CN113111302B - Information extraction method based on non-European space - Google Patents
Information extraction method based on non-European space Download PDFInfo
- Publication number
- CN113111302B CN113111302B CN202110431000.7A CN202110431000A CN113111302B CN 113111302 B CN113111302 B CN 113111302B CN 202110431000 A CN202110431000 A CN 202110431000A CN 113111302 B CN113111302 B CN 113111302B
- Authority
- CN
- China
- Prior art keywords
- space
- hyperbolic
- hyperbolic space
- information
- mapping
- 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
- 238000000605 extraction Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 34
- 238000013507 mapping Methods 0.000 claims abstract description 28
- 230000008569 process Effects 0.000 claims abstract description 12
- 238000009826 distribution Methods 0.000 claims abstract description 5
- 239000011159 matrix material Substances 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 13
- 230000004927 fusion Effects 0.000 claims description 12
- 238000004220 aggregation Methods 0.000 claims description 6
- 230000002776 aggregation Effects 0.000 claims description 6
- 125000004122 cyclic group Chemical group 0.000 claims description 5
- 238000012614 Monte-Carlo sampling Methods 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 12
- 238000013528 artificial neural network Methods 0.000 description 10
- 230000004913 activation Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000005520 cutting process Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012733 comparative method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000000644 propagated effect Effects 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
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/17—Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an information extraction method based on a non-European space, which comprises the steps of mapping a graph structure input by the European space to a hyperbolic space based on a basic algorithm of the hyperbolic space; by means of
Addition and addition
Multiplying redefines a conditional intensity function of Hawkes process in the hyperbolic space; and taking the curvature c as a parameter of the function, and iteratively updating the curvature c to obtain a trained hyperbolic space model, and describing the distribution of future events. The method of the invention maps future events into hyperbolic space by combining time and type information of past events, obtains better hierarchical structure to predict the future events, and can well represent large-scale hierarchical data in application aspect, so that the graph information is more accurate.
Description
Technical Field
The invention relates to the technical field of hyperbolic space of computers, in particular to an information extraction method based on a non-European space.
Background
To better describe hyperbolic space, the basic concept of differential geometry required for the general principles of neural networks in Euclidean space is briefly described herein, an n-dimensional manifold
Is a local available +.>
Space of approximation: it is a popularization of two-dimensional curved surface 2D concept in higher dimension, for +.>
Can be +.>
The cutting space at x->
Defined as->
First order linear approximation around x, +.>
Riemann metric->
Is an inner product g x :/>
Is of the type Riemann manifold->
Is a manifold with Riemann metric g>
Although the Riemann metric g appears to define geometry only locally, it yields the global distance by integrating the length of the shortest path between two points (the velocity vector present in the tangent space):
wherein,,
and γ (0) =, γ (1) = y, the minimum length path γ between two points x and y is called geodesic, and can be regarded as a generalization of a straight line in euclidean space, P x→y :T x M→T y M corresponds to the tangent vector moving along the geodesic line and defines a canonical method of connecting the tangent spaces, expo is mapped exponentially x A method is provided for cutting the space +.>
Is projected to a point on the manifold +.>
The poincare sphere model considered herein, exp x In the whole cutting space->
The method is well defined.
Disclosure of Invention
This section is intended to outline some aspects of embodiments of the invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section as well as in the description summary and in the title of the application, to avoid obscuring the purpose of this section, the description summary and the title of the invention, which should not be used to limit the scope of the invention.
The present invention has been made in view of the above-described problems occurring in the prior art.
Therefore, the invention provides an information extraction method based on non-European space, which can solve the problem that better node characteristic information and European space mapping distortion cannot be obtained.
In order to solve the technical problems, the invention provides the following technical scheme: comprises mapping a graph structure of European space input to a hyperbolic space based on a basic algorithm of the hyperbolic space; by means of
Add and->
Multiplying redefines a conditional intensity function of Hawkes process in the hyperbolic space; and taking the curvature c as a parameter of the function, and iteratively updating the curvature c to obtain a trained hyperbolic space model, and describing the distribution of future events.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the method also comprises the steps of graph information embedding of the hyperbolic space, information fusion of the hyperbolic space, category characteristic embedding of the hyperbolic space, information fusion, prediction, training and fine adjustment.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the graph information embedding of the hyperbolic space comprises the steps of mapping the adjacent matrix of the user node to the hyperbolic space, and then utilizing the GCN in the hyperbolic space to aggregate the adjacent matrix information of the graph structure.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the information fusion of the hyperbolic space comprises mapping specific time intervals and occurrence time matrixes of past occurrence events into the hyperbolic space for aggregation.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the embedding of category features of the hyperbolic space includes mapping a category matrix of all past occurrences into the hyperbolic space.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the information fusion comprises the steps of inputting the adjacent matrix and the feature matrix into the hyperbolic space GCN, and connecting the output of the GCN with the category matrix and the time matrix after mapping the input of the hyperbolic space.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the prediction comprises the steps of simulating and calling an inverse method and Monte Carlo sampling, and predicting the type and time of an event which occurs in the future.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the training comprises the steps of dividing a data set, selecting 10 events as one training, predicting event time and category of the next event, and training through a cross entropy loss function.
As a preferable embodiment of the non-european space based information extraction method of the present invention, the method further includes: the fine tuning comprises a series of fine tuning by selecting different batch size, different cyclic neural network models and learning rate, and the optimal output is completed.
The invention has the beneficial effects that: the method of the invention maps future events into hyperbolic space by combining time and type information of past events, obtains better hierarchical structure to predict the future events, and can well represent large-scale hierarchical data in application aspect, so that the graph information is more accurate.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. Wherein:
FIG. 1 is a schematic diagram of a Nostradamus model of a non-European space-based information extraction method according to an embodiment of the present invention;
FIG. 2 is a flow chart of a Nostradamus model of a non-European space-based information extraction method according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of root mean square error of a synthetic dataset according to a non-European space-based information extraction method according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a root mean square error of a Stack Overflow data set according to a non-European space based information extraction method according to an embodiment of the present invention;
FIG. 5 is a root mean square error diagram of a failure dataset of a self-service deposit and withdrawal machine according to a non-European space-based information extraction method according to an embodiment of the present invention;
fig. 6 is a schematic diagram of root mean square error of a financial transaction data set according to a non-european space-based information extraction method according to an embodiment of the present invention.
Detailed Description
So that the manner in which the above recited objects, features and advantages of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to the embodiments, some of which are illustrated in the appended drawings. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
While the embodiments of the present invention have been illustrated and described in detail in the drawings, the cross-sectional view of the device structure is not to scale in the general sense for ease of illustration, and the drawings are merely exemplary and should not be construed as limiting the scope of the invention. In addition, the three-dimensional dimensions of length, width and depth should be included in actual fabrication.
Also in the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper, lower, inner and outer", etc. are based on the orientation or positional relationship shown in the drawings, are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the apparatus or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
The terms "mounted, connected, and coupled" should be construed broadly in this disclosure unless otherwise specifically indicated and defined, such as: can be fixed connection, detachable connection or integral connection; it may also be a mechanical connection, an electrical connection, or a direct connection, or may be indirectly connected through an intermediate medium, or may be a communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
Example 1
Referring to fig. 1 and 2, for a first embodiment of the present invention, there is provided a non-european space-based information extraction method, including:
s1: the basic algorithm based on hyperbolic space maps the graph structure of the European space input to the hyperbolic space.
S2: by means of
Add and->
The multiplication redefines the conditional intensity function of the Hawkes process in hyperbolic space.
S3: and taking the curvature c as a parameter of a function, and iteratively updating the curvature c to obtain a trained hyperbolic space model, and describing the distribution of future events.
Preferably, in the embodiment, the graph convolutional neural network and the cyclic neural network are defined in hyperbolic space, time and space information of the past occurrence event is obtained by utilizing a Hooke process, a conditional intensity function similar to a parameter of the Hooke process is converted into a non-parameter form by utilizing the cyclic neural network to approximate the Hooke process, and the effect of predicting the future event is improved by fine adjustment of the parameter.
Specifically, the non-euro space-based information extraction method provided in the embodiment specifically includes the following steps:
graph information embedding of hyperbolic space, information fusion of hyperbolic space, category characteristic embedding of hyperbolic space, information fusion, prediction, training and fine adjustment;
(1) The graph information embedding of the hyperbolic space comprises the steps of mapping an adjacent matrix of a user node to the hyperbolic space, and then utilizing the GCN in the hyperbolic space to aggregate the adjacent matrix information of the graph structure;
(2) The information fusion of the hyperbolic space comprises the steps of mapping a specific time interval of a past occurrence event and an occurrence time matrix into the hyperbolic space for aggregation;
(3) Category feature embedding of the hyperbolic space comprises mapping a category matrix of all past occurrences into the hyperbolic space;
(4) The information fusion comprises the steps of inputting an adjacent matrix and a feature matrix into a hyperbolic space GCN, and connecting the output of the GCN with a category matrix and a time matrix after mapping the input hyperbolic space;
(5) The prediction comprises the steps of simulating and calling an inverse method and Monte Carlo sampling, and predicting the type and time of an event which occurs in the future;
(6) The training comprises dividing a data set, selecting 10 events as one training, predicting event time and category of the next event, and training through a cross entropy loss function;
(7) The fine tuning comprises a series of fine tuning by selecting different batch size, different cyclic neural network models and learning rate, and completing the optimal output.
In order to obtain a good graph embedding, namely characteristic information of nodes, and overcome mapping distortion of European space so as to obtain a good event time and type prediction effect, the embodiment adopts a Poincare sphere model, a graph convolution neural network and GRU to predict, so that the basic thought of non-European space-based information extraction is as follows:
mapping the graph structure input in European space to hyperbolic space, and adopting a method of combining the hyperbolic space and a graph convolution neural network;
in order to solve the hyperbolic space of the input mapping of the European space, the aggregation formula operation of the graph convolution neural network is replaced by a basic algorithm of the hyperbolic space;
to define conditions of Hawkes process in hyperbolic spaceIntensity function, the operation of the function is used
Add and->
Multiplying redefines the conditional intensity function;
in order to make the hyperbolic space model with proper curvature in each iteration, the curvature c is taken as a parameter of a loss function, and the curvature c is continuously and iteratively updated in the training of the model so as to solve the second challenge, so that the model can better describe the distribution of future events.
Example 2
Referring to fig. 3 to 6, in a second embodiment of the present invention, which is different from the first embodiment, there is provided experimental verification of a non-european space based information extraction method, specifically including:
for the euro-space diagram feature, the conventional graph roll-up neural network is used to capture the aggregate information between nodes by constructing a filter in the fourier domain, and the specific operation is as follows:
in order to learn more valuable information from the hierarchical structure, the embodiment selects a graph convolution neural network defined in hyperbolic space to obtain spatial dependency, and the specific implementation method is shown in the following formula.
Therefore, for the input features of the European space and the output of the model, the conversion of the European space features and the hyperbolic space features is realized according to an Exponential mapping (Exponential) and a logarithmic mapping (logarithmic), and the following quotients give closed deductions of the Exponential mapping and the logarithmic mapping:
for any point
When v+.0, exponential mapping +.>
Can be given by:
when y+.x, log mapping
Can be given by:
given hyperbolic spatial curvatures K of different layers l-1 and l in a neural network l-1 ,K l Nonlinear activation function using hyperbolic spaces of different curvatures
For nonlinear activation functions in the European space, the logarithmic mapping is used to map data to the European space for activation, and then the exponential mapping is used to return to the hyperbolic space, the operation is as follows.
Wherein the superscript H represents a hyperbolic space.
When obtaining the European space diagram, capturing the space characteristics of the diagram by adopting a two-layer hyperbolic space diagram convolutional neural network, wherein one layer of hyperbolic space diagram convolutional neural network layer is defined as the following formula:
wherein K is l-1 And K l Hyperbolic space curvature, W, of convolutions layers l-1 and l, respectively l Representing the weighting matrix of the l-1 layer, and the output hyperbolic spatial feature H of the last layer (l+1) The next stage is input as spatial information to predict the time and type of the next event occurrence.
In the hyperbolic space diagram neural Hawkes process stage, when output information H of a hyperbolic space diagram convolutional neural network (HGCN) is obtained (l+1) With past dynamically propagated sequences
The present embodiment proposes a HGNHP stage (Hyperbolic Graph Neural Hawkes Process Stage) based on a gated loop unit (GRU) and poincare sphere, which predicts the occurrence type and time of the next event by iterating the following five steps.
Referring to FIG. 2, the information of the past event and the output of the previous stage include { y } i-1 ,g,t i-1 As triples, the first term is
The read-heat coding of the middle node, the second item is the graph embedding learned from the first stage and represents the aggregation information of the corresponding node in the hyperbolic space, and the last item is the occurrence time t of the current event i-1 。
Hidden layer status update, as shown in FIG. 1, the detailed information of HGNHP units, h, is shown on the right side t-1 Representing t i-1 Hidden state of time, r t And u t Reset gate and update gate of GRU when current input is receivedAnd h t-1 After the hidden state, the calculating process of the hidden layer state update is as follows:
wherein,,
representing an activation function, g representing node aggregation information captured by a graph convolution neural network defined in hyperbolic space, k representing an event k embedding occurring at a current time, t representing a time interval from a previous occurrence of the event, W representing a weight, and b representing a deviation, so that an event propagation probability P (yn+1hn) and a conditional intensity function approximating Hawkes process can be calculated.
Event propagation probability given h i Decoding the hidden state into European space by exponential mapping, and calculating
The specific calculation is as follows:
wherein N is a graph
Node number of->
Deviation (I)>
It is k-th that is the th row of matrix V.
Referring to fig. 3, 4, 5 and 6, in the artificially generated dataset, the RMSE of noststradamus is 2.84, which is superior to all the comparative methods, in the automatic teller machine failure dataset, the RMSE of the most advanced method is 2.256, the RMSE of noststradamus is predicted to be 2.82, and the RMSE of noststradamus is predicted to be 10.05, which is similar to the most advanced method, constructing a graph-based dataset to verify that adding graph information can improve the performance of timestamp prediction, and these results indicate that noststradamus performs better than other existing methods when applied to the artificially synthesized dataset.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.
Claims (3)
1. An information extraction method based on non-European space is characterized by comprising the following steps: comprising the steps of (a) a step of,
mapping a graph structure input by an European space to a hyperbolic space based on a basic algorithm of the hyperbolic space, and further comprising graph information embedding of the hyperbolic space, information fusion of the hyperbolic space, category characteristic embedding of the hyperbolic space, information fusion, prediction, training and fine adjustment;
the graph information embedding of the hyperbolic space comprises the steps of mapping an adjacent matrix of a user node to the hyperbolic space, and then utilizing the GCN in the hyperbolic space to aggregate the adjacent matrix information of the graph structure;
the information fusion of the hyperbolic space comprises the steps of mapping a specific time interval and an occurrence time matrix of an event in the past into the hyperbolic space for aggregation;
the category characteristic embedding of the hyperbolic space comprises mapping a category matrix of all past occurrences into the hyperbolic space;
the information fusion comprises the steps of inputting the adjacent matrix and the feature matrix into the hyperbolic space GCN, and connecting the output of the GCN with the category matrix and the time matrix after mapping the input of the hyperbolic space;
the prediction comprises the steps of simulating and calling an inverse method and Monte Carlo sampling, and predicting the type and time of an event occurring in the future;
by means of
Add and->
Multiplying redefines a conditional intensity function of Hawkes process in the hyperbolic space;
and taking the curvature c as a parameter of the function, and iteratively updating the curvature c to obtain a trained self-service deposit and withdrawal machine fault hyperbolic space model, and describing the distribution of future events.
2. The non-european space based information extraction method of claim 1, wherein: the training comprises the steps of dividing a data set, selecting 10 events as one training, predicting event time and category of the next event, and training through a cross entropy loss function.
3. The non-european space based information extraction method of claim 2, wherein: the fine tuning comprises a series of fine tuning by selecting different batch size, different cyclic neural network models and learning rate, and the optimal output is completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110431000.7A CN113111302B (en) | 2021-04-21 | 2021-04-21 | Information extraction method based on non-European space |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110431000.7A CN113111302B (en) | 2021-04-21 | 2021-04-21 | Information extraction method based on non-European space |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113111302A CN113111302A (en) | 2021-07-13 |
| CN113111302B true CN113111302B (en) | 2023-05-12 |
Family
ID=76719112
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110431000.7A Active CN113111302B (en) | 2021-04-21 | 2021-04-21 | Information extraction method based on non-European space |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113111302B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113947088A (en) * | 2021-09-15 | 2022-01-18 | 阿里巴巴达摩院(杭州)科技有限公司 | Emotion polarity representation method, emotion polarity representation device and computer storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800504A (en) * | 2019-01-21 | 2019-05-24 | 北京邮电大学 | A kind of embedding grammar and device of heterogeneous information network |
| CN109857457A (en) * | 2019-01-29 | 2019-06-07 | 中南大学 | A kind of function level insertion representation method learnt in source code in the hyperbolic space |
| CN111782768A (en) * | 2020-06-30 | 2020-10-16 | 首都师范大学 | Fine-grained entity identification method based on hyperbolic space representation and label text interaction |
| CN112131490A (en) * | 2020-09-18 | 2020-12-25 | 东南大学 | Region-sensitive interest point recommendation method driven by knowledge graph |
| CN112287126A (en) * | 2020-12-24 | 2021-01-29 | 中国人民解放军国防科技大学 | Entity alignment method and device suitable for multi-mode knowledge graph |
| CN112328801A (en) * | 2020-09-28 | 2021-02-05 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Method for predicting group events by event knowledge graph |
-
2021
- 2021-04-21 CN CN202110431000.7A patent/CN113111302B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109800504A (en) * | 2019-01-21 | 2019-05-24 | 北京邮电大学 | A kind of embedding grammar and device of heterogeneous information network |
| CN109857457A (en) * | 2019-01-29 | 2019-06-07 | 中南大学 | A kind of function level insertion representation method learnt in source code in the hyperbolic space |
| CN111782768A (en) * | 2020-06-30 | 2020-10-16 | 首都师范大学 | Fine-grained entity identification method based on hyperbolic space representation and label text interaction |
| CN112131490A (en) * | 2020-09-18 | 2020-12-25 | 东南大学 | Region-sensitive interest point recommendation method driven by knowledge graph |
| CN112328801A (en) * | 2020-09-28 | 2021-02-05 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Method for predicting group events by event knowledge graph |
| CN112287126A (en) * | 2020-12-24 | 2021-01-29 | 中国人民解放军国防科技大学 | Entity alignment method and device suitable for multi-mode knowledge graph |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113111302A (en) | 2021-07-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111161535B (en) | Attention mechanism-based graph neural network traffic flow prediction method and system | |
| CN112988723A (en) | Traffic data restoration method based on space self-attention-diagram convolution cyclic neural network | |
| CN108763377B (en) | Multi-source telemetering big data feature extraction preprocessing method based on satellite fault diagnosis | |
| CN113905391A (en) | Ensemble learning network traffic prediction method, system, device, terminal, and medium | |
| CN111582610A (en) | Prediction method for family energy decomposition based on convolutional neural network | |
| CN115346372B (en) | Multi-component fusion traffic flow prediction method based on graph neural network | |
| CN113111302B (en) | Information extraction method based on non-European space | |
| CN115952424A (en) | Graph convolution neural network clustering method based on multi-view structure | |
| CN106355191A (en) | Deep generating network random training algorithm and device | |
| CN113688253A (en) | Hierarchical perception temporal knowledge map representation learning method | |
| CN113822419A (en) | Self-supervision graph representation learning operation method based on structural information | |
| CN113689052B (en) | Travel demand prediction method based on tensor product neural network | |
| CN117077771B (en) | Causal association driven aviation traffic network sweep effect prediction method | |
| Kim et al. | A daily tourism demand prediction framework based on multi-head attention CNN: The case of the foreign entrant in South Korea | |
| CN107451684A (en) | Stock market's probability forecasting method based on core stochastic approximation | |
| CN113688974A (en) | Mobile application recommendation method based on lightweight graph convolutional network | |
| CN113793267A (en) | Self-supervision single remote sensing image super-resolution method based on cross-dimension attention mechanism | |
| US20220164688A1 (en) | System and method for automated imputation for multi-state sensor data and outliers | |
| CN116664253B (en) | Project recommendation method based on generalized matrix decomposition and attention shielding | |
| CN113095480A (en) | Interpretable graph neural network representation method based on knowledge distillation | |
| CN114566048B (en) | Traffic control method based on multi-view self-adaptive space-time diagram network | |
| CN116106909A (en) | Radar echo extrapolation method, system and storage medium | |
| CN114861235A (en) | High-precision time sequence prediction method for inhibiting data leakage | |
| Kim et al. | A convenient approach for penalty parameter selection in robust lasso regression | |
| CN109102010B (en) | Image classification method based on bidirectional neural network structure |
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 |