In this paper, we introduce a generally applicable algorithm model named Macro-to-Micro Model (M 2M) which is derived from human thinking pattern. The data structure for the nearest neighbor problem based on M 2M can be built in O(n) time. It can also be finished in O(1) time by parallel technology. Moreover, the insertion, deletion and query operation can be completed in constant time without the problem of breaking the balance of tree. And the most noteworthy is that this data structure and preprocessing operation can be shared with most M2M algorithm, so that we can hugely improve the efficiency of the multi-operation problem like image processing and pattern recognition. We mainly focus on the nearest neighbour (NN) searching algorithm in this paper. The M2M approach can achieve the optimal expected time complexity. And the comparative experiment between M 2M and kd-tree shows the great advantage of the former.
[1]
Sunil Arya,et al.
An optimal algorithm for approximate nearest neighbor searching fixed dimensions
,
1998,
JACM.
[2]
Jon Louis Bentley,et al.
An Algorithm for Finding Best Matches in Logarithmic Expected Time
,
1977,
TOMS.
[3]
Bruce W. Weide,et al.
Optimal Expected-Time Algorithms for Closest Point Problems
,
1980,
TOMS.
[4]
Marshall W. Bern,et al.
Approximate Closest-Point Queries in High Dimensions
,
1993,
Inf. Process. Lett..
[5]
David Eppstein,et al.
Parallel Construction of Quadtrees and Quality Triangulations
,
1993,
WADS.
[6]
Alexander G. Gray,et al.
Efficient exact k-NN and nonparametric classification in high dimensions
,
2003,
NIPS 2003.
[7]
Ronald L. Rivest,et al.
Introduction to Algorithms, Second Edition
,
2001
.