CN112434812B - Knowledge graph link prediction method and system based on dual quaternion - Google Patents
Knowledge graph link prediction method and system based on dual quaternion Download PDFInfo
- Publication number
- CN112434812B CN112434812B CN202011346008.5A CN202011346008A CN112434812B CN 112434812 B CN112434812 B CN 112434812B CN 202011346008 A CN202011346008 A CN 202011346008A CN 112434812 B CN112434812 B CN 112434812B
- Authority
- CN
- China
- Prior art keywords
- knowledge graph
- quaternion
- dual
- triplet
- data
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/027—Frames
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a knowledge graph link prediction method and a system based on dual quaternion, wherein the method comprises the following steps: loading data and analyzing the data to obtain triplet data of the knowledge graph; training and parameter adjustment are carried out on a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph, so that a trained dual quaternion knowledge graph model is obtained; and predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result. The system comprises: the system comprises a data loading module, a model training module and a link prediction module. By using the method and the device, the problem that a plurality of relations exist between the head entity and the tail entity in the knowledge graph can be solved by effectively utilizing the characteristics of dual quaternions. The knowledge graph link prediction method and system based on dual quaternion can be widely applied to the field of knowledge graphs.
Description
Technical Field
The application belongs to the field of knowledge maps, and particularly relates to a knowledge map link prediction method and system based on dual quaternion.
Background
In the knowledge graph, each fact may be represented by a triplet (h, r, t), where h represents a head entity, t represents a tail entity, and r represents a relationship between the head entity and the tail entity. Knowledge graph representation learning is a basic and challenging work in the field of knowledge graphs, and has become an important component of knowledge graph completion. Knowledge-graph representation learning (knowledge-graph embedding) captures semantic information between entities and relationships in a knowledge-graph by mapping them to low-dimensional spatial vectors. There are a number of relationships between entities in a real scene and these relationships are completely uncorrelated. However, the current knowledge-graph representation learning model ignores this aspect and lacks consideration of the problem of relationship diversity among entities in the knowledge graph.
Disclosure of Invention
In order to solve the technical problems, the application aims to provide a knowledge graph link prediction method and a knowledge graph link prediction system based on dual quaternion, which are used for effectively solving the problem that a plurality of relations exist between a head entity and a tail entity by using pure quaternion vectors to represent the entities and unit dual quaternion vectors to represent the relations and modeling the relations as the combination of rotation and translation of a three-dimensional space.
The first technical scheme adopted by the application is as follows: a knowledge graph link prediction method based on dual quaternion comprises the following steps:
loading data and analyzing the data to obtain triplet data of the knowledge graph;
training and parameter adjustment are carried out on a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph, so that a trained dual quaternion knowledge graph model is obtained;
and predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result.
Further, the preset dual quaternion knowledge graph model comprises a dual quaternion knowledge graph embedding module, a scoring function module, a negative sampling module and a loss function module.
Further, the step of training and parameter adjustment on the preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph to obtain the trained dual quaternion knowledge graph model specifically comprises the following steps:
processing the triplet data of the knowledge graph to obtain an entity and a relation index;
initializing entities and relations in the knowledge graph based on the quaternion vector;
constructing a score function of the triplet;
generating a negative sample according to the triplet data of the knowledge graph;
and constructing a loss function according to the triplet data and the negative sample of the knowledge graph, and carrying out parameter adjustment on the score function based on the loss function to obtain the training-completed dual-quaternion knowledge graph model parameters.
Further, the step of initializing the entity and the relationship in the knowledge graph based on the quaternion vector specifically further includes:
initializing an entity in the knowledge graph by using a pure quaternion vector;
and initializing the relation in the knowledge graph by using a unit dual quaternion vector.
Further, the step of generating a negative sample according to the triplet data of the knowledge graph specifically includes:
for a correct triplet, randomly replacing the head and tail entities in the triplet forms a negative sample.
Further, the scoring function expression is as follows:
in the above, theInitializing the relation in the knowledge graph with a unit dual quaternion vector, dual quaternion,/->Representing the conjugation of dual quaternions, +.>Representing dual quaternion multiplication, < >>Representing the initialization of entities in a knowledge graph with pure quaternion vectors, the E being a dual unit and satisfying the E 2 ＝0。
Further, the loss function expression is as follows:
in the above formula, gamma is a super parameter, and f is r (h, t) represents a positive sample score, f r (h′ i ，t′ i ) Representing a negative sample score.
Further, the step of predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result specifically comprises the following steps:
inputting the triplet to be tested into a training-completed dual-quaternion knowledge graph model and calculating the score of the triplet;
and obtaining a prediction result according to the score, and judging whether the triplet is a fact or not.
The second technical scheme adopted by the application is as follows: a knowledge graph link prediction system based on dual quaternion comprises the following modules:
the data loading module is used for loading data and analyzing the data to obtain triplet data of the knowledge graph;
the model training module is used for training and parameter adjustment of a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph to obtain a trained dual quaternion knowledge graph model;
and the link prediction module is used for predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result.
Further, the model training module further comprises the following modules:
and the dual quaternion knowledge graph embedding module is used for processing the triplet data of the knowledge graph to obtain entity and relationship indexes, and initializing the entity and relationship in the knowledge graph based on quaternion vectors.
The scoring function module is used for constructing a scoring function of the triples and obtaining scores of the triples;
the negative sampling module is used for generating a negative sample according to the triplet data of the knowledge graph;
and the loss function module is used for constructing a loss function according to the triplet data and the negative sample of the knowledge graph and carrying out parameter adjustment on the score function based on the loss function.
The method and the system have the beneficial effects that: in the application, the dual quaternion is utilized to encode the knowledge graph, and the link prediction is further carried out by learning the vector representation of the entity and the relation in the knowledge graph, in addition, the vector representation of the pure quaternion is utilized to encode the entity in the knowledge graph, the relation is regarded as the combination of rotation and translation from the head entity to the tail entity, and the characteristic of the dual quaternion is effectively utilized to solve the problem that a plurality of relations exist between the head entity and the tail entity in the knowledge graph.
Drawings
FIG. 1 is a flow chart of steps of a knowledge-graph link prediction method based on dual quaternions in accordance with an embodiment of the present application;
fig. 2 is a block diagram of a knowledge-graph link prediction system based on dual quaternions according to an embodiment of the present application.
Detailed Description
The application will now be described in further detail with reference to the drawings and to specific examples. The step numbers in the following embodiments are set for convenience of illustration only, and the order between the steps is not limited in any way, and the execution order of the steps in the embodiments may be adaptively adjusted according to the understanding of those skilled in the art.
As shown in fig. 1, the application provides a knowledge graph link prediction method based on dual quaternion, which comprises the following steps:
s1, loading data and analyzing the data to obtain triplet data of a knowledge graph;
s2, training and parameter adjustment are carried out on a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph, so that a trained dual quaternion knowledge graph model is obtained;
s3, predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result.
Further as a preferred embodiment of the method, the preset dual-quaternion knowledge graph model includes a dual-quaternion knowledge graph embedding module, a scoring function module, a negative sampling module and a loss function module.
Further as a preferred embodiment of the method, the step of training and parameter adjusting the preset dual-quaternion knowledge graph model according to the triplet data of the knowledge graph to obtain a trained dual-quaternion knowledge graph model specifically includes:
processing the triplet data of the knowledge graph to obtain an entity and a relation index;
initializing entities and relations in the knowledge graph based on the quaternion vector;
specifically, the triplet (h, r, t) of the knowledge graph is processed, indexes are established for the entity and the relation, vector representations of the knowledge graph entity and the relation are initialized in a uniform distribution mode, and the relation of the entity in the knowledge graph to various semantic relation-free relations are modeled through the relation vector.
Constructing a score function of the triplet;
generating a negative sample according to the triplet data of the knowledge graph;
and constructing a loss function according to the triplet data and the negative sample of the knowledge graph, and carrying out parameter adjustment on the score function based on the loss function to obtain the training-completed dual-quaternion knowledge graph model parameters.
Further as a preferred embodiment of the present application, the initializing the entity and the relationship in the knowledge graph based on the quaternion vector specifically further includes:
initializing an entity in the knowledge graph by using a pure quaternion vector;
specifically, in the knowledge-graph entity embedding, each dimension in the entity vector is mapped to a three-dimensional space, and the entity vector in the pure quaternion modeling knowledge graph is used for representing. The entity in the knowledge graph is initialized by pure quaternion vector and is symbolized by symbolAnd (5) vector representation.
The definition of the quaternion extends from complex numbers, and is different from complex numbers in that the quaternion has three imaginary parts, and can be defined as:
q＝a+bi+cj+dk
for i, j, k satisfies:
i 2 ＝j 2 ＝k 2 ＝ijk＝-1
when the real part of the quaternion is zero, we define this quaternion as a pure quaternion as follows:
q＝bi+cj+dk
in the application, the entity vector in the knowledge graph is embedded by using the pure quaternion.
And initializing the relation in the knowledge graph by using a unit dual quaternion vector.
Specifically, the relation vector modeling can be used for representing the relation of the entity in the knowledge graph to various semantic-free relations. Initializing the relation vector in the knowledge graph by using unit dual quaternion and using symbolAnd (5) vector representation. Mapping the relation in the knowledge graph into the combination of three-bit space rotation and translation through unit dual quaternion, and representing the multi-relation without semantic relation in the head entity and the tail entity through different combinations of rotation and translation.
Dual quaternions are extensions to even numbers, while combining quaternions. Formally, we define dual quaternions:
δ＝p+∈q
wherein p and q are quaternions, ε is a dual unit and satisfy ε 2 =0. The dual four elements can be expanded into the following forms:
δ＝p 0 +p 1 i+p 2 j+p 3 k+∈(q 0 +q 1 i+q 2 j+q 3 k)
when the conjugate product of the dual four elements and itself is 1, namely:
we call this dual quaternion a unit dual quaternion. According to the above conditions, the unit dual quaternion satisfies:
p 0 q 0 +p 1 q 1 +p 2 q 2 +p 3 q 3 ＝0
the unit dual quaternion solution by using the conditions needs a large calculation amount and can only be solved in an exhaustive way. We have calculated the unit dual quaternion in another way. Defining quaternion:where u is a unit pure quaternion and. Definition of pure quaternions: n=n 1 i+n 2 j+n 3 k. Finally, we calculate the unit dual quaternion as follows:
through the calculation method, each dimension of the relation vector is a unit dual quaternion, and in the representation learning, the relation vector distinguishes a plurality of semantic-free relations between entity pairs in the knowledge graph through combination of rotation and translation.
Further as a preferred embodiment of the present application, the step of generating the negative sample according to the triplet data of the knowledge-graph specifically includes:
for a correct triplet, randomly replacing the head and tail entities in the triplet forms a negative sample.
Specifically, for a correct triplet (h, r, t), the head entity (h ', r, t) and the tail entity (h, r, t') randomly replaced therein form a negative sample.
Further as a preferred embodiment of the present application, the scoring function expression is as follows:
in the above, theInitializing the relation in the knowledge graph with a unit dual quaternion vector, dual quaternion,/->Representing the conjugation of dual quaternions, +.>Representing dual quaternion multiplication, < >>Representing the initialization of entities in a knowledge graph with pure quaternion vectors, the E being a dual unit and satisfying the E 2 ＝0。
For dual quaternion delta 1 ＝p 1 +∈q 1 And delta 2 ＝p 2 +∈q 2 Then there is a dual quaternion multiplication defined as follows:
p 1 p 2 、p 1 q 2 and q 1 p 2 Representing a quaternion multiplication. The multiplication of quaternions we have the following definition: quaternion p=p 0 +p 1 i+p 2 j+p 3 k and p=q 0 +q 1 i+q 2 j+q 3 k, we define the product of two numbers as follows:
pq＝p 0 q 0 -p 1 q 1 -p 2 q 2 -p 3 q 3 +(p 0 q 1 -p 1 q 0 -p 2 q 3 -p 3 q 2 )i+(p 0 q 2 -p 1 q 3 -p 2 q 0 -p 3 q 1 )j+(p 0 q 3 -p 1 q 3 -p 2 q 1 -p 3 q 0 )k
further as a preferred embodiment of the method, the loss function expression is as follows:
in the above formula, gamma is a super parameter, and f is r (h, t) represents a positive sample score, f r (h′ i ，t′ i ) Representing a negative sample score.
Further as a preferred embodiment of the method, the step of predicting the triplet to be tested according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result specifically includes:
inputting the triplet to be tested into a training-completed dual-quaternion knowledge graph model and calculating the score of the triplet;
and obtaining a prediction result according to the score, and judging whether the triplet is a fact or not.
As shown in fig. 2, a dual quaternion-based knowledge graph link prediction system includes the following modules:
the data loading module is used for loading data and analyzing the data to obtain triplet data of the knowledge graph;
the model training module is used for training and parameter adjustment of a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph to obtain a trained dual quaternion knowledge graph model;
and the link prediction module is used for predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result.
Further as a preferred embodiment of the system, the model training module further comprises the following modules:
and the dual quaternion knowledge graph embedding module is used for processing the triplet data of the knowledge graph to obtain entity and relationship indexes, and initializing the entity and relationship in the knowledge graph based on quaternion vectors.
The scoring function module is used for constructing a scoring function of the triples and obtaining scores of the triples;
the negative sampling module is used for generating a negative sample according to the triplet data of the knowledge graph;
and the loss function module is used for constructing a loss function according to the triplet data and the negative sample of the knowledge graph and carrying out parameter adjustment on the score function based on the loss function.
The content in the system embodiment is applicable to the method embodiment, the functions specifically realized by the method embodiment are the same as those of the system embodiment, and the achieved beneficial effects are the same as those of the system embodiment.
While the preferred embodiment of the present application has been described in detail, the application is not limited to the embodiment, and various equivalent modifications and substitutions can be made by those skilled in the art without departing from the spirit of the application, and these equivalent modifications and substitutions are intended to be included in the scope of the present application as defined in the appended claims.
Claims (6)
1. The knowledge graph link prediction method based on dual quaternion is characterized by comprising the following steps of:
loading data and analyzing the data to obtain triplet data of the knowledge graph;
training and parameter adjustment are carried out on a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph, so that a trained dual quaternion knowledge graph model is obtained;
predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result;
training and parameter adjustment are carried out on a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph to obtain a trained dual quaternion knowledge graph model, and the method specifically comprises the following steps:
processing the triplet data of the knowledge graph to obtain an entity and a relation index;
initializing entities and relations in the knowledge graph based on the quaternion vector;
constructing a score function of the triplet;
generating a negative sample according to the triplet data of the knowledge graph;
constructing a loss function according to the triplet data and the negative sample of the knowledge graph, and carrying out parameter adjustment on the score function based on the loss function to obtain the training-completed dual-quaternion knowledge graph model parameters;
the scoring function expression is as follows:
in the above, theRepresenting relationships in knowledge-graph initialized with unit dual quaternion vectors, dual quaternions,/->Representing the conjugation of dual quaternions, +.>Representing dual quaternion multiplication, < >>Representing the initialization of entities in a knowledge graph with pure quaternion vectors, the E being a dual unit and satisfying the E 2 ＝0；
The loss function expression is as follows:
in the above formula, gamma is a super parameter, and f is r (h, t) represents a positive sample score, f r (h′ i ,t′ i ) Representing a negative sample score.
2. The dual-quaternion-based knowledge graph link prediction method according to claim 1, wherein the preset dual-quaternion knowledge graph model comprises a dual-quaternion knowledge graph embedding module, a scoring function module, a negative sampling module and a loss function module.
3. The method for predicting a link of a knowledge graph based on dual quaternions according to claim 2, wherein the step of initializing entities and relationships in the knowledge graph based on quaternion vectors specifically further comprises:
initializing an entity in the knowledge graph by using a pure quaternion vector;
and initializing the relation in the knowledge graph by using a unit dual quaternion vector.
4. The method for predicting the link of the knowledge-graph based on the dual quaternion according to claim 3, wherein the step of generating the negative sample according to the triplet data of the knowledge-graph comprises the following steps:
for a correct triplet, randomly replacing the head and tail entities in the triplet forms a negative sample.
5. The method for predicting the link of the knowledge graph based on the dual quaternion according to claim 4, wherein the step of predicting the triplet to be tested according to the training completed knowledge graph model of the dual quaternion to obtain the prediction result specifically comprises the following steps:
inputting the triplet to be tested into a training-completed dual-quaternion knowledge graph model and calculating the score of the triplet;
and obtaining a prediction result according to the score, and judging whether the triplet is a fact or not.
6. The knowledge graph link prediction system based on the dual quaternion is characterized by comprising the following modules:
the data loading module is used for loading data and analyzing the data to obtain triplet data of the knowledge graph;
the model training module is used for training and parameter adjustment of a preset dual quaternion knowledge graph model according to the triplet data of the knowledge graph to obtain a trained dual quaternion knowledge graph model;
the link prediction module is used for predicting the to-be-detected triplet according to the training-completed dual-quaternion knowledge graph model to obtain a prediction result;
the model training module further comprises the following modules:
the dual quaternion knowledge graph embedding module is used for processing the triplet data of the knowledge graph to obtain entity and relationship indexes, and initializing the entity and relationship in the knowledge graph based on quaternion vectors;
the scoring function module is used for constructing a scoring function of the triples and obtaining scores of the triples;
the negative sampling module is used for generating a negative sample according to the triplet data of the knowledge graph;
the loss function module is used for constructing a loss function according to the triplet data and the negative sample of the knowledge graph and carrying out parameter adjustment on the score function based on the loss function;
the scoring function expression is as follows:
in the above, theRepresenting relationships in knowledge-graph initialized with unit dual quaternion vectors, dual quaternions,/->Representing the conjugation of dual quaternions, +.>Representing dual quaternion multiplication, < >>Representing the initialization of entities in a knowledge graph with pure quaternion vectors, the E being a dual unit and satisfying the E 2 ＝0；
The loss function expression is as follows:
in the above formula, gamma is a super parameter, and f is r (h, t) represents a positive sample score, f r (h′ i ,t′ i ) Representing a negative sample score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011346008.5A CN112434812B (en) | 2020-11-26 | 2020-11-26 | Knowledge graph link prediction method and system based on dual quaternion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011346008.5A CN112434812B (en) | 2020-11-26 | 2020-11-26 | Knowledge graph link prediction method and system based on dual quaternion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112434812A CN112434812A (en) | 2021-03-02 |
| CN112434812B true CN112434812B (en) | 2023-09-05 |
Family
ID=74699202
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011346008.5A Active CN112434812B (en) | 2020-11-26 | 2020-11-26 | Knowledge graph link prediction method and system based on dual quaternion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112434812B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117495538B (en) * | 2023-10-30 | 2024-08-13 | 北京融和云链科技有限公司 | Risk assessment method and model training method for order financing |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105824802A (en) * | 2016-03-31 | 2016-08-03 | 清华大学 | Method and device for acquiring knowledge graph vectoring expression |
| CN106295796A (en) * | 2016-07-22 | 2017-01-04 | 浙江大学 | Entity link method based on degree of depth study |
| CN106528609A (en) * | 2016-09-28 | 2017-03-22 | 厦门理工学院 | Vector constraint embedded transformation knowledge graph inference method |
| CN109840282A (en) * | 2019-03-01 | 2019-06-04 | 东北大学 | A kind of knowledge mapping optimization method based on fuzzy theory |
| CN109992670A (en) * | 2019-04-04 | 2019-07-09 | 西安交通大学 | A kind of map completion method of knowledge based map neighbour structure |
| CN111680109A (en) * | 2020-04-22 | 2020-09-18 | 北京三快在线科技有限公司 | Knowledge graph representation learning model training method and device and electronic equipment |
| CN111914094A (en) * | 2019-05-10 | 2020-11-10 | 中国人民大学 | Knowledge graph representation learning method based on ternary interaction |
-
2020
- 2020-11-26 CN CN202011346008.5A patent/CN112434812B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105824802A (en) * | 2016-03-31 | 2016-08-03 | 清华大学 | Method and device for acquiring knowledge graph vectoring expression |
| CN106295796A (en) * | 2016-07-22 | 2017-01-04 | 浙江大学 | Entity link method based on degree of depth study |
| CN106528609A (en) * | 2016-09-28 | 2017-03-22 | 厦门理工学院 | Vector constraint embedded transformation knowledge graph inference method |
| CN109840282A (en) * | 2019-03-01 | 2019-06-04 | 东北大学 | A kind of knowledge mapping optimization method based on fuzzy theory |
| CN109992670A (en) * | 2019-04-04 | 2019-07-09 | 西安交通大学 | A kind of map completion method of knowledge based map neighbour structure |
| CN111914094A (en) * | 2019-05-10 | 2020-11-10 | 中国人民大学 | Knowledge graph representation learning method based on ternary interaction |
| CN111680109A (en) * | 2020-04-22 | 2020-09-18 | 北京三快在线科技有限公司 | Knowledge graph representation learning model training method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112434812A (en) | 2021-03-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Uncertainty-aware blind image quality assessment in the laboratory and wild | |
| Zhang et al. | Hierarchical feature fusion with mixed convolution attention for single image dehazing | |
| Shrivastava et al. | Learning from simulated and unsupervised images through adversarial training | |
| Zhao et al. | Alike: Accurate and lightweight keypoint detection and descriptor extraction | |
| CN109919209B (en) | Domain self-adaptive deep learning method and readable storage medium | |
| WO2020228525A1 (en) | Place recognition method and apparatus, model training method and apparatus for place recognition, and electronic device | |
| CN112446423B (en) | Fast hybrid high-order attention domain confrontation network method based on transfer learning | |
| CN110647991B (en) | Three-dimensional human body posture estimation method based on unsupervised field self-adaption | |
| CN112115967B (en) | Image increment learning method based on data protection | |
| CN109949255A (en) | Image rebuilding method and equipment | |
| Mahapatra et al. | Controllable animation of fluid elements in still images | |
| CN110728219A (en) | 3D face generation method based on multi-column multi-scale graph convolution neural network | |
| TW202032423A (en) | Method for image processing and apparatus thereof | |
| US20230245351A1 (en) | Image style conversion method and apparatus, electronic device, and storage medium | |
| CN112232325B (en) | Sample data processing method and device, storage medium and electronic equipment | |
| WO2022166797A1 (en) | Image generation model training method, generation method, apparatus, and device | |
| CN112434812B (en) | Knowledge graph link prediction method and system based on dual quaternion | |
| CN117218300B (en) | Three-dimensional model construction method, three-dimensional model construction training method and device | |
| CN107729885B (en) | Face enhancement method based on multiple residual error learning | |
| Liang et al. | Video super-resolution reconstruction based on deep learning and spatio-temporal feature self-similarity | |
| Xia et al. | Meta-learning-based degradation representation for blind super-resolution | |
| CN112597847A (en) | Face pose estimation method and device, electronic equipment and storage medium | |
| CN112417991A (en) | Double-attention face alignment method based on hourglass capsule network | |
| CN115131605A (en) | Structure perception graph comparison learning method based on self-adaptive sub-graph | |
| CN115063847A (en) | Training method and device for facial image acquisition model |
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 |