LMW-tree is a generic template library written in C++ that implements several algorithms that use the m-way nearest neighbor tree structre to store their data. The algorithms and data structures are generic to support different data representations such as dense real valued and bit vectors, and sparse vectors. Additionally, it can index any object type that can form a prototype representation of a set of objects.
The algorithms are primarily focussed on comptutationally efficient clustering. Clustering is an unsupervised machine learning process that finds interesting patterns in data. It places similar items into clusters and dissimilar items into different clusters. The data structures and algorithms can also be used for nearest neighbor search, supervised learning and other machine learning applications.
The package includes EM-tree, K-tree, k-means, TSVQ, repeated k-means, clustering, random projections, random indexing, hashing, bit signatures.
LMW-tree is licensed under the BSD license.