Target data searching method and device, computer equipment and storage medium

Abstract

The embodiment of the invention discloses a target data searching method and device, computer equipment and a storage medium, and the method comprises the steps: constructing a multi-layer graph structure of a data feature library, the number of nodes in each layer in the multi-layer graph structure is in exponential decay probability distribution from a bottom layer to a top layer, and the bottom layer comprises all nodes corresponding to the data feature library; randomly selecting a node from a top layer as an entry node of a query target node; from the entry node of the top layer, querying a node, closest to the target node, of each layer as a query entry node of the next layer until the bottom layer; and querying a neighbor node close to the target node by using the query entry node of the upper layer in the bottom layer. According to the method, the node which is queried in the upper layer and is closest to the target node is used as the iteration starting point of the lower layer until the bottom layer, so that a large part of nodes which are far away from the target node can be filtered out in each layer during query, the query and operation time is saved, and the query efficiency is improved.

Description

technical field [0001] The present invention relates to data processing technology, more specifically to a target data search method, device, computer equipment and storage medium. Background technique [0002] In the era of surging data volume, it has always been a difficult problem to do high-efficiency and high-precision similarity search in a large amount of data, and how to manage these data is also facing many problems. [0003] Currently, there are several ways to search for data at the level of 100,000 or greater: [0004] The simplest way is to store the features sequentially. This kind of construction algorithm has low time complexity and does not need to occupy additional storage space. The problem is that the search efficiency is low. Each query needs to traverse all the data, and the efficiency drops sharply in scenarios with frequent queries and large amounts of data. [0005] The construction speed and search efficiency of the nearest neighbor retrieval algo...

AI Technical Summary

Problems solved by technology

[0007] The nearest neighbor retrieval technology based on graph theory is a popular method in recent years. It solves the problem of fast retrieval of high-dimensional data by sacrificing part of the accuracy in exchange for retrieval speed, but there is still the problem of high cost of building graphs.

Images