Method and equipment for acquiring minimum cut of directed graphs

Abstract

An embodiment of the invention provides a method and equipment for acquiring minimum cut of directed graphs. The method includes respectively partitioning each directed graph into at least two meeting point sub-graphs and at least two source point sub-graphs according to preset strategies; acquiring minimum cut sets of all the meeting point sub-graphs by means of parallel computation, acquiring equivalent meeting points of the directed graphs by the aid of communication between the meeting point sub-graphs, acquiring minimum cut sets of all the source point sub-graphs by means of parallel computation and acquiring equivalent source points of the directed graphs by the aid of communication between the source point sub-graphs; acquiring the minimum cut of the directed graphs according to all connected edges of the equivalent meeting points of the directed graphs and the equivalent source points of the directed graphs. The method and the equipment for acquiring the minimum cut of the directed graphs have the advantages that the minimum cut of the directed graphs are acquired from node sets in forms of the sub-graphs in relations with one another sequentially, accordingly, communication frequencies and synchronization frequencies among active nodes during parallel computation can be reduced, and the performance can be improved.

Description

technical field [0001] The invention relates to the field of network optimization, in particular to a method and equipment for obtaining the minimum cut of a directed graph. Background technique [0002] The maximum flow minimum cut problem is a classic problem in the field of network optimization, which can be widely used in the application background with the network as the form of expression. The methods proposed for this problem are all based on the parallel computing of the nodes in the network. Active nodes are processed in parallel through the overall synchronous parallel computing model (Bulk Synchronous Parallel Computing Model, BSP model), and each active node is calculated through superstep until there are no active nodes in the entire network, and the calculation ends. [0003] The inventors found that the above methods have the following problems: during the calculation process of each superstep, a large amount of communication and synchronization between active...

AI Technical Summary

Problems solved by technology

[0003] The inventor found that the above method has the following problems: During the calculation process of each superstep, a large amount of communication and synchronization between active nodes are required, which leads to network congestion and extremely low performance

Images