2012•Journal of SoftwareOpen access

c-Approximate Nearest Neighbor Query Algorithm Based on Learning for High-Dimensional Data

Pei-Sen YUAN, Chaofeng Sha, Xiao-Ling WANG, Aoying Zhou

Open full text 2 citations

Abstract

针对高维数据近似最近邻查询,在过滤-验证框架下提出了一种基于学习的数据相关的c-近似最近邻查询算法.证明了数据经过随机投影之后,满足语义哈希技术所需的熵最大化准则.把经过随机投影的二进制数据作为数据的类标号,训练一组分类器用来预测查询的类标号.在此基础上,计算查询与数据集中数据对象的海明距离.最后,在过滤后的候选数据集上计算查询的最近邻.与现有方法相比,该方法对空间需求更小,编码长度更短,效率更高.模拟数据集和真实数据集上的实验结果表明,该方法不仅能够提高查询效率,而且方便调控在查询质量和查询处理时间方面的平衡问题.

Open-access reader

About this research paper

What this paper is about

针对高维数据近似最近邻查询,在过滤-验证框架下提出了一种基于学习的数据相关的c-近似最近邻查询算法.证明了数据经过随机投影之后,满足语义哈希技术所需的熵最大化准则.把经过随机投影的二进制数据作为数据的类标号,训练一组分类器用来预测查询的类标号.在此基础上,计算查询与数据集中数据对象的海明距离.最后,在过滤后的候选数据集上计算查询的最近邻.与现有方法相比,该方法对空间需求更小,编码长度更短,效率更高.模拟数据集和真实数据集上的实验结果表明,该方法不仅能够提高查询效率,而且方便调控在查询质量和查询处理时间方面的平衡问题.

Why it matters

OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

针对高维数据近似最近邻查询,在过滤-验证框架下提出了一种基于学习的数据相关的c-近似最近邻查询算法.证明了数据经过随机投影之后,满足语义哈希技术所需的熵最大化准则.把经过随机投影的二进制数据作为数据的类标号,训练一组分类器用来预测查询的类标号.在此基础上,计算查询与数据集中数据对象的海明距离.最后,在过滤后的候选数据集上计算查询的最近邻.与现有方法相比,该方法对空间需求更小,编码长度更短,效率更高.模拟数据集和真实数据集上的实验结果表明,该方法不仅能够提高查询效率,而且方便调控在查询质量和查询处理时间方面的平衡问题.

Key concepts: k-nearest neighbors algorithm, Random projection, Computer science, Binary code, Hash function, Binary number, Nearest neighbor search, Hamming distance

Related papers

Back to paper searchBrowse research topicsOriginal source
c-Approximate Nearest Neighbor Query Algorithm Based on Learning for High-Dimensional Data — Research Paper | ScholarLens