Geometric Approach
The Main Problem:
- fast, indexed, complex searches of Terabytes in k-dim space
- searches are not necessary parallel to the axes => traditional indexing (b-tree) does not work
Geometric Approach:
- Use the geometric nature of the k-dimensional data
- Quantize data into containers of ‘friends’: objects of similar colors close on the sky stored together => efficient cache performance
- Containers represent a coarse grained density map of the data multidimensional index tree: k-d tree + r-tree