hnswlib
hnswlib copied to clipboard
Header-only C++/python library for fast approximate nearest neighbors
I was trying to reproduce the idea in [this paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/XboxInnerProduct.pdf), where you can prepend a number to your vectors and have l2 space behave exactly like dot product space (look...
Hey, I have noticed that in [getRandomLevel](https://github.com/nmslib/hnswlib/blob/master/hnswlib/hnswalg.h#L148), where the maximum level for a new element in the HNSW graph is generated, the generated number `r` is rounded to the largest...
@[yurymalkov](https://github.com/nmslib/hnswlib/commits?author=yurymalkov) I need to get the adjacency of the points in each layer. Thx.
I have init a HierarchicalNSW withm space=L2,maxelements=30000, M=16, ef_construction=200 ef=100 rand_seed=100. and then call addPoint to add feature, at first, it is very fast, but when it goes to 200,...
I see a huge potential of improvement at a very low cost. Currently indexes are very optimized in size (and one can control it with parameters) but we still have...
Hi, I'm wondering why there is no cosine space in hnswlib. How to use cosine similarity in C++ version? Thanks
- When I add some points and delete all points, unordered_map remains item (bug)
Dear team, This is a question on whether we can parallel hnsw on different computing node (so distributed memory) but not just on one node. If not, I split my...
Hi there, here's a tiny PR to fix some typos in the HNSW params' docs :slightly_smiling_face:
Thanks for the great implementation, the performance is impressive. I was wondering what it would take to add the ability to delete an element from the graph without degrading too...