Slipstream: Locality-Aware Graph Index Construction for Streaming Approximate Nearest Neighbor Search
Researchers have developed Slipstream, a novel method designed to accelerate approximate nearest neighbor search (ANNS) in streaming vector data. This approach leverages the continuity of vector streams by initiating searches from promising candidates identified during previous insertions, rather than starting from scratch. Slipstream has been integrated into popular libraries like Faiss and HNSWLib, demonstrating up to 30.8 times higher throughput while maintaining a recall rate of at least 0.95. AI
IMPACT Accelerates real-time vector search for applications like recommendation systems and similarity matching.