave directory is an excellent way to gain further insight. These examples serve as practical guides since SWIG/Octave does not presently offer a convenient mechanism to add documentation within wrapper code.