patann 0.0.20

PatANN is a massively parallel, distributed, and scalable in-memory/on-disk vector database library for efficient nearest neighbor search across large-scale datasets by finding vector patterns.

PatANN - Pattern-Aware Vector Database / ANN

Overview

PatANN is a massively parallel, distributed, and scalable in-memory/on-disk vector database library for efficient nearest neighbor search across large-scale datasets by finding vector patterns.

PatANN leverages patterns for data partitioning similar to Google ScANN, implements disk-based I/O similar to DiskANN, and employs search techniques like HNSWlib, resulting in an algorithm that combines the best features to outperform existing approaches.

Status

Beta Version: Currently uploaded for benchmarking purposes. Complete documentation and updates are under development. Not for production use yet.

Platforms

• Beta Version: Restricted to Linux to prevent premature circulation of beta versions
• Production Releases (late Feb 2024)*: Will support all platforms that are supported by mesibo

Key Features

• Faster Index building and Searching
• Supports both in-memory and on-disk operations
• Dynamic sharding to partition and load balance across servers
• Refined search, filtering and pagination
• Unlimited scalability without pre-specified capacity

Algorithmic Approach

• Combines modified NSW (Navigable Small World) graph with a novel pattern based partitioning algorithm
• Preliminary results show phenomenal performance in building index and searching
• Potential slight variations in lower-end matching
• Detailed research paper forthcoming

Contributions

We are seeking help to:

• Run additional datasets. So far, all tested datasets (including self-generated) exhibit patterns that helps algorithm. We have yet to test datasets without clear patterns or with uniform distribution.
• Validate and improve the algorithm

Contact

For support / questions, please contact: [e-mail]