CN113297500A - Social network isolated node link prediction method - Google Patents
Social network isolated node link prediction method Download PDFInfo
- Publication number
- CN113297500A CN113297500A CN202110697470.8A CN202110697470A CN113297500A CN 113297500 A CN113297500 A CN 113297500A CN 202110697470 A CN202110697470 A CN 202110697470A CN 113297500 A CN113297500 A CN 113297500A
- Authority
- CN
- China
- Prior art keywords
- vector
- network
- node
- semi
- link
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 89
- 238000013507 mapping Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000010801 machine learning Methods 0.000 claims description 7
- PRQROPMIIGLWRP-UHFFFAOYSA-N N-formyl-methionyl-leucyl-phenylalanin Chemical compound CSCCC(NC=O)C(=O)NC(CC(C)C)C(=O)NC(C(O)=O)CC1=CC=CC=C1 PRQROPMIIGLWRP-UHFFFAOYSA-N 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 abstract description 9
- 230000003042 antagnostic effect Effects 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000012545 processing Methods 0.000 abstract description 2
- 238000011161 development Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012706 support-vector machine Methods 0.000 description 3
- 238000007476 Maximum Likelihood Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 239000012466 permeate Substances 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- 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
-
- 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Neurology (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of social network link prediction, and particularly relates to a social network isolated node link prediction method. The invention provides a social network isolated node link prediction method for semi-supervised link prediction by adopting auxiliary information aiming at isolated node prediction tasks in social network link prediction. The method learns a mapping model through the attribute vector and the topological vector of the nodes in the known network; mapping the attribute vector of the node to be predicted into a topological vector by using the model; and generating an antagonistic network based on semi-supervision, completing a link prediction task by using the topology vector of the node to be predicted and the topology vector of the current network node, and verifying the feasibility and the advantage of the model on a social data set. The method can be used for processing the prediction problem of the isolated node in the social network link prediction process, and can get rid of the dependence on the sample containing the label to a certain extent in the prediction process.
Description
Technical Field
The invention belongs to the technical field of social network link prediction, and particularly relates to a social network isolated node link prediction method.
Background
With the progress of computer software and hardware level and the continuous development of network and communication means, many excellent social network applications are born and gradually permeate into the aspects of people's lives. As the number of users increases geometrically over the years, these platforms accumulate vast amounts of data that can be analyzed to yield information that is useful for production and life. In foreign countries, people often use Twitter and Facebook to chat with friends, sharing their own living states; and meanwhile, LinkedIn is used for carrying out activities such as extension of position, exchange of technical experience, acquisition of company information and the like. In China, people generally browse various hot news by using microblogs, perform activities such as praise, forward and post comments, and can also focus on accounts of other people to acquire the latest dynamics of the focused accounts. Meanwhile, the QQ and the WeChat provide the functions of real-time information exchange and text content browsing among friends for people. Social networks provide a number of application services including friend recommendations, product recommendations, knowledge network construction, etc., and the core process for implementing such services is to accurately mine relationships between various entities in the social network. This process may be referred to as social network entity link prediction, hereinafter simply link prediction.
One class of methods for link prediction is based on maximum likelihood estimation. Newman, Clauset and Moore consider links as a reflection of the internal hierarchy, and propose a maximum likelihood estimation algorithm to predict links based on the assumption. The method is suitable for networks with obvious hierarchical organization, such as family relation networks or food chain networks in local environments. Another type of method is a predictive method that utilizes node attributes. Heaukulani et al classify by directly extracting feature information related to nodes as input to a random forest. O' Madadhain et al think that the existing similarity measure is not completely applicable to all networks, and therefore propose to use a measure learning method to generate a measure between node attributes, and further predict the possibility of links existing between nodes.
In addition, some scholars propose link prediction methods based on node similarity. Link prediction methods based on node similarity are generally based on the following assumptions: the higher the similarity of two nodes, the greater the likelihood of a link existing between the two nodes. The core idea of the method is to judge whether the point pair is a link or not based on a threshold value of an index specifically describing the similarity of the point pair.
Therefore, many scholars propose different similarity indexes to measure the similarity of nodes, and common factors include common neighbors, cosine similarity, Jaccard coefficients and the like. The earliest used similarity evaluation index is a common neighbor, and the common neighbor means that if two nodes have more common neighbors, the two nodes can be considered to have stronger similarity. Many link prediction related studies are improved based on a common neighbor similarity index. Jaccard proposes that the Jaccard coefficient be used in link prediction. In recent years, various versions of link prediction similarity indexes have been studied and successfully applied to link prediction of various networks.
With the development of machine learning, some learners also propose a link prediction algorithm based on machine learning. Al Hasan M and the like apply a Support Vector Machine (SVM) model of machine learning to link prediction, and experimental verification is carried out on DBLP, BIOBASE and two data sets, so that the effectiveness of the SVM model relative to other machine learning algorithms is proved. Hasan et al use a plurality of different classification algorithms to perform a link prediction experiment on a real data set, most of which achieve a better experiment effect, and also prove the feasibility of adopting a link prediction algorithm for supervised learning.
However, in the development of the above-mentioned link prediction algorithm, researchers will mostly focus on the problem of performing link prediction on a complete network. But the demand in the real world can sometimes abstract the task of predicting the possible relationships of isolated nodes to the current network. The method has important significance for improving the service quality of the social network and has a strong practical application background. The biggest difficulty of such tasks is how to obtain topology information and other auxiliary information of the node to be predicted. The topology information of the node refers to the link relation between the node and other nodes in the network; the auxiliary information includes, for example, a User profile (User profile), text, and the like of the node. Since most social networks require users to register first when they first use, the social network platform can obtain the auxiliary information, such as user portrayal, before these newly joined users do not establish a connection with other users. Therefore, according to the behavior pattern of the new user, the link relation which may exist in the future of the user can be predicted by utilizing the initial auxiliary information.
Besides the problem that the link prediction task may encounter the isolated node prediction task in practical application, the supervised learning-based social network link prediction model relies on the label quality of the training data set, which brings high cost to the algorithm. And the semi-supervised model can efficiently finish model training in a data set using a small number of labels, so that the model adopting semi-supervised learning can be better suitable for a link prediction task.
Disclosure of Invention
The invention aims to provide a social network isolated node link prediction method.
The purpose of the invention is realized by the following technical scheme: the method comprises the following steps:
step 1: inputting a current network topology structure of the social network, and generating an embedded vector for a node in the current network by adopting a network embedding algorithm;
step 2: collecting attribute vectors of current network nodes, and learning a mapping model formed by a multi-layer perceptron MLP by using the attribute vectors of all the nodes and embedded vectors thereof;
if the dimension of the attribute vector of the user is D and the dimension of the user embedded vector is D, obviously, the number of neurons of the input layer of the multi-layer perceptron is D, and the number of neurons of the output layer is D; let the user attribute vector of input be X ═ X1,x2,x3,...,xD]The multilayer perceptron can be regarded as a function fMLP(x) Then the output of the input vector through the multi-layer perceptron is:
X'=[x'1,x'2,...,x'd]=fMLP(X)
in order to map the attribute vector to the embedded vector well, the model adopts the mean square error as the loss function, and the embedded vector corresponding to the user is set as Y ═ Y1,y2,...,yd]Then the loss function of the model is:
and step 3: collecting attribute vectors of isolated nodes, and learning by using the learned multi-layer perceptron mapping model to obtain potential embedded vectors of the isolated nodes;
and 4, step 4: splicing the embedded vector of the isolated node with the embedded vector of the current node in the network, and partially endowing the label to form a machine learning sample;
and 5: constructing a semi-supervised generation confrontation network;
step 6: learning a semi-supervised link prediction model from a sample by adopting a semi-supervised generation countermeasure network, and performing link prediction by using the semi-supervised link prediction model;
after the embedded vectors of the isolated nodes are obtained, the vectors of two users are spliced to form the input of a semi-supervised generation countermeasure network; let user u1Is [ x ] as an embedding vector1,x2,...,xd](ii) a User u2Is [ y ] as an embedding vector1,y2,...,yd]Then the input of the semi-supervised generation confrontation model is [ x ]1,x2,...,xd,y1,y2,...,yd;L]Wherein L represents a label to represent whether a link relationship exists between two user pairs; the user can be the existing node of the current network or an isolated node, and the input node pair vectors do not all have labels;
due to the existence of 3 classes of prediction results: real data, generated data, and tagged data, thus changing the output shape of the discriminator, changing the output of the scalar quantity into the output of a three-dimensional vector: [ Link, Unlink, Fake ]; wherein Link indicates that a Link exists between users; unlink indicates that no link exists; fake denotes generating samples;
the loss function of the discriminator adopts the sum of the supervision loss and the semi-supervision loss:
LD=Lsupervised+Lunsupervised
the loss function of the generator is:
LG=Ladversarial+Lfeature
Ladversarial=-Ex~g(x)log[1-D(y|x,y=K+1)]
in the semi-supervised link prediction model, a generator and a discriminator all adopt a fully-connected network, in order to prevent the occurrence of the phenomenon of mode collapse and the problem of overfitting, a batch regularization and dropout technology is adopted, and the generator and the discriminator adopt a symmetrical structure.
The invention has the beneficial effects that:
the invention provides a social network isolated node link prediction method for semi-supervised link prediction by adopting auxiliary information aiming at isolated node prediction tasks in social network link prediction. The method learns a mapping model through the attribute vector and the topological vector of the nodes in the known network; mapping the attribute vector of the node to be predicted into a topological vector by using the model; and generating an antagonistic network based on semi-supervision, completing a link prediction task by using the topology vector of the node to be predicted and the topology vector of the current network node, and verifying the feasibility and the advantage of the model on a social data set. The method can be used for processing the prediction problem of the isolated node in the social network link prediction process, and can get rid of the dependence on the sample containing the label to a certain extent in the prediction process.
Drawings
FIG. 1 is a diagram of a mapping model of a multi-layered perceptron.
Fig. 2 is a model diagram of a semi-supervised generation countermeasure network.
Fig. 3 is an overall structural diagram of a model of semi-supervised isolated node prediction.
Fig. 4 is a setting table for generating a countermeasure network.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The invention relates to the field of social network link prediction, in particular to a prediction method of isolated nodes in a social network. The invention aims to solve the prediction problem of isolated nodes in the link prediction process of a social network and get rid of the dependence on a sample containing a label to a certain extent in the prediction process.
The purpose of the invention is realized as follows:
step 1, inputting a current network topology structure of a social network, and generating an embedded vector for a node in the current network by adopting a certain network embedding algorithm;
step 2, collecting the attribute vectors of the current network nodes, and learning a mapping model formed by a multilayer perceptron (MLP) by using the attribute vectors of all the nodes and the embedded vectors thereof;
step 4, splicing the embedded vector of the isolated node with the embedded vector of the current node in the network, and partially endowing the labels to form a machine learning sample;
step 5, constructing a semi-supervised generation confrontation network;
and 6, learning from the sample by adopting the generated countermeasure network to obtain a semi-supervised link prediction model, and performing link prediction by using the semi-supervised link prediction model.
The invention provides a semi-supervised link prediction model by adopting auxiliary information aiming at an isolated node prediction task in social network link prediction. Firstly, a mapping model is learned through the attribute vectors and the topology vectors of the nodes in the known network, and the attribute vectors of the nodes to be predicted are mapped into the topology vectors by utilizing the model. And generating an antagonistic network based on semi-supervision, completing a link prediction task by using the topology vector of the node to be predicted and the topology vector of the current network node, and verifying the feasibility and the advantage of the model on a social data set.
Example 1:
1. mapping of node attribute vectors to embedded vectors
Denote as u the nodes already existing in the network1~uN1It has both attribute vector and network structure information, so in the invention, the network embedding algorithm is adopted as u first1~uN1Each node generates an embedded vector; then, a mapping is learned through the embedded vectors and the attribute vectors of the nodes, the mapping can well convert the attribute vectors of the nodes into corresponding embedded vectors, and a Multi-Layer Perceptron (MLP) is selected as a mapping model of the invention, as shown in fig. 1.
If the dimension of the attribute vector of the user is D and the dimension of the user embedded vector is D, obviously, the number of neurons of the input layer of the multi-layer perceptron is D, and the number of neurons of the output layer is D; let the user attribute vector of input be X ═ X1,x2,x3,...,xD]The multilayer perceptron can be regarded as a functionfMLP(x) Then the output of the input vector through the multi-layer perceptron is:
X'=[x'1,x'2,...,x'd]=fMLP(X) (1)
in order to map the attribute vector to the embedded vector well by the model, the invention adopts the mean square error as the loss function, and the embedded vector of the corresponding user is set as Y-Y1,y2,...,yd]Then the loss function of the model is:
2. semi-supervised link prediction method based on generation of countermeasure network
After the embedded vectors of the isolated nodes are obtained, the vectors of two users are spliced to form the input of a semi-supervised generation countermeasure network. Let user u1Is [ x ] as an embedding vector1,x2,...,xd](ii) a User u2Is [ y ] as an embedding vector1,y2,...,yd]Then the input of the semi-supervised generation countermeasure model in the present invention is [ x ]1,x2,...,xd,y1,y2,...,yd;L]And L represents a label to indicate whether a link relation exists between two user pairs. The user can be an existing node of the current network or an isolated node, and the input node has no label on the vector.
After the generation of the confrontation network model is born, students try to complete a semi-supervised learning task by generating the confrontation network through different technical angles. The invention uses related ideas for reference, and enables the generation of a countermeasure network to process data with classification labels by increasing the dimensionality of the output result of a discriminator:
in a conventional generative confrontation network, the discriminator D outputs a single scalar representing the magnitude of the probability that the input belongs to a real sample. In the invention, 3 types of prediction results exist: therefore, the invention changes the output shape of the discriminator and changes the output of the single scalar into the output of a three-dimensional vector: [ Link, Unlink, Fake ], where Link indicates the existence of links between users, Unlink indicates the absence of links, and Fake indicates the loss function of the discriminator for generating samples can be the sum of the supervised loss and the semi-supervised loss:
for the loss of the generator, the model of the invention adopts the sum of the loss of the classical generated sample and the loss of the feature matching. The feature matching loss is a difference between the distribution of the data generated by the generator after feature extraction by the discriminator and the distribution of the real data, and is represented as follows:
thus, the loss function of the generator can be expressed as follows:
semi-supervised generation of a link prediction model against a network is shown in fig. 2.
In order to prevent the occurrence of the phenomenon of mode collapse and the over-fitting problem, batch regularization (batch _ normalization) and dropout techniques are adopted. The generator and the arbiter use a symmetrical structure, the detailed setup of which is shown in fig. 4.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (1)
1. A social network isolated node link prediction method is characterized by comprising the following steps:
step 1: inputting a current network topology structure of the social network, and generating an embedded vector for a node in the current network by adopting a network embedding algorithm;
step 2: collecting attribute vectors of current network nodes, and learning a mapping model formed by a multi-layer perceptron MLP by using the attribute vectors of all the nodes and embedded vectors thereof;
if the dimension of the attribute vector of the user is D and the dimension of the user embedded vector is D, obviously, the number of neurons of the input layer of the multi-layer perceptron is D, and the number of neurons of the output layer is D; let the user attribute vector of input be X ═ X1,x2,x3,...,xD]The multilayer perceptron can be regarded as a function fMLP(x) Then the output of the input vector through the multi-layer perceptron is:
X'=[x'1,x'2,...,x'd]=fMLP(X)
in order to map the attribute vector to the embedded vector well, the model adopts the mean square error as the loss function, and the embedded vector corresponding to the user is set as Y ═ Y1,y2,...,yd]Then the loss function of the model is:
and step 3: collecting attribute vectors of isolated nodes, and learning by using the learned multi-layer perceptron mapping model to obtain potential embedded vectors of the isolated nodes;
and 4, step 4: splicing the embedded vector of the isolated node with the embedded vector of the current node in the network, and partially endowing the label to form a machine learning sample;
and 5: constructing a semi-supervised generation confrontation network;
step 6: learning a semi-supervised link prediction model from a sample by adopting a semi-supervised generation countermeasure network, and performing link prediction by using the semi-supervised link prediction model;
after the embedded vectors of the isolated nodes are obtained, the vectors of two users are spliced to form the input of a semi-supervised generation countermeasure network; let user u1Is [ x ] as an embedding vector1,x2,...,xd](ii) a User u2Is [ y ] as an embedding vector1,y2,...,yd]Then the input of the semi-supervised generation confrontation model is [ x ]1,x2,...,xd,y1,y2,...,yd;L]Wherein L represents a label to represent whether a link relationship exists between two user pairs; the user can be the existing node of the current network or an isolated node, and the input node pair vectors do not all have labels;
due to the existence of 3 classes of prediction results: real data, generated data, and tagged data, thus changing the output shape of the discriminator, changing the output of the scalar quantity into the output of a three-dimensional vector: [ Link, Unlink, Fake ]; wherein Link indicates that a Link exists between users; unlink indicates that no link exists; fake denotes generating samples;
the loss function of the discriminator adopts the sum of the supervision loss and the semi-supervision loss:
LD=Lsupervised+Lunsupervised
the loss function of the generator is:
LG=Ladversarial+Lfeature
Ladversarial=-Ex~g(x)log[1-D(y|x,y=K+1)]
in the semi-supervised link prediction model, a generator and a discriminator all adopt a fully-connected network, in order to prevent the occurrence of the phenomenon of mode collapse and the problem of overfitting, a batch regularization and dropout technology is adopted, and the generator and the discriminator adopt a symmetrical structure.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110697470.8A CN113297500B (en) | 2021-06-23 | 2021-06-23 | Social network isolated node link prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110697470.8A CN113297500B (en) | 2021-06-23 | 2021-06-23 | Social network isolated node link prediction method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113297500A true CN113297500A (en) | 2021-08-24 |
CN113297500B CN113297500B (en) | 2023-07-25 |
Family
ID=77329298
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110697470.8A Active CN113297500B (en) | 2021-06-23 | 2021-06-23 | Social network isolated node link prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113297500B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117349950A (en) * | 2023-12-05 | 2024-01-05 | 华设检测科技有限公司 | Assembly type roadbed optimization method based on generated countermeasure type network and random forest unit |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104134159A (en) * | 2014-08-04 | 2014-11-05 | 中国科学院软件研究所 | Method for predicting maximum information spreading range on basis of random model |
CN107784124A (en) * | 2017-11-23 | 2018-03-09 | 重庆邮电大学 | A kind of LBSN super-networks link Forecasting Methodology based on time-space relationship |
CN108449209A (en) * | 2018-03-17 | 2018-08-24 | 北京工业大学 | The social networks friend recommendation method merged based on routing information and nodal information |
CN109543114A (en) * | 2018-11-14 | 2019-03-29 | 南京邮电大学 | Heterogeneous Information network linking prediction technique, readable storage medium storing program for executing and terminal |
CN110677284A (en) * | 2019-09-24 | 2020-01-10 | 北京工商大学 | Heterogeneous network link prediction method based on meta path |
CN111475739A (en) * | 2020-05-22 | 2020-07-31 | 哈尔滨工程大学 | Heterogeneous social network user anchor link identification method based on meta-path |
CN111611785A (en) * | 2020-04-30 | 2020-09-01 | 北京理工大学 | Generation type confrontation network embedded representation learning method |
CN112767186A (en) * | 2021-01-26 | 2021-05-07 | 东南大学 | Social network link prediction method based on 7-subgraph topological structure |
WO2021096009A1 (en) * | 2019-11-15 | 2021-05-20 | 숭실대학교산학협력단 | Method and device for supplementing knowledge on basis of relation network |
CN113011282A (en) * | 2021-02-26 | 2021-06-22 | 腾讯科技(深圳)有限公司 | Graph data processing method and device, electronic equipment and computer storage medium |
-
2021
- 2021-06-23 CN CN202110697470.8A patent/CN113297500B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104134159A (en) * | 2014-08-04 | 2014-11-05 | 中国科学院软件研究所 | Method for predicting maximum information spreading range on basis of random model |
CN107784124A (en) * | 2017-11-23 | 2018-03-09 | 重庆邮电大学 | A kind of LBSN super-networks link Forecasting Methodology based on time-space relationship |
CN108449209A (en) * | 2018-03-17 | 2018-08-24 | 北京工业大学 | The social networks friend recommendation method merged based on routing information and nodal information |
CN109543114A (en) * | 2018-11-14 | 2019-03-29 | 南京邮电大学 | Heterogeneous Information network linking prediction technique, readable storage medium storing program for executing and terminal |
CN110677284A (en) * | 2019-09-24 | 2020-01-10 | 北京工商大学 | Heterogeneous network link prediction method based on meta path |
WO2021096009A1 (en) * | 2019-11-15 | 2021-05-20 | 숭실대학교산학협력단 | Method and device for supplementing knowledge on basis of relation network |
CN111611785A (en) * | 2020-04-30 | 2020-09-01 | 北京理工大学 | Generation type confrontation network embedded representation learning method |
CN111475739A (en) * | 2020-05-22 | 2020-07-31 | 哈尔滨工程大学 | Heterogeneous social network user anchor link identification method based on meta-path |
CN112767186A (en) * | 2021-01-26 | 2021-05-07 | 东南大学 | Social network link prediction method based on 7-subgraph topological structure |
CN113011282A (en) * | 2021-02-26 | 2021-06-22 | 腾讯科技(深圳)有限公司 | Graph data processing method and device, electronic equipment and computer storage medium |
Non-Patent Citations (4)
Title |
---|
牟卫华 等: "采用向量内积的并行相关算法", 《国防科技大学学报》 * |
王慧 等: "从信息学的角度分析复杂网络链路预测", 《小型微型计算机***》 * |
王智强: "社交网络链接预测方法研究", 《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》 * |
王飞程 等: "多层网络链路预测研究进展", 《计算机应用与软件》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117349950A (en) * | 2023-12-05 | 2024-01-05 | 华设检测科技有限公司 | Assembly type roadbed optimization method based on generated countermeasure type network and random forest unit |
CN117349950B (en) * | 2023-12-05 | 2024-02-02 | 华设检测科技有限公司 | Assembly type roadbed optimization method based on generated countermeasure type network and random forest unit |
Also Published As
Publication number | Publication date |
---|---|
CN113297500B (en) | 2023-07-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113095439A (en) | Heterogeneous graph embedding learning method based on attention mechanism | |
CN102629275B (en) | Face and name aligning method and system facing to cross media news retrieval | |
CN113255895B (en) | Structure diagram alignment method and multi-diagram joint data mining method based on diagram neural network representation learning | |
CN113961759B (en) | Abnormality detection method based on attribute map representation learning | |
CN112990295A (en) | Semi-supervised graph representation learning method and device based on migration learning and deep learning fusion | |
Xiao et al. | Link prediction based on feature representation and fusion | |
Xu et al. | A scientific research topic trend prediction model based on multi‐LSTM and graph convolutional network | |
CN113761250A (en) | Model training method, merchant classification method and device | |
CN113554100B (en) | Web service classification method for enhancing attention network of special composition picture | |
Li et al. | Residual attention graph convolutional network for web services classification | |
CN115686868A (en) | Cross-node-oriented multi-mode retrieval method based on federated hash learning | |
Zhang et al. | An intrusion detection method based on stacked sparse autoencoder and improved gaussian mixture model | |
Geng et al. | Applying the blockchain-based deep reinforcement consensus algorithm to the intelligent manufacturing model under internet of things | |
Wang et al. | Meta-learning adaptation network for few-shot link prediction in heterogeneous social networks | |
Song et al. | Spammer detection using graph-level classification model of graph neural network | |
CN114254738A (en) | Double-layer evolvable dynamic graph convolution neural network model construction method and application | |
CN113297500A (en) | Social network isolated node link prediction method | |
Li et al. | Capsule neural tensor networks with multi-aspect information for Few-shot Knowledge Graph Completion | |
Han et al. | A triple-branch neural network for knowledge graph embedding | |
CN116306834A (en) | Link prediction method based on global path perception graph neural network model | |
Chen et al. | A supervised link prediction method for dynamic networks | |
Liu et al. | Semantic Association and Decision‐Making for the Internet of Things Based on Partial Differential Fuzzy Unsupervised Models | |
İş et al. | A Profile Analysis of User Interaction in Social Media Using Deep Learning. | |
Xie et al. | L-BGNN: Layerwise trained bipartite graph neural networks | |
Chen et al. | Sampled in Pairs and Driven by Text: A New Graph Embedding Framework |
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 |