An Algorithm for Finding Best Matches in Logarithmic Expected Time
Summary (1 min read)
Analysis of the Performance
- The storage required for file organization is proportionalto the file size, N. The discriminating key number and partition value must be stored for each nonterminal node of the k-d tree.
- The computation required to build the k-d tree is easily derived.
- At each level of the tree, the entire set of key values must be scanned.
- The expected time performance of the search is not so easily derived.
Implementation
- The above discussion has centered on the expected number of records examined as the sole criterion for performance evaluation cf the algorithm.
- The results in Figure 5 show that in two dimensions near-asymptotic behavior occurs even for files as small as 128 records.
- The logarithmic behavior size increases is illustrated of the overall computation as the file for the k-d tree algorithm in Figure 5 , except that for eight dimensions the Comparison of Figure 3 to Figure 5 tation involved in building the tree increase is slightly faster.
- The fraction of computation spent on preprocessing decreases with increasing dimensionality.
Did you find this useful? Give us your feedback
Citations
46,906 citations
Cites methods from "An Algorithm for Finding Best Match..."
...Our keypoint descriptor has a 128-dimensional feature vector, and the best algorithms, such as the k-d tree (Friedman et al., 1977) provide no speedup over exhaustive search for more than about 10 dimensional spaces....
[...]
37,989 citations
23,600 citations
20,196 citations
7,159 citations
References
17,939 citations
7,159 citations
2,048 citations
1,014 citations
"An Algorithm for Finding Best Match..." refers background in this paper
...Such an optimization is known to be NP-complete [7] and thus very likely of nonpolynomial time complexity....
[...]
...HYAYIL, L., Am) RZVEST, R.L. Constructing optimal binary decision trees is NP-complete....
[...]
...Such an optimization is known to be NP-complete [7] and thus very likely of nonpolynomial time complexity....
[...]
776 citations