MRPT

MRPT (Multi-Resolution Proximity Trees) is an open-source library for fast approximate nearest neighbor (ANN) search in high-dimensional vector spaces. It is written in C++11 and provides Python bindings. MRPT is designed to be lightweight, easy to use, and efficient, making it suitable for vector database backends and other applications requiring ANN search.

• Source: GitHub Repository
• Category: Open-source
• Tags: open-source, ann, high-dimensional, vector-search

Features

• Approximate Nearest Neighbor Search: Efficient ANN search in high-dimensional spaces.
• C++11 Implementation: Core library written in modern C++ (header-only).
• Python Bindings: Easy integration with Python via pip package (pip install mrpt).
• Random Projection Trees: Uses a collection of random projection trees to index data for fast querying.
• Automatic Hyperparameter Tuning: Integrated autotuning algorithm; only the target recall level and number of neighbors need to be specified. No manual hyperparameter tuning required.
• Euclidean Distance Metric: Currently supports Euclidean distance for similarity measurement.
• Minimal Dependencies: Only dependency is Eigen (linear algebra library), bundled with the library.
• No Compilation Required (Header-only): For C++, just include the header file.
• Examples Provided: Minimal usage examples for both Python and C++.
• Open Source: MIT license.

Usage

• Python: Install via pip install mrpt.
• C++: Include the cpp/Mrpt.h header and link with Eigen.

Documentation

• C++ API documentation
• Algorithm description and complexity analysis
• Automatic hyperparameter tuning article

Pricing

• Free and open-source (MIT License)