From dd9cfa31a80350ee0bd1ec3d7c41a1e0a18a2297 Mon Sep 17 00:00:00 2001 From: Zhao Junwang Date: Tue, 16 Apr 2024 16:06:48 +0800 Subject: [PATCH] add source code link Signed-off-by: Zhao Junwang --- src/databases/vectordb/diskann.md | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/src/databases/vectordb/diskann.md b/src/databases/vectordb/diskann.md index 8949cef..b6f0217 100644 --- a/src/databases/vectordb/diskann.md +++ b/src/databases/vectordb/diskann.md @@ -40,6 +40,10 @@ HNSW，NSG 和 Vamana 本质上都是图算法，都可以抽象为两个操作 DiskANN 通过 BeamSearch(设置 beamwidth 一次读多个数据块) 和缓存最常访问的节点（eg. by caching all vertices that are C = 3 or 4 hops from the starting point s）来加速查询。 另外，DiskANN 将邻居节点的向量保存在磁盘索引文件中，来提高搜索的精度（Implicit Re-Ranking Using Full-Precision Vectors）。 +### Code + +- [microsoft/DiskANN](https://github.com/microsoft/DiskANN), Graph-structured Indices for Scalable, Fast, Fresh and Filtered Approximate Nearest Neighbor Search + ### Further readings - [FreshDiskANN: A Fast and Accurate Graph-Based ANN Index for Streaming Similarity Search](/assets/pdfs/FreshDiskANN_2021.pdf), 2021