PRE-ALPHA CODE: BEWARE

Author: Eric Christiansen

An implementation of metric ball trees for fast nearest neighbor in an arbitrary metric space. Unlike techniques like kd-trees or locality sensitive hashing, metric ball trees do not require the data be embedded, they simply require the user provide a metric function for comparing two objects. The hope is to provide a tool like the Fast Library for Approximate Nearest Neighbor (http://people.cs.ubc.ca/~mariusm/index.php/FLANN/FLANN) for metrics which are not easily embedded.

This library is benchmarked in the project at:

https://github.com/emchristiansen/benchmark-mbtree

Note the above is also a good place to look for usage examples. In addition, the benchmarks are discussed at:

http://emc-fall2011-fastdoppelganger.blogspot.com/

References:
http://en.wikipedia.org/wiki/Metric_tree