Complex network node influence maximization method based on depth auto-encoder

A complex network and autoencoder technology, applied in the field of maximizing the influence of complex network nodes based on deep autoencoders, can solve the problem of high time complexity, inability to mine node influence characteristics in complex networks, and unsuitable for large-scale networks and other problems to achieve the effect of avoiding the amount of calculation, reducing the feature dimension, and reducing time consumption

Inactive Publication Date: 2018-09-07
XIDIAN UNIV
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Existing technologies cannot mine the influence characteristics of nodes in complex networks, and there is a certain degree of blindness in the process of sele

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Complex network node influence maximization method based on depth auto-encoder
  • Complex network node influence maximization method based on depth auto-encoder
  • Complex network node influence maximization method based on depth auto-encoder

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0028] Complex networks are abstracted from various complex systems in the real world. Studying complex networks can help understand the deep relationship between individuals in the network, and influence analysis in complex networks plays a vital role in complex networks. Existing technologies cannot mine the influence characteristics of nodes in complex networks, and there is a certain blindness in the process of selecting nodes, and most of them are based on greedy algorithms, which have very high time complexity and are not suitable for large-scale network data processing and complex mining. Influence characteristics of network nodes. The present invention conducts research on this, and proposes a method for maximizing the influence of complex network nodes based on a deep autoencoder. refer to figure 1 , the implementation steps of the present invention are as follows:

[0029] (1) Construct complex network data: abstract the complex network in the real world into a top...

Embodiment 2

[0046] The method of maximizing the influence of complex network nodes based on deep autoencoder is the same as that in Embodiment 1. The estimated value of influence propagation within the scope of two layers is obtained in step 3 of the present invention, and the two-layer influence propagation of each node in the complex network needs to be set. The range equation is used to estimate the two-layer influence propagation range of complex network nodes, and its influence estimate is calculated by the following LIE equation:

[0047]

[0048] in, is the set of neighbor nodes of node v, S represents the set of nodes, k represents the number of nodes, Indicates the neighbor nodes of all nodes in the set S, Indicates the node set of the set S within the spread range of two layers of influence.

[0049] It can be seen that the estimated value of the spread of influence within the scope of the two layers of the present invention is the initial number of activated nodes σ 0 ...

Embodiment 3

[0052] The method for maximizing the influence of complex network nodes based on deep autoencoders is the same as that of Embodiment 1-2, and the construction of the codec model described in step 5 of the present invention specifically includes the following steps:

[0053] (5a) Construct the encoder: use the ReLu function f(x)=max(0,x) as the activation function of the encoder, input the influence approximation matrix M into the encoder equation, and use a single element x i , For example, calculate the deep feature z of node influence, Methods as below:

[0054] z=f(Wx i +b)

[0055] where W and b are the weights and biases in the encoder, respectively.

[0056] (5b) Constructing the decoder: use the ReLu function f(x)=max(0,x) as the activation function of the decoder, input the node influence deep feature z obtained in 5a) to the decoder, and obtain the heavy The element x′ in the influence approximation matrix of the structure i :

[0057] x' i =f(W'z+b')

[00...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The present invention discloses a complex network node influence maximization method based on a depth auto-encoder. The complex network node influence maximization problem is solved. The method comprises the steps of: constructing complex network data; determining initial influence of each node of a complex network; estimating influence propagation values in two layers of ranges; constructing a complex network influence approximate matrix; constructing a coding/decoding model; training the coding/decoding model, optimizing a target function of the coding/decoding model, and obtaining influencefeatures of each node of the complex network; selecting the influence node set of the complex network. Different from other technologies avariciously selecting nodes, the complex network node influence maximization method selects a depth auto-encoder to effectively execute the deep influence features of the complex network nodes and find a node set with potential influence through adoption of anunsupervised method, and the selected nodes can cause large-scale influence propagation. The complex network node influence maximization method based on a depth auto-encoder can be applied to variouscomplex networks in a real world.

Description

technical field [0001] The invention belongs to the technical field of complex networks, and relates to a method for selecting the most influential nodes in a network, in particular to a method for maximizing the influence of nodes in a complex network based on a deep autoencoder, which is used to discover important nodes in a complex network. Background technique [0002] Individuals in the natural world and their interrelationships can be abstracted into a complex network. The nodes in the network can represent individuals in the natural world, and the edges of the network represent the relationships between individuals. Complex networks permeate into various fields, and in-depth research on complex networks can reveal common laws hidden in nature, biology and human society. However, the connection relationship of nodes in complex networks is very complex. The study of nodes with important influence plays a vital role in information dissemination and diffusion. Discovering...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
IPC IPC(8): G06N3/04G06N3/08
CPCG06N3/08G06N3/048
Inventor 公茂果王善峰张天乐毛贻顺杨月磊牛旭东武越
Owner XIDIAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap