CN114691890A - QR decomposition-based time sequence knowledge graph completion method and device and electronic equipment - Google Patents
QR decomposition-based time sequence knowledge graph completion method and device and electronic equipment Download PDFInfo
- Publication number
- CN114691890A CN114691890A CN202210406181.2A CN202210406181A CN114691890A CN 114691890 A CN114691890 A CN 114691890A CN 202210406181 A CN202210406181 A CN 202210406181A CN 114691890 A CN114691890 A CN 114691890A
- Authority
- CN
- China
- Prior art keywords
- entity
- initial
- knowledge graph
- real
- timestamp
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000014509 gene expression Effects 0.000 claims abstract description 30
- 239000011159 matrix material Substances 0.000 claims description 120
- 230000006870 function Effects 0.000 claims description 41
- 230000004927 fusion Effects 0.000 claims description 22
- 238000011156 evaluation Methods 0.000 claims description 9
- 230000009469 supplementation Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 7
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 claims description 4
- 230000002123 temporal effect Effects 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 6
- 238000011161 development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 239000000969 carrier Substances 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 241001417539 Liza Species 0.000 description 1
- 238000010225 co-occurrence analysis Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to a time sequence knowledge graph completion method and device based on QR decomposition and electronic equipment. The method comprises the steps of obtaining initial embedded expressions of a head entity, a relation, a tail entity and a time stamp in a time sequence knowledge graph, wherein the initial embedded expressions are in a complex form; performing QR decomposition on a real part and an imaginary part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization respectively to obtain two orthogonal matrixes corresponding to the real part and the imaginary part of the timestamp, fusing the initial embedded representations of the head entity, the relation and the tail entity according to the two orthogonal matrixes to obtain the fused time embedded representation of the head entity, the relation and the tail entity, calculating the fraction of knowledge in the time sequence knowledge graph to be complemented according to the fused time embedded representation, predicting the missing content in the time sequence knowledge graph to be complemented through the fraction, and completing the time sequence knowledge graph complementation. The method can accurately and quickly construct the dynamic knowledge graph, and has better practical significance.
Description
Technical Field
The present application relates to the field of timing knowledge graph technologies, and in particular, to a method and an apparatus for complementing a timing knowledge graph based on QR decomposition, and an electronic device.
Background
Knowledge map (Knowledge Graph) is a series of different graphs displaying Knowledge development process and structure relationship in the book intelligence field, describing Knowledge resources and carriers thereof by using visualization technology, mining, analyzing, constructing, drawing and displaying Knowledge and mutual relation between Knowledge resources and Knowledge carriers. The knowledge graph is a modern theory which achieves the aim of multi-discipline fusion by combining theories and methods of applying subjects such as mathematics, graphics, information visualization technology, information science and the like with methods such as metrology introduction analysis, co-occurrence analysis and the like and utilizing a visualized graph to vividly display the core structure, development history, frontier field and overall knowledge framework of the subjects.
The knowledge map displays the complex knowledge field through data mining, information processing, knowledge measurement and graph drawing, reveals the dynamic development rule of the knowledge field, and provides a practical and valuable reference for subject research. Most of current knowledge graph researches concern static knowledge graphs which cannot change along with time, and the time-varying knowledge graphs are less explored. But timing information is very important because many structured knowledge is only valid within a specific interval, and the change in fact follows a time series. Recent research has begun to integrate timing information into knowledge-graph learning and knowledge-graph completion, referred to as timing knowledge graphs.
The knowledge graph completion algorithm can enable the knowledge graph to be more complete, and particularly the knowledge graph is generally constructed manually or is directly learned through representation learning. Although the features constructed by manual intervention usually have better interpretability but obviously consume too much resources, the new representation is automatically learned from data through representation learning, namely through a machine learning algorithm, and although the features can be automatically constructed according to tasks under less manual intervention, the interpretability of the constructed features is usually poor, and recently, methods for combining the two construction modes are provided, and a priori knowledge (such as rules, entity types, multi-hop paths and the like) is fused to the representation learning. Therefore, for the incomplete knowledge graph obtained manually or semi-automatically, particularly for the sparse implicit relationship or the implicit relationship that a plurality of relationships are not mined, if a method is available, completion can be carried out to a certain extent, and the knowledge graph can be more complete.
Knowledge Graph Completion (KGC) is currently mainly abstracted as a prediction problem, i.e. predicting missing parts in triples. So it can be divided into 3 subtasks: head entity prediction; predicting the relationship; and predicting a tail entity. Meanwhile, the knowledge graph completion algorithm can be divided into two types according to whether a new entity or a new relationship can be processed or not: static knowledge graph completion (Static KGC), which is used to complete implicit relationships between known entities. Only the fixed scenes of the entities and the relations can be processed, so the expansibility is poor; dynamic knowledge-graph completion (Dynamic KGC), which relates to more than entities or relations in the knowledge-graph G, namely some words which do not appear, or in the scene that the knowledge-graph is needed to be completed in the later period.
The difficulty of constructing a dynamic knowledge graph is high, and the speed and the accuracy of the conventional time sequence knowledge graph completion method are low.
Disclosure of Invention
In view of the above, it is desirable to provide a method, an apparatus and an electronic device for supplementing a time-series knowledge graph based on QR decomposition.
A QR decomposition-based time-series knowledge graph completion method, the method comprising:
an initial embedded representation of a head entity, a relationship, a tail entity, a timestamp in a time-series knowledge graph is obtained, the initial embedded representation being represented by a complex number including a real part and an imaginary part.
And performing QR decomposition on the real part and the imaginary part of the initial embedding representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedding representation of the timestamp.
And fusing the real part orthogonal matrix and the imaginary part orthogonal matrix with the initial embedded representation of the head entity, the relation and the tail entity respectively to obtain the fused time embedded representation of the head entity, the relation and the tail entity.
And calculating the scores of knowledge in the time sequence knowledge graph to be supplemented according to the fusion time embedded expression of the head entity, the relation and the tail entity and a preset scoring function, predicting the missing content in the time sequence knowledge graph to be supplemented according to the scores and a preset loss function, and completing the time sequence knowledge graph supplementation.
In one embodiment, the initial embedded representation is represented by a complex number, including a real part and an imaginary part.
Obtaining initial embedded representations of head entities, relationships, tail entities, timestamps in a time series knowledge graph, comprising:
and (4) performing random initialization on the head entity, the relation, the tail entity and the timestamp in the time sequence knowledge graph to obtain initial embedded representation of the head entity, the relation, the tail entity and the timestamp band.
Classifying the head entity and the tail entity as entities to generate an initial embedding matrix of the entities, and marking the initial embedding matrix as E|ε|×kGenerating an initial embedding matrix of relationships, denoted asGenerating an initial embedding matrix of time stamps, denotedWhere k is the dimension of the three initial embedded matrices, | ε |,andrespectively the number of entities, relationships and timestamps.
In one embodiment, the initial embedding of the timestamp is represented by:
eτ=eτ_real+eτ_imgi
wherein e isτFor an initial embedded representation of the time stamp, eτ_realIs the real part of the initial embedded representation of the timestamp, eτ_imgIs the imaginary part of the initial embedded representation of the timestamp, i being the imaginary unit.
Performing QR decomposition on the real part and the imaginary part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedded representation of the timestamp, wherein the QR decomposition comprises the following steps:
performing QR decomposition on the real part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix of the initial embedded representation of the timestamp; the expression of the real part QR decomposition is as follows:
eτ_real=Qτ_realRτ_real
wherein Q isτ_realOrthogonal matrix of real parts, R, being an initial embedded representation of the time stampτ_realA real R matrix of an initial embedded representation of the timestamp;
performing QR decomposition on the imaginary part of the initial embedding representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain an imaginary part orthogonal matrix of the initial embedding representation of the timestamp; the expression of the imaginary part QR decomposition is:
eτ_img=Qτ_imgRτ_img
wherein Qτ_imgAn imaginary orthogonal matrix, R, being an initial embedded representation of the time stampτ_imgAn imaginary R matrix of an imaginary orthogonal matrix of the initial embedded representation of the timestamp;
in one embodiment, fusing the real orthogonal matrix and the imaginary orthogonal matrix with the initial embedded representations of the head entity, the relationship, and the tail entity, respectively, to obtain a fused time embedded representation of the head entity, the relationship, and the tail entity, includes:
taking the real part orthogonal matrix and the imaginary part orthogonal matrix as a real part and an imaginary part respectively to form an orthogonal matrix about the time stamp; the expression for the orthogonal matrix of time stamps is:
Qτ=Qτ_real+Qτ_imgi
wherein QτAs an orthogonal matrix with respect to time stamps, Qτ_realIs an orthogonal matrix of real parts, Qτ_imgIs an imaginary orthogonal matrix, i is an imaginary unit.
And multiplying the orthogonal matrix related to the timestamp with the initial embedded representation of the head entity, the relation and the tail entity respectively to obtain the fusion time embedded representation of the head entity, the relation and the tail entity.
In one embodiment, the method for completing the time sequence knowledge graph completion includes the steps of calculating scores of knowledge in the time sequence knowledge graph to be completed according to fusion time embedded expression of a head entity, a relation and a tail entity and a preset scoring function, predicting missing content in the time sequence knowledge graph to be completed according to the scores and a preset loss function, and completing the time sequence knowledge graph completion, and includes the steps of:
obtaining a test set comprising a plurality of valid triples, the triples comprising: head entity, relationship, tail entity;
and forming an entity set by the head entity and the tail entity of all the triples.
Replacing the head entity or the tail entity of each triple with an entity set except the head entity or the tail entity contained in the triple to obtain a damaged triple, comparing all the damaged triples with the existing triples of the knowledge graph, deleting the damaged triples existing in the knowledge graph, and forming the remaining damaged triples into a negative example set.
And calculating the scores of the effective triples and the damaged triples according to a preset scoring function, and arranging the scores in a descending order.
And predicting missing contents in the time sequence knowledge graph to be supplemented according to the scores, the preset loss function and the preset evaluation index, and completing the time sequence knowledge graph supplementation.
In one embodiment, the scores of the effective triples and the damaged triples are calculated according to a preset scoring function, and the scores are arranged in the order from small to large, wherein the preset scoring function in the step is as follows:
wherein f (h, r, t) is a fraction, hτFor fused time-embedded representation of the head entity, rτIs a head entity hτAnd tail entity tτIn the context of (a) or (b),is a tail entity tτIs given by a time stamp, | | | | | non-woven phosphor1/2To obtain the L1 or L2 norm.
In one embodiment, the predetermined evaluation index is that the loss of the knowledge graph reaches a predetermined threshold, and the preset loss function expression is:
wherein, T+Is a collection of all valid triples; t is a unit of-Is a set of negative examples; f (h, r, t) is the fraction of valid triples; f (h ', r, t') is the fraction of the damage triplet; h and h' are head entities; t and t' are tail entities, and r is a relationship; gamma is a spacing hyperparameter, gamma>0;[]+Max (0, γ + f (h, r, t) -f (h ', r, t')).
A QR decomposition-based temporal knowledge graph complementation apparatus, the apparatus comprising:
and the initial embedded representation acquisition module is used for acquiring initial embedded representations of the head entity, the relation, the tail entity and the timestamp in the time sequence knowledge graph, wherein the initial embedded representations are expressed by complex numbers and comprise real parts and imaginary parts.
And the QR decomposition module is used for carrying out QR decomposition on the real part and the imaginary part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedded representation of the timestamp.
And the fusion time embedded representation determining module is used for fusing the real part orthogonal matrix and the imaginary part orthogonal matrix with the initial embedded representations of the head entity, the relation and the tail entity respectively to obtain the fusion time embedded representations of the head entity, the relation and the tail entity.
And the knowledge graph completion module is used for calculating the scores of knowledge in the time sequence knowledge graph to be completed according to the fusion time embedded expression of the head entity, the relation and the tail entity and a preset scoring function, predicting the missing content in the time sequence knowledge graph to be completed according to the scores and a preset loss function, and completing the time sequence knowledge graph completion.
The QR decomposition-based time sequence knowledge graph completion method, the QR decomposition-based time sequence knowledge graph completion device and the electronic equipment are characterized in that initial embedded expressions of head entities, relations, tail entities and time stamps in the time sequence knowledge graph are obtained, and the initial embedded expressions are expressed in a complex form; performing QR decomposition on a real part and an imaginary part of the initial embedded representation of the timestamp by Gram-Schmidt orthogonalization respectively to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix of the timestamp, fusing the real part orthogonal matrix and the imaginary part orthogonal matrix of the timestamp, and the initial embedded representations of a head entity, a relation and a tail entity to obtain a fused time embedded representation of the head entity, the relation and the tail entity, calculating the fraction of knowledge in the time sequence knowledge graph to be supplemented according to the fused time embedded representation of the head entity, the relation and the tail entity and a preset scoring function, predicting the missing content in the time sequence knowledge graph to be supplemented according to the fraction and the preset loss function, and completing the time sequence knowledge graph supplementation. The method can accurately and quickly construct the dynamic knowledge map, and has better practical significance.
Drawings
FIG. 1 is a schematic flow chart diagram illustrating a QR decomposition-based temporal knowledge graph completion method in one embodiment;
FIG. 2 is a schematic flow chart diagram illustrating the completion step of the temporal knowledge graph in one embodiment;
FIG. 3 is a block diagram of a QR decomposition-based sequential knowledge-graph completion apparatus in one embodiment;
FIG. 4 is a diagram of the internal structure of an electronic device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The QR decomposition-based time sequence knowledge graph completion method is applied to the aspect of information mining, and information analysis is carried out by utilizing the network figure relation graph.
In one embodiment, as shown in fig. 1, there is provided a QR decomposition-based time-series knowledge-graph completion method, which includes the following steps:
step 100: an initial embedded representation of a head entity, a relationship, a tail entity, a timestamp in a time-series knowledge graph is obtained, the initial embedded representation being represented by complex numbers including real and imaginary parts.
Specifically, the timing knowledge graph includes a plurality of quadruples consisting of head entities, relationships, tail entities and timestamps. The quadruplet in the time sequence knowledge graph is extracted from a plurality of facts described by a telecommunication case character relation graph, for example, the description of the telecommunication case character relation graph comprises the following three facts, namely, a loan data of a loan client is purchased to an organization in 10, 11 and 10 months in 2020, wherein the loan data comprises king of a loan data; zhang a person hires a Liji for his work from 1 month in 2019 to 12 months in 2021; li's one induces Wang' to pay the 'agency service fee' by telephone in 10, 30 months in 2020. The above three facts are described as a quadruplet (Zhangyi, Purchase, Wangzhi, 2020/10/11); (Zhang-Kao, hire, Liza, 2019/1-2021/12); (Lieqi, decoy, Wangzao, 2020/10/30). And connecting the four-tuples by using nodes and edges to form a time sequence knowledge graph, wherein the nodes represent head entities or tail entities in the four-tuples, and the edges represent relations and time pairs in the four-tuples.
A relationship is a relationship between a head entity and a tail entity.
Obtaining corresponding initial embedded representation e by using random initialization for head entity, relation, tail entity and time stamp in time sequence knowledge graphh,er,etAnd eτ(ii) a Initial embedded representations of the head entity, the relationships, the tail entity, and the timestamps are obtained using the TransE method. Accordingly, ehAnd etInitial embedding of a head entity and a tail entity respectively, classifying the head entity and the tail entity into entities, generating an entity initial embedding matrix, and marking as E|ε|×kGenerating a relational initial embedding matrix, and recordingGenerating an initial embedding matrix of time stamps, denotedWhere k is the initial embedding dimension, where it is assumed that the three initial embedding dimensions are the same; the number of entities, relationships, and timestamps are referred to as epsilon,andand (4) showing. The embedded representations are all expressed in complex numbers, and the dimensions of the real part and the imaginary part are consistent. Initial embedded representation e of head entity, relationship, tail entity, timestamph,er, etAnd eτThe expression of (a) is:
eh=eh_real+eh_imgi
er=er_real+er_imgi
et=et_real+et_imgi
eτ=eτ_real+eτ_imgi
wherein,eh_realAnd eh_imgRespectively the real and imaginary parts, e, of the initial embedded representation of the head entityr_realAnd er_imgRespectively the real and imaginary parts of the initial embedded representation of the relation, et_realAnd et_imgRespectively the real and imaginary parts of the initial embedded representation of the real tail body, i being the imaginary unit.
Step 102: and performing QR decomposition on the real part and the imaginary part of the initial embedding representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedding representation of the timestamp.
Step 104: and respectively fusing the real part orthogonal matrix and the imaginary part orthogonal matrix with the initial embedded representation of the head entity, the relation and the tail entity to obtain the fusion time embedded representation of the head entity, the relation and the tail entity.
Step 106: and calculating the scores of knowledge in the time sequence knowledge graph to be supplemented according to the fusion time embedded expression of the head entity, the relation and the tail entity and a preset scoring function, predicting the missing content in the time sequence knowledge graph to be supplemented according to the scores and a preset loss function, and completing the time sequence knowledge graph supplementation.
In the QR decomposition-based time sequence knowledge graph completion method, initial embedded expressions of a head entity, a relation, a tail entity and a time stamp in the time sequence knowledge graph are obtained, and the initial embedded expressions are expressed in a complex form; performing QR decomposition on a real part and an imaginary part of the initial embedded representation of the timestamp by Gram-Schmidt orthogonalization respectively to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix of the timestamp, fusing the real part orthogonal matrix and the imaginary part orthogonal matrix of the timestamp, and the initial embedded representations of a head entity, a relation and a tail entity to obtain a fused time embedded representation of the head entity, the relation and the tail entity, calculating the fraction of knowledge in the time sequence knowledge graph to be supplemented according to the fused time embedded representation of the head entity, the relation and the tail entity and a preset scoring function, predicting the missing content in the time sequence knowledge graph to be supplemented according to the fraction and the preset loss function, and completing the time sequence knowledge graph supplementation. The method can accurately and quickly construct the dynamic knowledge graph, and has better practical significance.
In one embodiment, step 100 comprises: using random initialization for head entities, relations, tail entities and timestamps in a time sequence knowledge graph to obtain initial embedded representations of the head entities, the relations, the tail entities and timestamp bands; classifying the head entity and the tail entity into entities to generate an initial embedding matrix of the entities, and marking the initial embedding matrix as E|ε|×kGenerating an initial embedding matrix of relationships, denoted asGenerating an initial embedding matrix of time stamps, denotedWhere k is the dimension of the three initial embedding matrices, | ε |,andthe number of entities, relationships and timestamps, respectively;
in one embodiment, the initial embedding of the timestamp is represented as:
eτ=eτ_real+eτ_imgi
wherein e isτFor an initial embedded representation of the time stamp, eτ_realIs the real part of the initial embedded representation of the timestamp, eτ_imgIs the imaginary part of the initial embedded representation of the timestamp, i is the imaginary unit; step 102 comprises: performing QR decomposition on the real part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix of the initial embedded representation of the timestamp; the expression of the real part QR decomposition is as follows:
eτ_real=Qτ_realRτ_real
wherein Q isτ_realOrthogonal matrix of real parts, R, being an initial embedded representation of the time stampτ_realReal part of initial embedded representation of time stampAnd (4) an R matrix.
Performing QR decomposition on the imaginary part of the initial embedding representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain an imaginary part orthogonal matrix of the initial embedding representation of the timestamp; the expression of the imaginary part QR decomposition is:
eτ_img=Qτ_imgRτ_img
wherein Q isτ_imgAn imaginary orthogonal matrix, R, being an initial embedded representation of the time stampτ_imgAn imaginary R matrix of an imaginary orthogonal matrix of the initial embedded representation of the timestamp.
In one embodiment, step 104 includes: respectively taking the real part orthogonal matrix and the imaginary part orthogonal matrix as a real part and an imaginary part to form an orthogonal matrix about the time stamp; the expression for the orthogonal matrix of time stamps is:
Qτ=Qτ_real+Qτ_imgi
wherein Q isτBeing an orthogonal matrix with respect to time stamp, Qτ_realIs an orthogonal matrix of real parts, Qτ_imgIs an imaginary orthogonal matrix, i is an imaginary unit.
And multiplying the orthogonal matrix related to the timestamp with the initial embedded representation of the head entity, the relation and the tail entity respectively to obtain the fusion time embedded representation of the head entity, the relation and the tail entity.
Specifically, a complex phase multiplication formula is used:
z1z2=(a+bi)(c+di)=ac+adi+bci+bdi2=(ac-bd)+(ad+bc)i
associating the orthogonal matrix with the time stamp with the initial embedded representation e of the initial embedded representation of the head entity, the relation, the tail entity, respectivelyh,erAnd etFusing to obtain the embedded expression h of the fusion time of the head entity, the relation and the tail entityτ,rτ,tτ:
hτ=ehQτ=(eh_realQτ_real-eh_imgQτ_img)+(eh_realQτ_img+eh_imgQτ_real)i
rτ=erQτ=(er_realQτ_real-er_imgQτ_img)+(er_realQτ_img+er_imgQτ_real)i
tτ=etQτ=(et_realQτ_real-et_imgQτ_img)+(et_realQτ_img+et_imgQτ_real)i
In one embodiment, step 106 specifically includes:
step 200: obtaining a test set comprising a plurality of valid triplets, a triplet comprising: head entity, relationship, tail entity.
Step 202: and forming an entity set by the head entity and the tail entity of all the triples.
Step 204: replacing the head entity or the tail entity of each triple with an entity set except the head entity or the tail entity contained in the triple to obtain a damaged triple, comparing all the damaged triples with the existing triples of the knowledge graph, deleting the damaged triples existing in the knowledge graph, and forming the remaining damaged triples into a negative example set.
Step 206: calculating the scores of the effective triples and the damaged triples according to a preset scoring function, and arranging the scores in a descending order; and predicting missing contents in the time sequence knowledge graph to be supplemented according to the scores, the preset loss function and the preset evaluation index, and completing the time sequence knowledge graph supplementation.
In one embodiment, the predetermined scoring function in step 206 is:
wherein f (h, r, t) is a fraction, hτFor fused time-embedded representation of the head entity, rτIs a head entity hτAnd tail entity tτIn the context of (a) or (b),is a tail entity tτIs given by a time stamp, | | | | | non-woven phosphor1/2To find the L1 or L2 norm.
In one embodiment, the predetermined evaluation index is that the loss of the knowledge-graph reaches a predetermined threshold, and the preset loss function expression is:
wherein, T+Is a collection of all valid triples; t is-Is a set of negative examples; f (h, r, t) is the fraction of valid triples; f (h ', r, t') is the fraction of damaged triples; h and h' are head entities; t and t' are tail entities, and r is a relationship; gamma is a spacing hyperparameter, gamma>0;[]+Max (0, γ + f (h, r, t) -f (h ', r, t')).
γ + f (h, r, t) -f (h ', r, t') it should be understood that, although the various steps in the flow charts of FIGS. 1-2 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps of fig. 1-2 may include multiple sub-steps or phases that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or phases may not necessarily be sequential, but may be performed in turn or alternating with other steps or at least a portion of the sub-steps or phases of other steps.
In one embodiment, as shown in fig. 3, there is provided a QR decomposition-based time-series knowledge-graph complementing apparatus, including: an initial embedded representation acquisition module, a QR decomposition module of an initial embedded representation of a timestamp, a fusion time embedded representation determination module, and a knowledge graph completion module, wherein:
the initial embedded representation acquisition module is used for acquiring initial embedded representations of a head entity, a relation, a tail entity and a time stamp in the time sequence knowledge graph, wherein the initial embedded representations are expressed by complex numbers and comprise a real part and an imaginary part;
the QR decomposition module is used for carrying out QR decomposition on the real part and the imaginary part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedded representation of the timestamp;
a fusion time embedded representation determining module, configured to fuse the real part orthogonal matrix and the imaginary part orthogonal matrix with the initial embedded representation of the head entity, the relationship, and the tail entity, respectively, to obtain a fusion time embedded representation of the head entity, the relationship, and the tail entity;
and the knowledge graph completion module is used for calculating the scores of knowledge in the time sequence knowledge graph to be completed according to the fusion time embedded expression of the head entity, the relation and the tail entity and a preset scoring function, predicting the missing content in the time sequence knowledge graph to be completed according to the scores and a preset loss function, and completing the time sequence knowledge graph completion.
In one embodiment, the initial embedded representation obtaining module is further configured to use random initialization for a head entity, a relationship, a tail entity, and a timestamp in the time-series knowledge graph to obtain an initial embedded representation of the head entity, the relationship, the tail entity, and a timestamp zone; classifying the head entity and the tail entity as entities to generate an initial embedding matrix of the entities, and marking the initial embedding matrix as E|ε|×kGenerating an initial embedding matrix of relationships, denoted asGenerating an initial embedding matrix of time stamps, denotedWhere k is the dimension of the three initial embedded matrices, | ε |,andthe number of entities, relationships and timestamps, respectively;
in one embodiment, the initial embedding of the timestamp is represented as:
eτ=eτ_real+eτ_imgi
wherein e isτFor an initial embedded representation of the time stamp, eτ_realIs the real part of the initial embedded representation of the timestamp, eτ_imgIs the imaginary part of the initial embedded representation of the timestamp, i is the imaginary unit; the QR decomposition module of the initial embedded representation of the timestamp is also used for carrying out QR decomposition on the real part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix of the initial embedded representation of the timestamp; the expression of the real part QR decomposition is as follows:
eτ_real=Qτ_realRτ_real
wherein Q isτ_realOrthogonal matrix of real parts, R, being an initial embedded representation of the time stampτ_realThe initial embedded real part R matrix of the representation of the time stamp.
Performing QR decomposition on the imaginary part of the initial embedding representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain an imaginary part orthogonal matrix of the initial embedding representation of the timestamp; the expression of the imaginary part QR decomposition is:
eτ_img=Qτ_imgRτ_img
wherein Q isτ_imgAn imaginary orthogonal matrix, R, being an initial embedded representation of the time stampτ_imgAn imaginary R matrix of an imaginary orthogonal matrix of the initial embedded representation of the timestamp.
In one embodiment, the fused time-embedded representation determining module is further configured to form an orthogonal matrix with respect to the time stamp by using the real orthogonal matrix and the imaginary orthogonal matrix as a real part and an imaginary part, respectively; the expression for the orthogonal matrix of time stamps is:
Qτ=Qτ_real+Qτ_imgi
wherein Q isτBeing an orthogonal matrix with respect to time stamp, Qτ_realIs an orthogonal matrix of real parts, Qτ_imgIs an imaginary orthogonal matrix, i is an imaginary unit.
And multiplying the orthogonal matrix related to the timestamp with the initial embedded representation of the head entity, the relation and the tail entity respectively to obtain the fusion time embedded representation of the head entity, the relation and the tail entity.
In one embodiment, the knowledge-graph completion module is further configured to obtain a test set comprising a plurality of valid triples, the triples including: head entity, relationship, tail entity; forming an entity set by the head entity and the tail entity of all the triples; replacing the head entity or the tail entity of each triple with an entity set except the head entity or the tail entity contained in the triple to obtain a damaged triple, comparing all the damaged triples with the existing triples of the knowledge graph, deleting the damaged triples existing in the knowledge graph, and forming the remaining damaged triples into a negative example set; calculating the scores of the effective triples and the damaged triples according to a preset scoring function, and arranging the scores in a descending order; and predicting missing contents in the time sequence knowledge graph to be supplemented according to the scores, the preset loss function and the preset evaluation index, and completing the time sequence knowledge graph supplementation.
In one embodiment, the preset scoring function in the knowledge-graph completion module is:
wherein f (h, r, t) is a fraction, hτFor fused time-embedded representation of the head entity, rτIs a head entity hτAnd tail entity tτIn the context of (a) or (b),as tail entity tτIs given by a time stamp, | | | | | non-woven phosphor1/2To obtain the L1 or L2 norm.
In one embodiment, the predetermined evaluation index in the knowledge-graph completion module is that the loss of the knowledge-graph reaches a predetermined threshold, and the preset loss function expression is as follows:
wherein, T+Is a collection of all valid triples; t is-Is a set of negative examples; f (h, r, t) is the fraction of valid triples; f (h ', r, t') is the fraction of damaged triples; h and h' are head entities; t and t' are tail entities, and r is a relationship; gamma is a spacing hyperparameter, gamma>0;[]+For max (0, γ + f (h, r, t) -f (h ', r, t')), i.e.: when the value of γ + f (h, r, t) -f (h ', r, t') is greater than 0, the value of the predetermined loss function is γ + f (h, r, t) -f (h ', r, t'), and the value of γ + f (h, r, t) -f (h ', r, t') is less than or equal to 0, the value of the predetermined loss function is 0.
Specific limitations on the QR decomposition-based time-series knowledge graph completion apparatus can be referred to the above limitations on the QR decomposition-based time-series knowledge graph completion method, and details thereof are not repeated here. The modules in the QR decomposition-based time-series knowledge map complementing device can be wholly or partially realized by software, hardware and a combination thereof. The modules may be embedded in a hardware form or may be independent of a processor in the electronic device, or may be stored in a memory in the electronic device in a software form, so that the processor calls and executes operations corresponding to the modules.
In one embodiment, an electronic device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 4. The electronic device comprises a processor, a memory, a network interface, a display screen and an input device which are connected through a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic equipment comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the electronic device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement a QR decomposition-based sequential knowledge-graph completion method. The display screen of the electronic equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the electronic equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the configuration shown in fig. 4 is a block diagram of only a portion of the configuration associated with the present application, and does not constitute a limitation on the electronic device to which the present application is applied, and a particular electronic device may include more or less components than those shown in the drawings, or combine certain components, or have a different arrangement of components.
In one embodiment, an electronic device is provided, comprising a memory storing a computer program and a processor implementing the steps of the above-described method embodiments when executing the computer program.
All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (9)
1. A QR decomposition-based time-series knowledge graph completion method is characterized by comprising the following steps:
acquiring initial embedded representations of a head entity, a relation, a tail entity and a time stamp in a time sequence knowledge graph, wherein the initial embedded representations are expressed by complex numbers and comprise a real part and an imaginary part;
performing QR decomposition on the real part and the imaginary part of the initial embedding representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedding representation of the timestamp;
fusing the real part orthogonal matrix and the imaginary part orthogonal matrix with a head entity, a relation and an initial embedding representation of a tail entity respectively to obtain a fusion time embedding representation of the head entity, the relation and the tail entity;
and calculating the scores of knowledge in the time sequence knowledge graph to be supplemented according to the fusion time embedded expression of the head entity, the relation and the tail entity and a preset scoring function, and predicting the missing content in the time sequence knowledge graph to be supplemented according to the scores and a preset loss function to complete the time sequence knowledge graph supplementation.
2. The method of claim 1, wherein the initial embedded representation is represented by a complex number, comprising a real part and an imaginary part;
obtaining an initial embedded representation of head entities, relationships, tail entities, timestamps in a time series knowledge graph, comprising:
performing random initialization on a head entity, a relation, a tail entity and a timestamp in a time sequence knowledge graph to obtain initial embedded representation of the head entity, the relation, the tail entity and the timestamp;
classifying the head entity and the tail entity as entities to generate an initial embedding matrix of the entities, and marking the initial embedding matrix as E|ε|×kGenerating an initial embedding matrix of relationships, denoted asGenerating an initial embedding matrix of time stamps, denotedWhere k is the dimension of the three initial embedded matrices, | ε |,andrespectively the number of entities, relationships and timestamps.
3. The method of claim 1, wherein the initial embedding of the timestamp is represented by:
eτ=eτ_real+eτ_imgi
wherein e isτFor an initial embedded representation of the time stamp, eτ_realIs the real part of the initial embedded representation of the timestamp, eτ_imgIs the imaginary part of the initial embedded representation of the timestamp, i is the imaginary unit;
performing QR decomposition on the real part and the imaginary part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedded representation of the timestamp, wherein the QR decomposition comprises the following steps:
performing QR decomposition on the real part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix of the initial embedded representation of the timestamp; the expression of the real part QR decomposition is as follows:
eτ_real=Qτ_realRτ_real
wherein Q isτ_realOrthogonal matrix of real parts, R, being an initial embedded representation of the time stampτ_realA real R matrix of an initial embedded representation of the timestamp;
performing QR decomposition on the imaginary part of the initial embedding representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain an imaginary part orthogonal matrix of the initial embedding representation of the timestamp; the expression of the imaginary part QR decomposition is:
eτ_img=Qτ_imgRτ_img
wherein Q isτ_imgWhen isImaginary orthogonal matrix, R, of the initial embedded representation of the timestampτ_imgAn imaginary R matrix of an imaginary orthogonal matrix of the initial embedded representation of the timestamp.
4. The method of claim 1, wherein fusing the real orthogonal matrix and the imaginary orthogonal matrix with initial embedded representations of a head entity, a relationship, and a tail entity, respectively, to obtain a fused time embedded representation of the head entity, the relationship, and the tail entity, comprises:
taking the real part orthogonal matrix and the imaginary part orthogonal matrix as a real part and an imaginary part respectively to form an orthogonal matrix about the time stamp; the expression for the orthogonal matrix of time stamps is:
Qτ=Qτ_real+Qτ_imgi
wherein Q isτBeing an orthogonal matrix with respect to time stamp, Qτ_realIs an orthogonal matrix of real parts, Qτ_imgIs an imaginary orthogonal matrix, i is an imaginary unit;
and multiplying the orthogonal matrix related to the timestamp with the initial embedded representation of the head entity, the relation and the tail entity respectively to obtain the fusion time embedded representation of the head entity, the relation and the tail entity.
5. The method of claim 1, wherein the completion of the sequential knowledge graph is accomplished by calculating scores of knowledge in the sequential knowledge graph to be completed based on the fused time-embedded representation of the head entity, the relationship, and the tail entity and a predetermined scoring function, and predicting missing content in the sequential knowledge graph to be completed based on the scores and a predetermined loss function, comprising:
obtaining a test set comprising a plurality of valid triplets, the triplets comprising: head entity, relationship, tail entity;
forming an entity set by the head entity and the tail entity of all the triples;
replacing a head entity or a tail entity of each triple with an entity set except the head entity or the tail entity contained in the triple to obtain a damaged triple, comparing all the damaged triples with the existing triples of the knowledge graph, deleting the damaged triples existing in the knowledge graph, and forming the remaining damaged triples into a negative example set;
calculating the scores of the effective triples and the damaged triples according to a preset scoring function, and arranging the scores in a descending order;
and predicting missing contents in the time sequence knowledge graph to be supplemented according to the scores, the preset loss function and the preset evaluation index, and completing the time sequence knowledge graph supplementation.
6. The method according to claim 5, wherein the scores of the effective triples and the damaged triples are calculated according to a preset scoring function, and the scores are arranged in a descending order, wherein the preset scoring function in the step is as follows:
7. The method according to claim 6, characterized in that missing contents in the time sequence knowledge graph to be complemented are predicted according to the scores, the preset loss function and a preset evaluation index, so that the time sequence knowledge graph is complemented, wherein in the step, the preset evaluation index is that the loss of the knowledge graph reaches a preset threshold value; the default loss function is:
wherein, T+Is a set of all valid triples, T-Is T+F (h, r, t) is the fraction of valid triples, f (h ', r, t') is the fraction of corrupted triples, h and h 'are head entities, t and t' are tail entities, r is a relationship, γ>0 is interval ultra parameter [ alpha ], [ alpha ] is an]+Max (0, γ + f (h, r, t) -f (h ', r, t')).
8. A QR decomposition-based temporal knowledge map completion apparatus, the apparatus comprising:
an initial embedded representation obtaining module, configured to obtain an initial embedded representation of a head entity, a relationship, a tail entity, and a timestamp in a time-series knowledge graph, where the initial embedded representation is expressed by a complex number and includes a real part and an imaginary part;
the QR decomposition module is used for carrying out QR decomposition on the real part and the imaginary part of the initial embedded representation of the timestamp by adopting Gram-Schmidt orthogonalization to obtain a real part orthogonal matrix and an imaginary part orthogonal matrix corresponding to the real part and the imaginary part of the initial embedded representation of the timestamp;
a fusion time embedding representation determining module, configured to fuse the real part orthogonal matrix and the imaginary part orthogonal matrix with the initial embedding representation of the head entity, the relationship, and the tail entity, respectively, to obtain a fusion time embedding representation of the head entity, the relationship, and the tail entity;
and the knowledge graph completion module is used for calculating the scores of knowledge in the time sequence knowledge graph to be completed according to the fusion time embedded expression of the head entity, the relation and the tail entity and a preset scoring function, predicting the missing content in the time sequence knowledge graph to be completed according to the scores and a preset loss function, and completing the time sequence knowledge graph completion.
9. An electronic device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method according to any one of claims 1-7 when executing the computer program.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210406181.2A CN114691890B (en) | 2022-04-18 | Sequential knowledge graph completion method and device based on QR decomposition and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210406181.2A CN114691890B (en) | 2022-04-18 | Sequential knowledge graph completion method and device based on QR decomposition and electronic equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114691890A true CN114691890A (en) | 2022-07-01 |
CN114691890B CN114691890B (en) | 2024-09-06 |
Family
ID=
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116842958A (en) * | 2023-09-01 | 2023-10-03 | 北京邮电大学 | Time sequence knowledge graph completion method, entity prediction method based on time sequence knowledge graph completion method and device thereof |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113360664A (en) * | 2021-05-31 | 2021-09-07 | 电子科技大学 | Knowledge graph complementing method |
CN114238658A (en) * | 2021-12-21 | 2022-03-25 | 杭州峰景科技有限公司 | Link prediction method and device of time sequence knowledge graph and electronic equipment |
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113360664A (en) * | 2021-05-31 | 2021-09-07 | 电子科技大学 | Knowledge graph complementing method |
CN114238658A (en) * | 2021-12-21 | 2022-03-25 | 杭州峰景科技有限公司 | Link prediction method and device of time sequence knowledge graph and electronic equipment |
Non-Patent Citations (1)
Title |
---|
陈恒;李冠宇;祁瑞华;朱毅;郭旭;: "改进的Tucker分解知识图谱补全算法", 数学的实践与认识, no. 16, 23 August 2020 (2020-08-23) * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116842958A (en) * | 2023-09-01 | 2023-10-03 | 北京邮电大学 | Time sequence knowledge graph completion method, entity prediction method based on time sequence knowledge graph completion method and device thereof |
CN116842958B (en) * | 2023-09-01 | 2024-02-06 | 北京邮电大学 | Time sequence knowledge graph completion method, entity prediction method based on time sequence knowledge graph completion method and device thereof |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Schlegel et al. | Towards a rigorous evaluation of XAI methods on time series | |
EP3627759A1 (en) | Method and apparatus for encrypting data, method and apparatus for training machine learning model, and electronic device | |
US10282360B2 (en) | Uniform chart formatting based on semantics in data models | |
CN113157752B (en) | Scientific and technological resource recommendation method and system based on user portrait and situation | |
CN106530010A (en) | Collaborative filtering method and apparatus based on fusion of time factor | |
CN116823409B (en) | Intelligent screening method and system based on target search data | |
US8036922B2 (en) | Apparatus and computer-readable program for estimating man-hours for software tests | |
EL-Sagheer | Inferences in constant-partially accelerated life tests based on progressive type-II censoring | |
Ahamed et al. | Semantic segmentation of self-supervised dataset and medical images using combination of u-net and neural ordinary differential equations | |
Longo | Explainable Artificial Intelligence: First World Conference, xAI 2023, Lisbon, Portugal, July 26–28, 2023, Proceedings, Part II | |
Hesamian et al. | Exponentially weighted moving average control chart based on normal fuzzy random variables | |
Muschalik et al. | isage: An incremental version of SAGE for online explanation on data streams | |
CN114049185A (en) | Cost accounting method and device, computer equipment and storage medium | |
CN114691890A (en) | QR decomposition-based time sequence knowledge graph completion method and device and electronic equipment | |
Wang et al. | Inference on semiparametric transformation model with general interval-censored failure time data | |
CN114691890B (en) | Sequential knowledge graph completion method and device based on QR decomposition and electronic equipment | |
CN114358821B (en) | Commodity detail feature extraction method, commodity detail feature extraction device, computer equipment and storage medium | |
Yan et al. | Asymptotic analysis of a size-structured population model with infinite states-at-birth | |
JP2022067897A (en) | Information processing method and information processing program | |
CN113095047A (en) | Experiment report generation method and device, computer equipment and storage medium | |
CN110782287A (en) | Entity similarity calculation method and device, article recommendation system, medium and equipment | |
US20240289622A1 (en) | Ai model recommendation based on system task analysis and interaction data | |
Angus et al. | Tracking the exploration and exploitation in stochastic population-based nature-inspired algorithms using recurrence plots | |
US20240177045A1 (en) | Method for determining decoherence model of qubit and computer readable storage medium | |
Rajesh et al. | Estimation of shannon entropy in the presence of length-bias and type I censoring |
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 |