Cover tree

The cover tree is a type of data structure in computer science that is specifically designed to facilitate the speed-up of a nearest neighbor search. It is a refinement of the Navigating Net data structure, and related to a variety of other data structures developed for indexing intrinsically low-dimensional data.^[1]

The tree can be thought of as a hierarchy of levels with the top level containing the root point and the bottom level containing every point in the metric space. Each level C is associated with an integer value i that decrements by one as the tree is descended. Each level C in the cover tree has three important properties:

Nesting: $C_{i} \subseteq C_{i-1}$
Covering: For every point $p \in C_{i-1}$ , there exists a point $q \in C_{i}$ such that the distance from $p$ to $q$ is less than or equal to $2^{i}$ and exactly one such $q$ is a parent of $p$ .
Separation: For all points $p,q \in C_i$ , the distance from $p$ to $q$ is greater than $2^{i}$ .

1 Complexity
2 See also
3 References

Complexity

Find

Like other metric trees the cover tree allows for nearest neighbor searches in $O(\eta*\log{n})$ where $\eta$ is a constant associated with the dimensionality of the dataset and n is the cardinality. To compare, a basic linear search requires $O(n)$ , which is a much worse dependence on $n$ . However, in high-dimensional metric spaces the $\eta$ constant is non-trivial, which means it cannot be ignored in complexity analysis. Unlike other metric trees, the cover tree has a theoretical bound on its constant that is based on the dataset's expansion constant or doubling constant (in the case of approximate NN retrieval). The bound on search time is $O(c^{12} \log{n})$ where $c$ is the expansion constant of the dataset.

Insert

Although cover trees provide faster searches than the naive approach, this advantage must be weighed with the additional cost of maintaining the data structure. In a naive approach adding a new point to the dataset is trivial because order does not need to be preserved, but in a cover tree it can take $O(c^6 \log{n})$ time. However, this is an upper-bound, and some techniques have been implemented that seem to improve the performance in practice.^[2]

Space

The cover tree uses implicit representation to keep track of repeated points. Thus, it only requires O(n) space.

References

^ Kenneth Clarkson. Nearest-neighbor searching and metric space dimensions. In G. Shakhnarovich, T. Darrell, and P. Indyk, editors, Nearest-Neighbor Methods for Learning and Vision: Theory and Practice, pages 15--59. MIT Press, 2006.
^ http://hunch.net/~jl/projects/cover_tree/cover_tree.html

Alina Beygelzimer, Sham Kakade, and John Langford. Cover Trees for Nearest Neighbor. In Proc. International Conference on Machine Learning (ICML), 2006.
JL's Cover Tree page. John Langford's page links to papers and code.
A C++ Cover Tree implementation on GitHub.

Trees in computer science

Binary trees	Binary search tree (BST) · Cartesian tree · Top tree · T-tree

Self-balancing binary search trees	AA tree · AVL tree · Red–black tree · Scapegoat tree · Splay tree · Treap

B-trees	B+ tree · B*-tree · B^x-tree · UB-tree · 2-3 tree · 2-3-4 tree · (a,b)-tree · Dancing tree · Htree

Tries	Suffix tree · Radix tree · Ternary search tree · X-fast trie · Y-fast trie

Binary space partitioning (BSP) trees	Quadtree · Octree · k-d tree · Implicit k-d tree · vp-tree

Non-binary trees	Exponential tree · Fusion tree · Interval tree · PQ tree · Range tree · SPQR tree · Van Emde Boas tree

Spatial data partitioning trees	R-tree · R+ tree · R* tree · X-tree · M-tree · Segment tree · Hilbert R-tree · Priority R-tree

Other trees	Heap · Hash tree · Finger tree · Metric tree · Cover tree · BK-tree · Doubly-chained tree · iDistance · Link-cut tree · Fenwick tree