按年被提到
20172017 · 12 次2026
系统、数据库与网络›数据库与向量检索·arXiv:1712.01208 · cs.DB

The Case for Learned Index Structures

学习型索引结构的论证
Tim Kraska, Alex Beutel, Ed H. Chi, Jeffrey Dean, Neoklis Polyzotis · 2017

主张用学习到的模型取代 B-Tree 等索引,最高提速 70%

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 25 人首次 2017-12-08 · 最近 2026-06-04
25不同的人
15活跃月份
31讨论串
979被引(OpenAlex)

论文把索引看作模型:B-Tree 是把键映射到有序数组中位置的模型,因此可以用包括深度学习在内的其他模型来替代,称为学习型索引。作者从理论上分析了学习型索引何时优于传统索引、设计中的主要挑战,初步结果显示在若干真实数据集上用神经网络可比缓存优化的 B-Tree 快最多 70%,同时内存占用小一个数量级。

HN 为什么提它

HN 上的人把它当作用机器学习重新设计数据库和系统底层组件的代表想法,多数提及集中在 2017 至 2018 年,称它好读、有启发,并常以它为例说明神经网络可替代 B-Tree。

方法难度 3 / 5约 60 分钟经得起时间
读之前最好知道
1B-Tree 与哈希索引等数据库索引结构
2神经网络基础
3缓存与内存层次(cache hierarchy)
4数据库系统基础

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
强烈推荐sulam · 2017-12

“I highly recommend people give this paper a read.”

强烈建议大家读这篇论文
Ask HN: What was the best CS paper you read in 2017?原评论 ↗
强烈推荐jacksmith21006 · 2018-05

“A fantastic paper on this from Jeff Dean at Google I suggest you read.”

极力推荐 Google 的 Jeff Dean 这篇论文
Google announces a new generation for its TPU machine learning hardware原评论 ↗
强烈推荐KirinDave · 2018-10

“but the paper is an incredibly cool read and very approachable: The Case for Learned Index Structures”

称这篇论文非常酷且通俗易懂,强烈推荐
Ask HN: What's your favorite elegant/beautiful algorithm?原评论 ↗
推荐jacksmith21006 · 2018-02

“Thought one of the more interesting ideas is the Jeff Dean paper”

称 Jeff Dean 这篇论文是更有意思的想法之一
Chip advances require fresh ideas, speakers say原评论 ↗
推荐jacksmith21006 · 2018-02

“Probably my favorite of late is from Jeff Dean”

称 Jeff Dean 这篇是自己近来最喜欢的
Google discloses Microsoft Edge security flaw before a patch is ready原评论 ↗
推荐jacksmith21006 · 2018-04

“But my favorite new thing with these chips is the Jeff Dean paper.”

称这些芯片带来的最爱是 Jeff Dean 这篇论文
The Looming Battle Over AI Chips原评论 ↗
看全部 20 条 →
推荐jacksmith21006 · 2018-05

“My favorite paper right now is”

称这是自己目前最喜欢的论文
Google embraces, extends, and extinguishes原评论 ↗
推荐jacksmith21006 · 2018-05

“Most interesting thing I have seen for for database indexes is this paper.”

称这是自己见过最有意思的数据库索引相关论文
Algorithms Behind Modern Storage Systems原评论 ↗
推荐glenrivard · 2018-10

“Paper is from Jeff Dean and a worth while read, IMO.”

推荐 Jeff Dean 的这篇论文,认为值得一读
BlazingSQL – A GPU SQL Engine for RAPIDS Open-Source Software from Nvidia原评论 ↗
推荐ndr · 2019-08

“You should look at the paper they mention The Case for Learned Index Structures[0].”

建议去看对方提到的这篇学习型索引论文
Machine learning primitives in rustc (2018)原评论 ↗
推荐kevindamm · 2025-06

“The topic of learned indexes is worth investigating”

认为学习型索引这个主题值得研究
The probability of a hash collision (2022)原评论 ↗
引用pg314 · 2017-12

“Under some assumptions it does handle inserts.”

指出在某些假设下学习型索引可以处理插入
Machine Learning for Systems and Systems for Machine Learning [pdf]原评论 ↗
引用visarga · 2017-12

“they replace B-Trees with neural nets:”

提到论文用神经网络取代 B-Tree
DeepZip: Lossless Compression Using Recurrent Networks [pdf]原评论 ↗
引用yorwba · 2018-05

“speed up some classical algorithm based on heuristics by modeling the data with a neural network”

拿它说明用神经网络对数据建模,可加速基于启发式的经典算法
An O(N) Sorting Algorithm: Machine Learning Sorting原评论 ↗
引用malisper · 2020-07

“The idea was you could train a model to predict the location of a row based on a primary key.”

解释其思路:训练模型根据主键预测行的位置
Neural programmer better than Quicksort原评论 ↗
引用seg_lol · 2020-10

“It would be something analogous to the 'learned indexes' paper.”

拿学习型索引论文作类比
Using Lidar to Add Autofocus to a Manual Focus Lens原评论 ↗
引用emayljames · 2021-01

“Our initial results show, that by using neural nets we are able to outperform cache-optimized B-Trees by up to 70% in speed while saving an order-of-magnitude in memory over several real-world data sets.”

引用结果:神经网络索引速度最多比缓存优化的 B-Tree 快 70%,内存省一个数量级
A Recursive Model learned index structure原评论 ↗
引用gopalv · 2021-07

“The learned index[1] is basically a variation of this sort of”

指出学习型索引基本上是这类方法的一种变体
B-tree Path Hints原评论 ↗
引用robertsdionne · 2022-07

“The Case for Learned Index Structures: The Case for Learned Index Structures”

举学习型索引为冷门数据结构的例子
Ask HN: What are some cool but obscure data structures you know about?原评论 ↗
引用hamilyon2 · 2026-01

“When NPU will become available in most systems, we will see learned indexes like in The Case for Learned Index Structures on client side.”

预言 NPU 普及后,客户端会出现这类学习型索引
Dell's CES 2026 chat was the most pleasingly un-AI briefing I've had in 5 years原评论 ↗