CN111401514A - Semi-supervised symbol network embedding method and system based on improved graph convolutional network - Google Patents
Semi-supervised symbol network embedding method and system based on improved graph convolutional network Download PDFInfo
- Publication number
- CN111401514A CN111401514A CN202010090482.XA CN202010090482A CN111401514A CN 111401514 A CN111401514 A CN 111401514A CN 202010090482 A CN202010090482 A CN 202010090482A CN 111401514 A CN111401514 A CN 111401514A
- Authority
- CN
- China
- Prior art keywords
- network
- symbol
- matrix
- propagation
- supervised
- 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.)
- Pending
Links
Images
Classifications
-
- 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
-
- 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
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The utility model provides a semi-supervised symbol network embedding method and system based on an improved graph convolutional network, which leads in relational network data to form a directed symbol network; calculating an adjacency matrix of a directed symbol network to obtain a propagation adjacency matrix, and activating the propagation adjacency matrix by using a symbol function to obtain a directed activation propagation adjacency matrix; constructing a symbol Laplace matrix, and applying the symbol Laplace matrix by the graph convolution network to realize the improvement of the graph convolution network; the improved graph convolution network is used for carrying out convolution operation on the input adjacent matrix to obtain embedding results of different degrees, and the prediction problem of network link information can be solved.
Description
Technical Field
The disclosure belongs to the technical field of network data or information expression display, and particularly relates to a semi-supervised symbol network embedding method and system based on an improved graph convolutional network.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the rapid development of social media and the gradual maturity of deep learning technologies, network representation learning becomes the focus of attention in the industry and academia. Network representation requires that original topological structure and semantic information of a network are kept unchanged while low-dimensional potential representation of nodes is learned. For example, in a comment trust network, each user can be represented by a multidimensional vector, and information expression of the user on the network can be quantized, so that a trust sub-network with the user as a starting point is mined, the trust sub-network can be perfected through a certain symbol propagation rule, and a trust network with huge information content is constructed for subsequent use. Thus, learned feature representations are an important basis for various tasks based on graphs. At present, most network embedding task methods adopt a deep learning method, potential node information expressions are mined through a multilayer network, so that the final embedding result is more representative, wherein a Graph Convolution Network (GCN) provides a new research idea for network expression learning.
However, the inventor knows that the current graph convolution network only supports an undirected unsigned network and cannot be directly applied to a directed symbolic network, that is, the original graph convolution network has the excellent property of logarithmic matrix and semi-positive matrix by means of the laplace matrix of the unsigned network, and the Fourier transform is applied to realize the graph convolution operation of a spectral domain. However, the directed sign network does not have the excellent property, and the original graph convolution network cannot learn the negative relation in the sign network, so that the final embedding result is seriously unbalanced, and the potential value cannot be effectively created for the related field. If the negative side connection in the symbol network is ignored, the symbol network is treated as an unsigned network, a satisfactory embedding result cannot be obtained, and the subsequent tasks cannot be carried out: for example, the direction and the symbol of the edge in the directed symbol network cannot be effectively processed, the propagation problem of the symbol in the symbol network cannot be solved, and further the form of performing the spectrum domain convolution in the directed symbol network cannot be realized.
In summary, there is no effective solution to the series of problems existing when the popular graph convolution network method is applied to the symbol network.
Disclosure of Invention
The invention provides a semi-supervised symbol network embedding method and system based on an improved graph convolutional network, which are used for solving the problems. The processing capacity of the graph convolution network on negative symbols and directed symbols is successfully improved, so that each node in the network obtains a unique characteristic vector of the node, and the prediction problem of link information or data in network representation can be effectively solved.
According to some embodiments, the following technical scheme is adopted in the disclosure:
a first object of the present disclosure is to provide a semi-supervised symbol network embedding method based on an improved graph convolutional network, comprising the following steps:
importing relational network data to form a directed symbol network;
calculating an adjacency matrix of a directed symbol network to obtain a propagation adjacency matrix, and activating the propagation adjacency matrix by using a symbol function to obtain a directed activation propagation adjacency matrix;
constructing a symbol Laplace matrix, and applying the symbol Laplace matrix by the graph convolution network to realize the improvement of the graph convolution network;
and carrying out convolution operation on the input adjacent matrix by using the improved graph convolution network to obtain embedding results of different degrees.
According to the technical scheme, the directional activation propagation adjacency matrix is constructed, the sign Laplace matrix is combined, the directional sign network can be effectively applied to the graph convolution network, meanwhile, weight information is eliminated, sign information is reserved, the direction and the sign of the edge in the directional sign network are effectively processed, the problem of directional propagation of the sign is solved, and the problem of prediction of link information or data in network representation can be effectively solved. .
As an alternative embodiment, the directed symbol network internal symbol propagation is based on structure balance theory propagation and obtains a higher-order symbol relationship.
As an alternative embodiment, the directed sign network is represented by a adjacency matrix, where a sign of an edge pointing from one point to another is positive, the corresponding matrix element is 1, when negative, the corresponding matrix element is-1, and when unknown, the corresponding matrix element is 0.
As an alternative embodiment, the propagation adjacency matrix is the sum of an adjacency matrix and a transposed matrix of the adjacency matrix and a unit matrix.
As an alternative embodiment, the symbolic laplacian matrix is constructed by constructing a degree matrix and an adjacency matrix of a graph based on the directional activation propagation adjacency matrix on the basis of the unsigned network laplacian matrix.
As an alternative embodiment, the specific process of applying the symbol laplacian matrix by the graph convolution network includes: carrying out spectrum decomposition on the symbol Laplacian matrix, defining a Fourier forward and inverse transformation rule on a graph by taking a feature vector of the symbol Laplacian matrix as a Fourier transformation base, and realizing the conversion of a symbol network to a frequency domain;
and converting the convolution kernel into a frequency domain to realize convolution operation.
As an alternative implementation mode, the embedded information of the symbolic network is obtained by using the improved graph convolution network, and a corresponding feature vector is obtained for each node in the network.
A second object of the present disclosure is to provide a semi-supervised symbol network embedding system based on an improved graph convolutional network, comprising:
the symbol propagation module is configured to calculate an adjacency matrix of the directed symbol network, and further obtain a propagation adjacency matrix;
a symbolic network processing module configured to activate the propagation adjacency matrix using a symbolic function, resulting in a directed activation propagation adjacency matrix;
the network feature extraction module is configured to construct a symbol Laplace matrix, and the graph convolution network applies the symbol Laplace matrix to improve the graph convolution network;
and the graph convolution module is configured to perform convolution operation on the input adjacent matrix by using the improved graph convolution network to obtain embedding results of different degrees.
A third object of the present disclosure is to provide a computer-readable storage medium, wherein a plurality of instructions are stored, and the instructions are adapted to be loaded by a processor of a terminal device and execute the semi-supervised symbol network embedding method based on the modified graph convolutional network.
A fourth object of the present disclosure is to provide a terminal device, comprising a processor and a computer-readable storage medium, the processor being configured to implement instructions; the computer readable storage medium is used for storing a plurality of instructions which are suitable for being loaded by a processor and executing the semi-supervised symbol network embedding method based on the improved graph convolutional network.
Compared with the prior art, the beneficial effect of this disclosure is:
the method can realize the convolution on the symbol network, expand the directionality of the symbol network, excavate deep information among nodes in the network, carry out state edge connection and obtain higher benefit, and the bottom layer takes a graph convolution algorithm as a core to carry out convolution operation of the symbol network, thereby further improving the reliability and the accuracy of information embedded in the network.
The method is used as a propagation rule of the signed network based on the structure balance theory, and can accurately acquire the high-order sign relation, so that the sign propagation information acquired by the unsigned label is more perfect.
The method and the device calculate and propagate the adjacency matrix based on the adjacency matrix, retain original weight information, screen conflicting 0-order symbol information, remove wrong polarity relation in a symbol network and improve accuracy.
The propagation adjacency matrix is activated by using the sign function to obtain the directed activation propagation adjacency matrix, the weight information is eliminated, the sign information is reserved, the problem of negative edges is solved by combining the new definition of the Laplace matrix, the sign network is successfully applied to the graph convolution method, and the problems that the direction and the sign of the edges in the directed sign network cannot be effectively processed, the propagation problem of the sign in the sign network cannot be solved, and further the form of spectrum domain convolution in the directed sign network cannot be realized in the prior art are solved.
The method has wide technical application fields, such as research on user credibility in a comment trust network, recommendation and avoidance of dislike of users in an E-commerce platform, and research and recommendation strategies on user similarity in various large-flow platforms.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure and are not to limit the disclosure.
FIG. 1 is a four symbol trigonometric relationship in a symbol undirected graph;
FIG. 2 is a network visualization result diagram of a sampling example in example 1;
FIG. 3 is a graph of the predicted results in example 1;
FIG. 4 is a flowchart in example 1;
fig. 5 is a system architecture diagram in embodiment 2.
The specific implementation mode is as follows:
the present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present disclosure. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example 1:
a semi-supervised symbol network embedding method based on an improved graph convolution network is characterized in that a graph convolution algorithm is used as a basis, the graph convolution algorithm is applied to various symbol networks in an expanded mode, embedding information of the symbol networks is obtained by a deep learning method, and a unique feature vector of each node in the network is obtained.
The specific process is shown in fig. 4, and comprises the following steps:
1. importing interactive data among users in a comment website, and constructing a comment symbol network;
in the comment website, the comment of each user can be represented by other users, namely the reaction of one user to the comment of another user has the following two basic conditions: trusting the language of the user and not trusting the language of the user, and based on the method, a basic comment symbol network model can be constructed.
2. Defining a symbol propagation basic rule, wherein the propagation rule of symbols in a symbol network can be defined according to the task type of the symbol network, and the method specifically comprises the following steps:
the basic rules of the internal symbol propagation of the embodiment are defined according to the task requirements. The embodiment applies the balance theory to the graph convolution method, and defines the symbol propagation rule as follows.
For the symbolic network, Heider et al proposes a structure balance theory as a basic theory for symbolic propagation in this embodiment, and some basic rules defined in the theory have been generally applied to a link symbol prediction task in the symbolic network and become the basic theory of the symbolic network. The basic structural balance theory is defined as the structural balance triangle, as shown in FIG. 1.
In fig. 1, an example of four symbol trigonometric relationships in a symbol undirected graph. Wherein (a, b) is a balanced trigonometric relationship and (c, d) is an unbalanced trigonometric relationship. The positive relationship between nodes is represented by '+'; the negative relationship is represented by '-'.
This is the underlying triangular balanced structure, which shows in the social network: friends of friends are friends, enemies of friends are enemies, enemies of enemies are friends, and the triangular structure conforming to the rule is a balanced triangular relationship (a, b), otherwise, the triangular structure is called an unbalanced triangular relationship (c, d). Similarly, the basic rules similar to the propagation of the positive and negative polarity relation can be carved in a trust network and other many symbol networks. Just because of the existence of the structure balance triangle, we can propagate and obtain a high-order (n > ═ 2) symbolic relationship by means of the structure balance theory, for example, it is known that any two sides can obtain a predicted value of the third side in the balance structure (a, b), we call the third side as a virtual side, the existence of the virtual side is the basis of the balance theory as a constraint, the appearance of the constraint information makes the symbolic propagation information obtained by the unsigned label more perfect, and the advantage of the semi-supervised symbolic propagation based on the graph will be reflected: during the symbol propagation, a large amount of unlabeled data will flow through and be given predictive labels, and the implied knowledge will play a role in the spectrum domain graph convolution.
3. In order to apply the symbolic network to the graph convolution network, the embodiment defines the form of the directed active propagation adjacency matrix. First, the basic definition of an unsigned network is introduced.
An unsigned network may be defined abstractly as G (V, E) where V {1, 2.., n } is the set of nodes of the graph and E is the set of edges of the graph, where E (i, j) ∈ E, E (i, j) ∈ {0,1}, where E (i, j) } 0 indicates that there is no edge pointing from point i to point j, E (i, j) } 1 indicates that there is an edge (i, j ∈ V) pointing from point i to point j.
The sign network further defines signs on the edge on the basis of an unsigned network, i.e., G ═ V, E, W, where W (i, j) ∈ W, W (i, j) ∈ { -1,0,1 }. W (i, j) — 1 indicates that the sign of the edge E (i, j) is negative, W (i, j) ═ 0 indicates that the sign of the edge E (i, j) is unknown, W (i, j) ═ 1 indicates that the sign of the edge E (i, j) is positiven*nExpressed as shown in equation (1):
further, in order to enable the input symbol network adjacency matrix to be successfully applied in the graph convolution, the embodiment gives the directed symbol network propagation adjacency matrix a based on the symbol propagationsignThe input is converted by the processing as shown in the formula (2).
Asign＝A+AT+I (2)
Wherein I is a unit array. A. thesignEach element in (a) can be calculated by equation (3):
Asignthe matrix not only retains original weight information, but also screens conflicting 0-order symbol information, for example, i and j are friends, j and i are enemies, the wrong polarity relationship in the symbol network is eliminated, the symbol propagation rule is applied to the matrix, and meanwhile, a self-loop is added to lay down symbol propagation in subsequent graph convolution.
The propagation of the elements of the adjacency matrix towards the symbol network can therefore be expressed in the following form.
It can be seen that: if A is a friend of B, B is a friend of A, then they must be friends of each other (symbol weight labeled 2); a is a friend of B, the relationship of B to A is unknown, then A and B are likely friends (symbol weight labeled 1); a is a friend of B, B is an enemy of A (symbol weight is marked as 0); the same goes on.
Of course, AsignMay interfere with the evolution and generation of the normal structural equilibrium triangle. Therefore, we activate A using the sign functionsignObtaining a directed active propagation adjacency matrix
As shown in equation (5).
Wherein sgn (×) is a sign function, and its mathematical expression is shown in formula (6)
I.e., the present embodiment eliminates the weight information and retains the symbol information. The elements of the directed symbol network activation propagation adjacency matrix may represent a form as shown in equation (7).
4. The present embodiment defines the form of the symbolic Laplacian matrix in order that the symbolic network may be applied to the graph convolution algorithm, the unsigned network Laplacian matrix L∈ R used in the graph convolution networkn*nAs shown in equation (8)
L＝D-A (8)
Its medium matrix D ∈ Rn*nAs shown in formula (9)
A is the adjacency matrix of the figure, A ∈ Rn*nGraph convolution networks use Laplace matrices for unsigned networks, andfor the introduction of negative edges in the symbol network, the solution is obtained according to the related definition proposed for the symbol networksign∈Rn*nAs shown in equation (10)
Its medium matrix
As shown in formula (11)
Through the definition of the symbolic network related to the embodiment, the symbolic network is successfully applied to the graph convolution method.
5. After a new definition form of the Laplace matrix is given, the spectral decomposition is successfully applied to the symbol Laplace matrix to obtain
Lsign＝UΛUT(12)
Wherein U is a matrix of feature vectors,
Λ is a diagonal matrix of eigenvalues, Λ ═ diag (λ)1,λ2,...,λn)。
Then, taking the eigenvector U of the symbol Laplace matrix as the base of the Fourier transform, the Fourier forward and inverse transform rule on the graph is defined:
where P represents any node vector in the graph. The feature vectors extracted by the symbol Laplace matrix spectrum decomposition are used as the basis of Fourier transform to realize the conversion of a symbol network into a frequency domain, and the convolution operation is realized by converting a convolution kernel into the frequency domain:
wherein
Representing the graph convolution rule of the symbol network, and X is a convolution kernel matrix. And can then be applied to the graph convolution method.
6. And applying the obtained symbol network series definition to a Graph Convolution Network (GCN) to obtain an embedding result of a target network, namely a feature vector of each user.
Specifically, the step 5 includes: the high-order symbolic reachable matrix is used as the input of the GCN, the symbolic Laplace matrix is applied, the whole network is composed of two graph convolution layers, the output of each layer is activated by a tanh function, and therefore negative symbols can be effectively expressed and transmitted in the network. And inputting the directional activation propagation adjacency matrix, and testing a graph convolution method of a spectral domain by means of a symbol Laplace matrix and Fourier transform so as to obtain an embedding result of the target network finally.
As a specific example, the specific description of embodiment 1 is made by using network data as part of symbols in the relationships data set. The operation flow of this embodiment will be described, wherein there are 15216 user nodes, and 597179 edges are shared in the network, wherein 525204 edges are positive edges, and 71975 edges are negative edges.
TABLE 1 partial symbol network data in the Epinions dataset
| Input data parameters | Value of |
| Number of network nodes | 15216 |
| Inter-node side information | 597179 |
| Inter-node forward side information | 525204 |
| Negative side information between nodes | 71975 |
Inputting the network into this embodiment, it is necessary to convert the adjacency matrix into a form of a directed active propagation adjacency matrix. Because the number of nodes in the example is too many, complete visualization presentation analysis is not convenient, and 7 nodes are not selected from the nodes and visualized as shown in fig. 2. The initial adjacency matrix is shown as equation (16). In fig. 2, the dotted line edge indicates that the relationship between the two nodes is positive, and the solid line edge indicates that the relationship between the two nodes is negative. The corresponding directed propagation adjacency matrix of the example is obtained, as shown in equation (17).
And converting the output calculation after the symbol network processing into a symbol Laplace matrix to obtain a result shown as a formula (18).
And carrying out convolution operation on the network feature extraction result to realize the application of graph convolution on the symbol network and obtain the embedded information of the symbol network node.
In order to visually display the process of embedding information to reconstruct the trust network, the invention uses the result as the code of the trust symbol network, and uses an inner product decoding layer to quantize and express a similarity measurement matrix between nodes in the network, thereby obtaining a reconstructed symbol network adjacency matrix. And the reverse conversion symbol network propagates the adjacency matrix, so that the prediction result of the system on the user potential symbol relationship in the trust symbol network is obtained. Finally, the predicted comment trust network can be displayed visually, a predicted symbolic relationship, a trusted relationship or an untrusted relationship exists between any two users, and the user comment website can selectively open comments or hide comments for each user according to the predicted symbolic relationship, so that the user can make a correct decision.
The effect of this example can be visualized by fig. 3, where the network embedding result is applied to the symbolic network link prediction task, where the abscissa represents the number of experiments.
Example 2:
a symbolic network application system based on an improved Graph Convolution Network (GCN) applies a graph convolution method applied to an unsigned network to a symbolic network, defines symbolic network symbolic propagation rules and is used for calculating a defined directional activation propagation adjacency matrix and a defined symbolic Laplace matrix.
As shown in fig. 5, the method specifically includes:
and the symbol propagation module can define the propagation rule of the symbols in the symbol network according to the task type of the symbol network and is applied to the semi-supervised symbol network embedding method.
And the symbol network processing module formats the input of the system and converts the input adjacency matrix into a form of a directional activation propagation adjacency matrix defined by the system.
And the network characteristic extraction module is used for acquiring the output of the network processing module and converting the output into a symbol Laplace matrix to lay a cushion for the application of subsequent graph convolution.
And (3) applying a graph convolution module, initializing a weight matrix, and performing convolution operation on the output of the network feature extraction module to realize the application of graph convolution on the symbol network.
Example 3:
a computer readable storage medium having stored therein a plurality of instructions adapted to be loaded by a processor of a terminal device and to execute a modified graph convolutional network based semi-supervised symbol network embedding method.
Example 4:
a terminal device comprising a processor and a computer readable storage medium, the processor being configured to implement instructions; the computer readable storage medium is used for storing a plurality of instructions which are suitable for being loaded by a processor and executing the semi-supervised symbol network embedding method based on the improved graph convolutional network.
As will be appreciated by one skilled in the art, embodiments of the present disclosure may be provided as a method, system, or computer program product. Accordingly, the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present disclosure may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and so forth) having computer-usable program code embodied therein.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure, and various modifications and changes may be made to the present disclosure by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
Although the present disclosure has been described with reference to specific embodiments, it should be understood that the scope of the present disclosure is not limited thereto, and those skilled in the art will appreciate that various modifications and changes can be made without departing from the spirit and scope of the present disclosure.
Claims (10)
1. A semi-supervised symbol network embedding method based on an improved graph convolutional network is characterized in that: the method comprises the following steps:
importing relational network data to form a directed symbol network;
calculating an adjacency matrix of a directed symbol network to obtain a propagation adjacency matrix, and activating the propagation adjacency matrix by using a symbol function to obtain a directed activation propagation adjacency matrix;
constructing a symbol Laplace matrix, and applying the symbol Laplace matrix by the graph convolution network to realize the improvement of the graph convolution network;
and carrying out convolution operation on the directed activation propagation adjacency matrix by using the improved graph convolution network to obtain embedding results of different degrees.
2. The semi-supervised symbol network embedding method based on the improved graph convolutional network as claimed in claim 1, wherein: and the symbol propagation in the directed symbol network is based on the structural balance theory propagation and obtains a high-order symbol relationship.
3. The semi-supervised symbol network embedding method based on the improved graph convolutional network as claimed in claim 1, wherein: the directed sign network is represented by an adjacency matrix, and when the sign of an edge pointing from one point to another is positive, the corresponding matrix element is 1, and is negative, the corresponding matrix element is-1, and when unknown, the corresponding matrix element is 0.
4. The semi-supervised symbol network embedding method based on the improved graph convolutional network as claimed in claim 1, wherein: the propagation adjacency matrix is the sum of the adjacency matrix, the transposed matrix of the adjacency matrix and the unit matrix.
5. The semi-supervised symbol network embedding method based on the improved graph convolutional network as claimed in claim 1, wherein: the symbol Laplace matrix is obtained by constructing a degree matrix and an adjacent matrix of a graph on the basis of the unsigned network Laplace matrix and based on the directed activation propagation adjacent matrix.
6. The semi-supervised symbol network embedding method based on the improved graph convolutional network as claimed in claim 1, wherein: the specific process of applying the symbol laplacian matrix by the graph convolution network comprises the following steps: carrying out spectrum decomposition on the symbol Laplacian matrix, defining a Fourier forward and inverse transformation rule on a graph by taking a feature vector of the symbol Laplacian matrix as a Fourier transformation base, and realizing the conversion of a symbol network to a frequency domain;
and converting the convolution kernel into a frequency domain to realize convolution operation.
7. The semi-supervised symbol network embedding method based on the improved graph convolutional network as claimed in claim 1, wherein: and acquiring the embedded information of the symbolic network by using the improved graph convolution network, and acquiring a corresponding feature vector for each node in the network.
8. A semi-supervised symbol network embedding system based on an improved graph convolutional network is characterized in that: the method comprises the following steps:
the symbol propagation module is configured to calculate an adjacency matrix of the directed symbol network, and further obtain a propagation adjacency matrix;
a symbolic network processing module configured to activate the propagation adjacency matrix using a symbolic function, resulting in a directed activation propagation adjacency matrix;
the network feature extraction module is configured to construct a symbol Laplace matrix, and the graph convolution network applies the symbol Laplace matrix to improve the graph convolution network;
and the graph convolution module is configured to perform convolution operation on the input adjacent matrix by using the improved graph convolution network to obtain embedding results of different degrees.
9. A computer-readable storage medium characterized by: a plurality of instructions are stored, the instructions are suitable for being loaded by a processor of a terminal device and executing the semi-supervised symbol network embedding method based on the improved graph convolutional network as set forth in any one of claims 1 to 7.
10. A terminal device is characterized in that: the system comprises a processor and a computer readable storage medium, wherein the processor is used for realizing instructions; the computer readable storage medium is used for storing a plurality of instructions, the instructions are suitable for being loaded by a processor and executing the semi-supervised symbol network embedding method based on the improved graph convolutional network as set forth in any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010090482.XA CN111401514A (en) | 2020-02-13 | 2020-02-13 | Semi-supervised symbol network embedding method and system based on improved graph convolutional network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010090482.XA CN111401514A (en) | 2020-02-13 | 2020-02-13 | Semi-supervised symbol network embedding method and system based on improved graph convolutional network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111401514A true CN111401514A (en) | 2020-07-10 |
Family
ID=71428397
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010090482.XA Pending CN111401514A (en) | 2020-02-13 | 2020-02-13 | Semi-supervised symbol network embedding method and system based on improved graph convolutional network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111401514A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112035334A (en) * | 2020-09-15 | 2020-12-04 | 深圳市欢太科技有限公司 | Abnormal equipment detection method and device, storage medium and electronic equipment |
| CN112529069A (en) * | 2020-12-08 | 2021-03-19 | 广州大学华软软件学院 | Semi-supervised node classification method, system, computer equipment and storage medium |
| CN113362963A (en) * | 2021-05-27 | 2021-09-07 | 山东师范大学 | Method and system for predicting side effects among medicines based on multi-source heterogeneous network |
-
2020
- 2020-02-13 CN CN202010090482.XA patent/CN111401514A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112035334A (en) * | 2020-09-15 | 2020-12-04 | 深圳市欢太科技有限公司 | Abnormal equipment detection method and device, storage medium and electronic equipment |
| CN112035334B (en) * | 2020-09-15 | 2023-01-31 | 深圳市欢太科技有限公司 | Abnormal equipment detection method and device, storage medium and electronic equipment |
| CN112529069A (en) * | 2020-12-08 | 2021-03-19 | 广州大学华软软件学院 | Semi-supervised node classification method, system, computer equipment and storage medium |
| CN112529069B (en) * | 2020-12-08 | 2023-10-13 | 广州大学华软软件学院 | Semi-supervised node classification method, system, computer equipment and storage medium |
| CN113362963A (en) * | 2021-05-27 | 2021-09-07 | 山东师范大学 | Method and system for predicting side effects among medicines based on multi-source heterogeneous network |
| CN113362963B (en) * | 2021-05-27 | 2024-04-02 | 山东师范大学 | Method and system for predicting side effects among medicines based on multi-source heterogeneous network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109582956B (en) | Text representation method and device applied to sentence embedding | |
| CN109960810B (en) | Entity alignment method and device | |
| KR20220005416A (en) | Method for training multivariate relationship generation model, electronic device and medium | |
| CN111401514A (en) | Semi-supervised symbol network embedding method and system based on improved graph convolutional network | |
| CN107944545B (en) | Computing method and computing device applied to neural network | |
| JP2020523699A (en) | Generate point of interest copy | |
| JP7438303B2 (en) | Deep learning model training methods, natural language processing methods and devices, electronic devices, storage media and computer programs | |
| CN112214608A (en) | Text generation method, medium, device and computing equipment based on knowledge reasoning | |
| CN113254716B (en) | Video clip retrieval method and device, electronic equipment and readable storage medium | |
| CN113254729B (en) | Multi-modal evolution characteristic automatic conformal representation method based on dynamic hypergraph network | |
| CN112883198B (en) | Knowledge graph construction method and device, storage medium and computer equipment | |
| Huai et al. | Zerobn: Learning compact neural networks for latency-critical edge systems | |
| CN108171328A (en) | A kind of convolution algorithm method and the neural network processor based on this method | |
| US20210089904A1 (en) | Learning method of neural network model for language generation and apparatus for performing the learning method | |
| Nepomuceno et al. | On the use of interval extensions to estimate the largest Lyapunov exponent from chaotic data | |
| CN113505583A (en) | Sentiment reason clause pair extraction method based on semantic decision diagram neural network | |
| CN115358397A (en) | Parallel graph rule mining method and device based on data sampling | |
| CN114385964B (en) | State space model calculation method, system and equipment of multi-element fractional order system | |
| CN111310996B (en) | User trust relation prediction method and system based on graph self-coding network | |
| JP2018041300A (en) | Machine learning model generation device and program | |
| CN111199153B (en) | Word vector generation method and related equipment | |
| CN116049691A (en) | Model conversion method, device, electronic equipment and storage medium | |
| CN111414989B (en) | User trust relationship network link prediction method and system based on gating mechanism | |
| CN110457692B (en) | Method and device for learning compound word representation | |
| CN114332472A (en) | Data processing method and device based on graph neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200710 |