Distributed ADMM machine learning method of adaptive network topology

Abstract

The invention relates to a distributed ADMM machine learning method for adaptive network topology, which belongs to the field of machine learning, and comprises the following steps of: dividing nodes into a management node and a plurality of working nodes, and abstracting the working nodes into upper-layer nodes and lower-layer nodes; a global convex optimization problem is decomposed into a plurality of local convex optimization problems for the connected network and solved, a global optimal solution is obtained by coordinating a local optimal solution, and a machine learning method comprises two parts of node detection and iterative calculation; in a node detection process, a working node runs updating of an iterative calculation part, and in addition, an upper-layer node feeds back completion of single iteration to a management node when each iteration is completed; when the position of an upper node is selected, all traversal possibilities are avoided through a greedy thought, and dynamic selection is adopted, so that the influence of link delay in the network is as small as possible.

Description

technical field [0001] The invention belongs to the technical field of machine learning, in particular to a distributed ADMM machine learning method based on an adaptive network topology. Background technique [0002] In recent years, with the rapid development of the information industry and the continuous expansion of the Internet, big data and machine learning have been used more and more frequently in business. In the field of machine learning, a large amount of high-dimensional data comes from different nodes, which puts forward high requirements for computing power. In this case, it is difficult for a single node to solve such problems, but distributed machine learning algorithms can better adapt to the situation. [0003] Alternating Direction Method of Multipliers (ADMM) is an optimization method for constrained problems widely used in machine learning. It greatly reduces the cost of a single problem by decomposing the global problem into local problems, and the lo...

AI Technical Summary

Problems solved by technology

However, since the communication between nodes requires the use of the network, the overall operation will be affected by the network conditions

If there are inappropriate node interactions in the calculation, the overall calculation process will be greatly slowed down due to the influence of network delay, and if the selected object is not suitable, the data of the corresponding communication group may be polluted, resulting in data convergence The slowdown of speed and the increase of convergence accuracy error

Images