All Tools
N
Vector DBFreeOpen Source
NMSLIB
Fast similarity search for non-metric spaces
Apache-2.0
ABOUT
Most vector similarity search libraries assume metric spaces (where the triangle inequality holds), but many real-world distance functions — KL-divergence, cosine distance, quadratic form distance — are non-metric and incompatible with standard spatial indexing. NMSLIB provides efficient search algorithms including VP-trees and neighborhood graphs that work in non-metric spaces, enabling accurate similarity search for applications where distances don't follow metric axioms.
INSTALL
pip install nmslibINTEGRATION GUIDE
1. Search images by visual similarity using non-metric distance functions like quadratic form distance
2. Perform efficient nearest-neighbor search on probabilistic embeddings where KL-divergence is the natural distance
3. Evaluate and benchmark different k-NN algorithms across a variety of distance metrics and datasets
4. Build recommendation systems using custom similarity measures that don't satisfy the triangle inequality
5. Index and query high-dimensional feature vectors with exotic distance functions for scientific data analysis
TAGS
vector-searchnearest-neighborsimilarity-searchknnnon-metricpythoncpp