知识卡片

普通读书笔记卡 · 6766.c · 生成模型:GPT-6

分层近邻图逐级细化候选

理解等级:中级

HNSW以多层图组织向量,上层较稀疏,帮助快速接近查询区域,再到较密层搜索。图连边是edge而非vertex,搜索通常还需底层候选扩展;层数、连接和搜索预算有代价,不是任意停在上层也保留同样准确度。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Next, let’s look at graph-based algorithms. A good representative of graph-based algorithms, the Hierarchical Navigable Small World (HNSW)51 algorithm approaches the problem from the bottom up. It starts by building navigable small world (NSW) graphs, which are graphs where each vector is connected to its closest neighbors by a vertex. To understand the intuition of it, think of the Facebook connections graph—everyone is connected directly

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第10章《真实世界中的大语言模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-10.html) - 结论依据:摘录中关于分层近邻图逐级细化候选的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Next, let’s look at graph-based algorithms. A good representative of graph-based algorithms, the Hierarchical Navigable Small World (HNSW)51 algorithm approaches the problem from the bottom up. It starts by building navigable small world (NSW) graphs, which are graphs where each vector is connected to its closest neighbors by a vertex. To understand the intuition of it, think of the Facebook connections graph—everyone is connected directly