基于离散监督哈希算法的非结构化网络数据库在线检索方法
An online retrieval method of an unstructured network database based on a discrete supervised hash algorithm
赵鑫1
作者信息
摘要
受非结构化网络数据关联性弱、数据收敛特性差的影响,对相关数据库进行检索时,检索精度难以得到保障,为此,提出基于离散监督哈希算法的非结构化网络数据库在线检索方法研究.首先,基于离散监督哈希算法将非结构化网络数据样本的显式语义标签利用相应的二进制码进行表示,并借助拉格朗日乘子对其进行离散化处理,使得哈希函数对于学习非结构化网络数据显式语义关联快速收敛.然后,按照相同的方式对所有数据及查询请求进行编码.最后,将与查询请求哈希编码匹配程度最高的数据作为优先检索输出结果.在测试结果中,设计检索方法在图像查询文本指令,以及文本查询图像指令下的检索平均精度均值始终稳定在较高水平,且检索时间最高仅为7s,具有实用性.
Abstract
Due to the weak correlation of unstructured network data and poor data convergence characteristics,it is difficult to guarantee the retrieval accuracy when searching relevant databases.Therefore,the online retrieval method of unstructured network database based on discrete supervised hash algorithm is proposed.First,the explicit semantic label of the unstructured network data sample is represented using the corresponding binary code,and discretized with the help of Lagrange multipliers,so that the hash function can quickly converge for learning the explicit semantic correlation of the unstructured network data.Then,all the data and query requests are encoded in the same way.Finally,the data with the highest match with the query request hash encoding is used as the priority retrieval output.In the test results,the average retrieval accuracy of the design retrieval method under the im-age query text instruction and the text query image instruction is always stable at a high level,and the highest retrieval time is only 7 s,which is practical.
关键词
离散监督哈希算法/非结构化网络数据库/在线检索/SDMH/显式语义标签/二进制码/拉格朗日乘子Key words
discrete supervised hash algorithm/unstructured network database/online search/SDMH/explicit semantic label/binary code/Lagrange multiplier引用本文复制引用
出版年
2024
NETL
NSTL
万方数据