fast-RNN

Fast Reciprocal Nearest Neighbors

Abstract

We release here software for accelerating the Reciprocal Nearest Neighbors (RNN) clustering algorithm, i.e. the fast-RNN. We speed up the nearest neighbor chains construction via a novel dynamic slicing strategy for projection search. A C++ implementation is made publicly available.

Downloads

• fast-RNN V2.0 (Source code) and Linux (32 and 64) binaries. Last Update: 11-April-2012.
• fast-RNN V1.0 (Source code). Last Update: 11-March-2011.
• Standard RNN. Last Update: 4-March-2011.
• Download data in Matlab format (370 MB) to replicate the experiments in our paper.

You may download the paper too:

Fast Reciprocal Nearest Neighbors Clustering, R. J. López-Sastre, D. Oñoro-Rubio, P. Gil-Jiménez, S. Maldonado-Bascón. Signal Processing, 2012.

Citing

If you make use of the fast-RNN software, please cite the following reference in any publications:

@ARTICLE{fastrnn2012,
  author = {Lopez-Sastre, R.~J. and Onoro-Rubio, Daniel and Gil-Jimenez, P. and Maldonado-Bascon, S.},
  title = {Fast Reciprocal Nearest Neighbors Clustering},
  journal = {Signal Processing},
  year = {2012},
  volume = {92},
  pages = {270--275}
}