3 ms·
Is this more or less a k-d tree as a service? Where any distance function can be used to index the data? Or is it something different?
by CoolGuySteve 5y ago
Is this more or less a k-d tree as a service? Where any distance function can be used to index the data?
Or is it something different?
- monkeybutton 5y agoA k-d tree gives you exact answers to nearest neighbour queries.
- srean 5y agoA k-d tree is a data structure. Whether you use that for exact nearest neighbor query or approximate is up to the algorithm used. K-d trees work well for a handful of dimension beyond that it becomes quite expensive.
- hamilyon2 5y agoI thought k-d trees were useless in high-dimentional case. So, it must be something else.
- contravariant 5y agoI'd say they're about as useful as euclidean distance is.
- ahurmazda 5y agoMore or less but as always the devil is in the detail. Here is a paper[1] that summarizes issues with naive approaches. Incidentally. the proposed solution (Hierarchical NSW) in this paper performed fairly well in the industry benchmarks [1] https://arxiv.org/ftp/arxiv/papers/1603/1603.09320.pdf https://arxiv.org/ftp/arxiv/papers/1603/1603.09320.pdf