KGraph

KGraph is an open-source library for fast approximate nearest neighbor (ANN) search in high-dimensional vector spaces, with applications in vector database solutions and similarity search.

Features

• k-NN Graph Construction: Builds k-nearest neighbor graphs for datasets.
• Online k-NN Search: Supports fast approximate search using the constructed k-NN graph as an index.
• Heuristic Algorithms: Implements generic and fast heuristic algorithms for ANN search.
• C++ API: Main API is in C++, providing maximum flexibility and performance. Users define custom similarity functions via oracles (callback classes).
• Python Wrapper: Provides a Python API (kgraph), supporting Euclidean and Angular distances on rows of NumPy matrices.
• Parameter Tuning: Both index construction and search support various optional parameters to tune performance.
• Efficient Oracle Implementations: Includes oracle implementations for common similarity measures.
• No Assumptions on Similarity Properties: Algorithms do not assume properties like the triangle inequality.
• Installation: Can be built using CMake or a provided Makefile; Python API installable with python setup.py install.
• Open Source: Source code available on GitHub under an open-source license.

Pricing

KGraph is open-source and free to use.

Links

• GitHub Repository
• Doxygen Documentation