CN113469261B - Source identification method and system based on infection map convolution network - Google Patents
Source identification method and system based on infection map convolution network Download PDFInfo
- Publication number
- CN113469261B CN113469261B CN202110786345.4A CN202110786345A CN113469261B CN 113469261 B CN113469261 B CN 113469261B CN 202110786345 A CN202110786345 A CN 202110786345A CN 113469261 B CN113469261 B CN 113469261B
- Authority
- CN
- China
- Prior art keywords
- node
- graph
- nodes
- feature
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a source identification method and a source identification system based on an infective diagram convolutional network, which relate to the technical field of network exploration type search and comprise the following steps: step S1: inputting a Laplace matrix subjected to symmetrical normalization and a characteristic vector V of each node; step S2: based on the feature optimization layer of the graph neural network, iteratively updating the graph neural network based on vectorized feature input, and optimizing a feature vector V; step S3: selecting and distributing different weights according to different types of nodes based on a plurality of IGCN network layers to perform feature optimization, and updating a feature vector V; step S4: inputting the updated feature vector V into a feedforward neural network, and outputting the classification probability obtained by learning; step S5: and defining the source identification problem as a graph classification problem, performing back propagation by using a cross entropy loss function, and learning a feature vector V of an input node. The method can improve the accuracy of source prediction under the condition of model independence.
Description
Technical Field
The invention relates to the technical field of network exploration type search, in particular to a source identification method and system based on an infective diagram convolution network.
Background
Due to the explosion of Facebook, Twitter, WeChat and other social applications, rapidly expanding social networks are becoming more attractive and a vast cyber-society is forming. Meanwhile, due to the rapidity and convenience of the social network, the social network can be easily utilized to promote the spread of malicious information such as false information and malicious software. Such rumors or error messages are widely spread in social networks, and may have a huge negative impact on society.
The Chinese invention patent with the publication number of CN112699312A discloses a community detection method and a system based on a generation confrontation network and membership graph model, which relate to the technical field of text network exploration type search and comprise the following steps: step 1: generating a to-be-detected graph based on the original data; step 2: generating a membership graph model based on the to-be-detected graph to initialize and generate a countermeasure network; and step 3: optimizing and generating a countermeasure network by using a strategy gradient method based on a membership graph model; and 4, step 4: and predicting member nodes and relationships among members of each community according to the optimized generation countermeasure network. The method can combine community detection and graph characterization learning in a unified learning framework, and greatly improves the model performance.
In the prior art, a given node v with an infectioniIdentifying rumor sources is a challenging but valuable research for detecting and controlling rumor propagation in social networks. Another application of source identification is to find the source of an epidemic, namely Patient Zero (Patient Zero, P0), which helps to gain insight into the spread of infection and to efficiently allocate resources. However, it appears that heuristic methods are only applicable to a variety of rumor sources. Without knowing the underlying propagation model, there is a lack of a rigorous and efficient approach to the single-source problem, and for this reason, there is a need to propose a solution to this problem. In recent years, a new deep learning model GNN has emerged in machine learning and data mining, which is well suited for graph data, which is the basis of the source identification problem. However, due to the uniqueness of the source node, the single source problem is not a simple node classification problem and thus cannot be solved using this idea.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a source identification method and a source identification system based on an infection graph convolutional network.
According to the source identification method and system based on the infectious map convolutional network provided by the invention, the scheme is as follows:
in a first aspect, a method for source identification based on an infection graph convolution network is provided, where the method includes:
step S1: inputting a Laplace matrix subjected to symmetrical normalization and a characteristic vector V of each node, and inputting the input and the characteristic vector V into a model as input parameters by the obtained Laplace matrix, wherein the Laplace matrix comprises state information of a network, and the characteristic vector V is the state information of each node;
step S2: based on the feature optimization layer of the graph neural network, iteratively updating the graph neural network based on vectorized feature input, and optimizing a feature vector V;
step S3: selecting and distributing different weights according to different types of nodes based on a plurality of IGCN network layers to perform feature optimization, and updating a feature vector V;
step S4: inputting the updated feature vector V into a feedforward neural network, outputting classification probability obtained by learning, and finally determining whether the node is an infected source node according to the classification probability so as to make accurate judgment;
step S5: and defining the source identification problem as a graph classification problem, performing back propagation by using a cross entropy loss function, and learning a feature vector V of an input node.
Preferably, the step S1 includes: infection status, infected neighbors, uninfected neighbors, cross-center and tightness-center intra-infection map structures, and normalize each feature with a normalization layer.
Preferably, the update formula of the neural network in step S2 is as follows:
wherein A ∈ R|V|"V" is the adjacency matrix of the target graph, denoted by HlA feature vector representing the l-th node of the GCN;
Hl+1a GCN model representing layer l + 1; wlIs the trainable weight of the GCN level, and sigma represents the nonlinear activation level;representing an adjacency matrix;a representation degree matrix; i denotes the value of the identity matrix.
Preferably, the step S3 includes:
step S3.1: there are four ways to update the feature vector from infected nodes and uninfected nodes: propagation between uninfected and uninfected nodes; propagation from infected node to uninfected node; the propagation from uninfected node to infected node and between infected nodes is based on four ways: assigning different weights;
step S3.2: the contribution of different neighboring nodes is controlled using a plurality of trainable parameters while updating the feature vector of the target node.
Preferably, the step S3.1 of assigning different weights includes:
A′=I+μ1A1+μ2A2+μ3A3+μ4A4
wherein:
A1=(pij)∈R|V|×|V|,
A2=(qij)∈R|V|×|V|,
A3=(sij)∈R|V|×|V|,
A4=(tij)∈R|V|×|V|
wherein A' represents the updated node weight distribution, μ1,μ2,μ3,μ4Respectively representing different weights assigned to each different node; i represents a node with a node number i; j represents a node with the node number j; v represents the number of nodes in the graph; r represents the code number of the infection map.
More specifically:
preferably, the feature update formula in the IGCN layer in step S3.2 is:
Hl+1=σ(A′·Hl)。
preferably, in step S4, the dimension of the feedforward neural network is | V | × N, | V | is the number of nodes, N is the size of the hidden state, and the SOFTMAX classifier is used for classification.
Preferably, the node with the highest classification probability in step S4 is a predicted value, and is represented as:
O=FC(Y)
where Y is the output of the IGCN layer, o is the output of the feedforward neural network,is the probability vector of the node that is the source.
Preferably, the step S5 adopts cross entropy loss as a loss function, which is defined as follows:
where | V | is the number of nodes, and y ═ y1,y2,…,y(v)) Is trueThe value of the unique non-zero element at the ith position is 1 if the id of the corresponding rumor source is i, and y is the estimated value given by softmax; and to reduce overfitting, an L2 regularization method is used in the equation with | | ω | | survival2The coefficient is denoted as λ, where L2 regularization is a general punitive method to prevent function overfitting.
In a second aspect, there is provided a source identification system based on an infection graph convolutional network, the system comprising:
module M1: inputting a Laplace matrix subjected to symmetrical normalization and a characteristic vector V of each node, and inputting the input and the characteristic vector V into a model as input parameters by the obtained Laplace matrix, wherein the Laplace matrix comprises state information of a network, and the characteristic vector V is the state information of each node;
module M2: based on the feature optimization layer of the graph neural network, iteratively updating the graph neural network based on vectorized feature input, and optimizing a feature vector v;
module M3: selecting and distributing different weights according to different types of nodes based on a plurality of IGCN network layers to perform feature optimization, and updating a feature vector v;
module M4: inputting the updated feature vector v into a feedforward neural network, outputting the classification probability obtained by learning, and finally determining whether the node is an infected source node according to the classification probability so as to make accurate judgment;
module M5: and defining the source identification problem as a graph classification problem, performing back propagation by using a cross entropy loss function, and learning a feature vector v of an input node.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention establishes an IGCN model to adapt to the infected network, thereby improving the prediction accuracy of the source under the condition of no relation of the model;
2. the invention can more accurately narrow the range of the target source.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is an infection model;
FIG. 2 is a propagation model;
FIG. 3 is a structural framework of the IGCN;
FIG. 4 is a feature update process in the GCN layer;
fig. 5 shows four possible updating methods for feature vectors.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will aid those skilled in the art in further understanding the present invention, but are not intended to limit the invention in any manner. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the invention.
An embodiment of the present invention provides a source identification method based on an infection map convolutional network, which is shown in fig. 1 and fig. 2 and includes:
step S1: inputting a Laplace matrix subjected to symmetrical normalization and a characteristic vector V of each node, and inputting the input and the characteristic vector V into a model as input parameters by the obtained Laplace matrix, wherein the Laplace matrix contains state information of a network, and the characteristic vector V is the state information of each node. This step included a 5-point infection map structure: infected state, infected neighbors, uninfected neighbors, cross-centers, and tightness-centers, and normalize each feature with a normalization layer.
Step S2: and based on the feature optimization layer of the graph neural network, iteratively updating the graph neural network based on vectorized feature input, and optimizing the feature vector v. The update formula of the graph neural network is as follows:
wherein A ∈ R|V|Where X is V is the adjacency matrix of the target graph, H1A feature vector representing the 1 st node of the GCN; hl +1A GCN model representing layer l + 1; wlIs the trainable weight of the GCN level, σ denotes the non-linear activation level;representing an adjacency matrix;a representation degree matrix; i denotes the value of the identity matrix.
Step S3: referring to fig. 3 and 4, feature optimization is performed by selectively assigning different weights according to different types of nodes based on multiple IGCN network layers, and the feature vector V is updated. The method also comprises the following steps:
step S3.1: there are four ways to update the feature vectors from infected nodes and uninfected nodes: propagation between uninfected and uninfected nodes; propagation from infected node to uninfected node; propagation from uninfected node to infected node and between infected nodes is based on four approaches: different weights are assigned:
A′=I+μ1A1+μ2A2+μ3A3+μ4A4
wherein:
A1=(pij)∈R|V|×|V|,
A2=(qij)∈R|V|×|V|,
A3=(sij)∈R|V|×|V|,
A4=(tij)∈R|V|×|V|;
wherein A' represents the updated node weight distribution, μ1,μ2,μ3,μ4Respectively representing different weights assigned to each different node; i represents a node with a node number i; j represents a node with the node number j; v represents the number of nodes in the graph; r represents the code number of the infection map.
More specifically:
referring to fig. 5, step S3.2: four trainable parameters are used to control the contribution of different neighbor nodes while updating the feature vector of the target node. The feature update formula in the IGCN layer is:
Hl+1=σ(A′·Hl)。
step S4: inputting the updated feature vector V into a feedforward neural network, outputting the classification probability obtained by learning, and finally determining whether the node is an infected source node according to the classification probability so as to make accurate judgment. The dimensionality of the feedforward neural network is | V | multiplied by N, | V | is the number of nodes, N is the size of a hidden state, and a SOFTMAX classifier is adopted for classification. The node with the highest classification probability is a predicted value, and is expressed as:
O=FC(Y)
whereinY is the output of the IGCN layer, O is the output of the feedforward neural network,is the probability vector of the node that is the source.
Step S5: and defining the source identification problem as a graph classification problem, performing back propagation by using a cross entropy loss function, and learning a feature vector V of an input node.
The cross-entropy loss is taken as a loss function, which is defined as follows:
where | V | is the number of nodes, and y ═ y1,y2,…,y(v)) Is true value, if id of corresponding rumor source is i, the value of unique non-zero element at ith position is 1, y is the estimated value given by softmax; and to reduce overfitting, an L2 regularization method is used in the equation with | | ω | | survival2The coefficient is denoted as λ, where L2 regularization is a general punitive method to prevent function overfitting.
It is well within the knowledge of a person skilled in the art to implement the system and its various devices, modules, units provided by the present invention in a purely computer readable program code means that the same functionality can be implemented by logically programming method steps in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units included in the system for realizing various functions can also be regarded as structures in the hardware component; means, modules, units for performing the various functions may also be regarded as structures within both software modules and hardware components for performing the method.
The foregoing description has described specific embodiments of the present invention. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (9)
1. A source identification method based on an infection graph convolution network is characterized by comprising the following steps:
step S1: inputting a Laplace matrix subjected to symmetrical normalization and a characteristic vector V of each node, and inputting the obtained Laplace matrix and the characteristic vector V into a model together as input parameters, wherein the Laplace matrix comprises state information of a network, and the characteristic vector V is the state information of each node;
step S2: based on the feature optimization layer of the graph neural network, iteratively updating the graph neural network based on vectorized feature input, and optimizing a feature vector V;
step S3: selecting and distributing different weights according to different types of nodes to perform feature optimization based on a plurality of infection graph convolution network layers, and updating a feature vector V, wherein the different types of nodes are as follows: infected nodes and uninfected nodes;
step S4: inputting the updated feature vector V into the feedforward neural network, and outputting the classification probability distribution obtained by the feedforward neural network;
step S5: defining a source identification problem as a graph classification problem, performing back propagation on output probability distribution by using a cross entropy loss function, learning a feature vector V of an input node, and finally determining whether the node is an infected source node according to classification probability so as to make accurate judgment;
the update formula of the graph neural network in step S2 is as follows:
wherein A ∈ R|V|×|V|Is an adjacency matrix of the target graph, using HlA feature vector representing the l-th node of the GCN;
2. The method for source identification based on the infection graph convolution network according to claim 1, wherein the state information of the feature vector V in step S1 includes: infection status, infected neighbors, uninfected neighbors, cross-center and tightness-center intra-infection map structures, and normalize each feature with a normalization layer.
3. The method for source identification based on the infection graph convolutional network of claim 1, wherein the step S3 comprises:
step S3.1: there are four ways to update the feature vectors from infected nodes and uninfected nodes: propagation between uninfected and uninfected nodes; propagation from infected node to uninfected node; propagation from uninfected node to infected node and between infected nodes is based on four approaches: assigning different weights;
step S3.2: the contribution of different neighboring nodes is controlled using a plurality of trainable parameters while updating the feature vector of the target node.
4. The infection graph convolutional network-based source identification method of claim 3, wherein the assigning of different weights in step S3.1 comprises:
A′=I+μ1A1+μ2A2+μ3A3+μ4A4
wherein:
A1=(pij)∈R|V|×|V|,
A2=(qij)∈R|V|×|V|,
A3=(sij)∈R|V|×|V|,
A4=(tij)∈R|V|×|V|;
wherein A' represents the updated node weight distribution, μ1,μ2,μ3,μ4Respectively representing different weights assigned to each different node; i represents a node with a node number i; j represents a node with the node number j; v represents the number of nodes in the graph; r represents a code number which is the infection map;
more specifically:
5. the infection map convolutional network-based source identification method of claim 3, wherein the feature update formula in the infection map convolutional network layer in the step S3.2 is as follows:
Hl+1=σ(A′·Hl)。
6. the method for source identification based on the infectogram convolution network of claim 1, wherein the dimension of the feedforward neural network in step S4 is | V | × N, | V | is the number of nodes, N is the size of the hidden state, and the classification is implemented by using a SOFTMAX classifier.
7. The method for source identification based on the infection graph convolutional network of claim 6, wherein the node with the highest classification probability in step S5 is a predicted value represented as:
O=FC(Y)
8. The method for source identification based on the infection graph convolution network according to claim 1, wherein the step S5 adopts cross entropy loss as a loss function, which is defined as follows:
wherein | V | is nodalNumber, if id of corresponding rumor source is i, the value of unique non-zero element at i-th position is 1, y is the estimated value vector given by softmax function, which is expressed as y ═ y (y ═ y1,y2,...,y(v)) Each value in the vector represents the probability of whether the corresponding node is infected; and in order to reduce overfitting, λ | | ω | | calvities in the equation2Representing the L2 canonical method, using | ω | ceiling2It is shown that the coefficients, with a factor of lambda,is the probability vector of the node as the source; among them, the L2 regularization is a general punitive method to prevent overfitting of the function.
9. A source identification system based on an infection graph convolutional network, comprising:
module M1: inputting a Laplace matrix subjected to symmetrical normalization and a characteristic vector V of each node, and inputting the obtained Laplace matrix and the characteristic vector V into the model together as input parameters, wherein the Laplace matrix comprises state information of a network, and the characteristic vector V is the state information of each node;
module M2: based on the feature optimization layer of the graph neural network, iteratively updating the graph neural network based on vectorized feature input, and optimizing a feature vector V;
module M3: selecting and distributing different weights according to different types of nodes based on a plurality of infection graph convolution network layers to perform feature optimization, and updating a feature vector V, wherein the different types of nodes are as follows: infected nodes and uninfected nodes;
module M4: inputting the updated characteristic vector V into the feedforward neural network, and outputting the class probability distribution of the feedforward neural network;
module M5: defining a source identification problem as a graph classification problem, performing backward propagation on probability distribution obtained by forward propagation by using a cross entropy loss function, learning a feature vector V of an input node, and finally determining whether the node is an infected source node or not according to classification probability so as to make accurate judgment;
the update formula of the neural network of the graph in the module M2 is as follows:
wherein A ∈ R|V|×|V|Is an adjacency matrix of the target graph, denoted by HlA feature vector representing the l-th node of the GCN;
Hl+1a GCN model representing layer l + 1; w is a group oflIs the trainable weight of the GCN level, and sigma represents the nonlinear activation level;representing an adjacency matrix;representing the normalized Laplace matrix;a representation degree matrix; i denotes the value of the identity matrix.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110786345.4A CN113469261B (en) | 2021-07-12 | 2021-07-12 | Source identification method and system based on infection map convolution network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110786345.4A CN113469261B (en) | 2021-07-12 | 2021-07-12 | Source identification method and system based on infection map convolution network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113469261A CN113469261A (en) | 2021-10-01 |
CN113469261B true CN113469261B (en) | 2022-07-15 |
Family
ID=77879868
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110786345.4A Active CN113469261B (en) | 2021-07-12 | 2021-07-12 | Source identification method and system based on infection map convolution network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113469261B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114693464B (en) * | 2022-03-08 | 2023-04-07 | 电子科技大学 | Self-adaptive information propagation source detection method |
CN115049415A (en) * | 2022-07-20 | 2022-09-13 | 北京工商大学 | Social media false news detection method based on community propagation structure |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112199608A (en) * | 2020-11-03 | 2021-01-08 | 北京中科研究院 | Social media rumor detection method based on network information propagation graph modeling |
CN112231562A (en) * | 2020-10-15 | 2021-01-15 | 北京工商大学 | Network rumor identification method and system |
CN112597699A (en) * | 2020-12-14 | 2021-04-02 | 新疆师范大学 | Social network rumor source identification method integrated with objective weighting method |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111160452A (en) * | 2019-12-25 | 2020-05-15 | 北京中科研究院 | Multi-modal network rumor detection method based on pre-training language model |
CN113052263A (en) * | 2021-04-23 | 2021-06-29 | 东南大学 | Small sample image classification method based on manifold learning and high-order graph neural network |
-
2021
- 2021-07-12 CN CN202110786345.4A patent/CN113469261B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112231562A (en) * | 2020-10-15 | 2021-01-15 | 北京工商大学 | Network rumor identification method and system |
CN112199608A (en) * | 2020-11-03 | 2021-01-08 | 北京中科研究院 | Social media rumor detection method based on network information propagation graph modeling |
CN112597699A (en) * | 2020-12-14 | 2021-04-02 | 新疆师范大学 | Social network rumor source identification method integrated with objective weighting method |
Non-Patent Citations (2)
Title |
---|
Multiple Rumor Source Detection with Graph Convolutional Networks;Ming Dong 等;《CIKM"19》;20191107;第569-578页 * |
基于图卷积网络的谣言鉴别研究;米源 等;《CNKI网络首发》;20201022;第1-9页 * |
Also Published As
Publication number | Publication date |
---|---|
CN113469261A (en) | 2021-10-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2022121289A1 (en) | Methods and systems for mining minority-class data samples for training neural network | |
CN113469261B (en) | Source identification method and system based on infection map convolution network | |
Vargas et al. | Spectrum-diverse neuroevolution with unified neural models | |
CN111881342A (en) | Recommendation method based on graph twin network | |
CN112508085A (en) | Social network link prediction method based on perceptual neural network | |
Chen et al. | Inferring fuzzy cognitive map models for gene regulatory networks from gene expression data | |
CN112085161B (en) | Graph neural network method based on random information transmission | |
CN110263236B (en) | Social network user multi-label classification method based on dynamic multi-view learning model | |
Xiong et al. | Multi-feature fusion and selection method for an improved particle swarm optimization | |
Sarswat et al. | A novel two-step approach for overlapping community detection in social networks | |
CN114463540A (en) | Segmenting images using neural networks | |
Barman et al. | A neuro-evolution approach to infer a Boolean network from time-series gene expressions | |
Ding et al. | An online learning neural network ensembles with random weights for regression of sequential data stream | |
Chiu et al. | An evolutionary approach to compact dag neural network optimization | |
Chen et al. | Optimal output tracking of switched Boolean networks | |
Kroos et al. | Neuroevolution for sound event detection in real life audio: A pilot study | |
Liu et al. | Gradient‐Sensitive Optimization for Convolutional Neural Networks | |
CN117113274A (en) | Heterogeneous network data-free fusion method and system based on federal distillation | |
Jiang et al. | ARAe-SOM+ BCO: an enhanced artificial raindrop algorithm using self-organizing map and binomial crossover operator | |
CN116956081A (en) | Heterogeneous social network distribution outward generalization-oriented social label prediction method and system | |
Wang et al. | A modified minmax k-means algorithm based on PSO | |
CN113704570B (en) | Large-scale complex network community detection method based on self-supervision learning type evolution | |
CN106815653B (en) | Distance game-based social network relationship prediction method and system | |
Alruwaili et al. | Red Kite Optimization Algorithm With Average Ensemble Model for Intrusion Detection for Secure IoT | |
CN110059806B (en) | Multi-stage weighted network community structure detection method based on power law function |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |