List supervision based hash sorting method

Abstract

The invention relates to a hash sorting method based on list supervision, comprising: calculating the Euclidean distance between a query point and each data point; sorting according to the order of the Euclidean distance from small to large, and obtaining a sorted list of data points in Euclidean space ;Convert the query point and all data points into binary codes through the hash function; divide the binary codes of the query point and data points into sub-blocks of the same length, and have different weights for each sub-block, calculate the query point and each Hamming distance between data points; calculate sorting information according to Hamming distance, and get the sorting list of each data point in Hamming space; calculate all queries according to the sorting list of data points in Euclidean space and Hamming space Point the total loss function; optimize the total loss function to obtain the optimal weight of each sub-block; according to the optimal weight, obtain the optimal hash function. It can effectively sort the data in the Hamming space and improve the accuracy of the query.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a hash sorting method based on list supervision. Background technique [0002] As the most widely used technique in data retrieval, nearest neighbor query has always been a research hotspot. Generally speaking, the nearest neighbor query is to find the data most similar to the query object from the database according to the similarity between the data. It is widely used in information retrieval, pattern recognition, data mining and many other fields. However, with the continuous development of modern society and the updating of information technology, the data generated by people every day has shown explosive growth in terms of both volume and dimension. How to obtain the data you want most through the nearest neighbor query technology is a Problems that need to be solved urgently. [0003] At present, there are mainly two methods to solve the nearest neighbor query problem: one...

AI Technical Summary

Problems solved by technology

This method generally performs well in low-dimensional spaces. However, when the dimension rises sharply, problems such as "curse of dimensionality" will occur, and its search efficiency is close to that of linear search.

The second is the currently popular hash-based search method. The hash method has attracted more and more researchers' attention in recent years because of its better retrieval speed and lower storage cost. Its core idea is to The eigenvectors in the high-dimensional space are converted into low-dimensional binary strings in the Hamming space, and at the same time, the similarity between the original data can be preserved, but the problem is that after the data points are converted into binary codes, when searching, you need Compare the Hamming distance between the data point and the query point, and the Hamming distance is a discrete integer value. There may be multiple points with the same Hamming distance as the query point, so that the most similar data cannot be obtained

[0006] Although the hash method based on list supervision makes full use of the sorting information, when considering that the Hamming distance between all points in the database and the query point is equal, it is impossible to sort each point more accurately, resulting in sorting errors

Images