Hypergraph iteration method based on two-hop graph and application of hypergraph iteration method

Abstract

According to the hypergraph iteration method based on the two-hop graph and the application of the hypergraph iteration method. By constructing the two-hop graph, the hyperedges are directly communicated, updating of the hyperedges is completed, and the iteration processing speed of hypergraph tasks is increased. A two-hop graph construction step includes: constructing an edge between the hyperedges containing the common vertex, and storing the degrees and the distributed vertex information of the hyperedges by the hyperedges. A vertex distribution step includes: distributing vertexes contained by only one hyperedge, and then distributing common vertexes of the hyperedges. A public vertex information obtaining step includes: analyzing different hypergraph tasks, determining a message valueformula and a hyperedge value formula suitable for a two-hop graph, and determining the public vertex information to be stored by each hyperedge; combining iterative processing based on a binary hopgraph with a Push and Pull-based message acquisition mechanism. Experiments are carried out on a plurality of data sets and a hypergraph learning algorithm, and experiment results verify the high efficiency and expandability of the method.

Description

technical field [0001] The invention belongs to the field of computer large-scale graph data processing, and in particular relates to a hypergraph iteration method based on a two-hop graph and its application. Background technique [0002] In the setting of machine learning problems, it is usually assumed that the relationship between objects is binary, so it is natural to use a graph model to model such a binary relationship. Each vertex in the graph represents an object, and each An edge represents a relationship between two objects. Such a graph can be directed or undirected, depending on whether the binary relations between objects are symmetric or not. For example, friendship relationships in a community can form an undirected graph. As for directed graphs, a well-known example is the World Wide Web. A hyperlink in a web page can be viewed as a directed edge, because assuming that web page A has a hyperlink to web page B, web page B may not have a hyperlink to web pa...

AI Technical Summary

Problems solved by technology

Existing distributed graph computing systems are inefficient in processing hypergraph tasks, and a small number of existing distributed hypergraph iterative processing systems lack hypergraph data and hypergraph task-aware optimization techniques, and processing efficiency is difficult to meet the requirements of big data analysis. need

Images