You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I tried NGT in some data, and I am very impressed by its recall and speed. I noticed that most of the examples are for topK, and I am wondering this algorithm is fit for range_search – search for all vectors within a specific radius of a given vector.
Is this algorithm fit for range_search (by make size=data_size and control radius)? If so, do we have data compared with other approaches (classic inverted files, vp-tree, or other approaches)? Thanks!
The text was updated successfully, but these errors were encountered:
NGT can basically deal with range search excluding the python bindings. When you set a radius, the radius is only set to the initial radius instead of infinity for the KNN search. Although I have not exactly compared NGT with other methods by using range search, I think inverted-file-based methods are advantageous for wide range search compared to graph-based methods.
I tried NGT in some data, and I am very impressed by its recall and speed. I noticed that most of the examples are for topK, and I am wondering this algorithm is fit for
range_search
– search for all vectors within a specific radius of a given vector.In the python wrapper, the radius setting is not exposed https://github.com/yahoojapan/NGT/blob/master/python/src/ngtpy.cpp#L126. However, according to the C++ example https://github.com/yahoojapan/NGT/blob/master/bin/search/search.cpp#L54, we can set the radius along with size.
Is this algorithm fit for
range_search
(by makesize=data_size
and controlradius
)? If so, do we have data compared with other approaches (classic inverted files, vp-tree, or other approaches)? Thanks!The text was updated successfully, but these errors were encountered: