一种哈希检索方法

• Format: Patent
• Language: Chinese
• Published: 01.04.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

本发明公开了一种哈希检索方法，特点是通过引入Spark分布式集群，将原始数据和查询数据平均分配至Spark分布式集群的每个节点中进行并行计算，其中包括了排序计算和根据损失函数通过梯度下降法对其中的参数进行迭代更新，最终根据迭代完成的最终二进制编码矩阵分别对原始数据集和查询数据集进行哈希编码，通过比较海明距离得到每个查询数据的查询结果，完成哈希检索过程，优点是降低了整体训练过程所需时间，通过实验结果表明，与其他哈希检索方法相比，采用本发明所述哈希检索方法能够大幅度降低算法的训练时间，从而提高训练效率。 The invention discloses a hash retrieval method. The method is characterized in that a Spark distributed cluster is introduced, the original data and the query data are averagely distributed to each node of the Spark distributed cluster for parallel computing; the parallel computing comprises sorting calculation and iterative updating of parameters in a loss function through a gradient descent method; finally, hash coding is respectively performed on the original data set and the query data set according to the final binary coding matrix after iteration is completed, and the hash retrieval process is completed by comparing the Hamming distance to get the query result of each query data. The Hash retrieval method has the advantages that the time required by the