nanoflann icon indicating copy to clipboard operation
nanoflann copied to clipboard

Published 20 hours ago • verified publisher icon jlblancoc

Re-use previous node for faster lookup?

Open Philip-Trettner opened this issue 6 years ago • 2 comments

I'm doing a lot of findNeighbors (for k = 1) close to each other. The results (and tree traversal) should be pretty coherent. Is there a way to re-use the previous computation for potential speed-up? Like starting at the node where the previous nearest-neighbor is located or something?

Philip-Trettner avatar May 14 '18 11:05 Philip-Trettner

I'd also be interested in knowing if this is possible. We're traversing a regular domain domain queries which are always close to the previous query point.

If this information were to be leveraged, should we expect significant improvements in efficiency?

jmansour avatar May 30 '19 07:05 jmansour

If not starting from the root of the tree, how can you 100% sure the previous saved points is the nearest neighbour?

libralibra avatar Jul 04 '19 19:07 libralibra