CN115859059B - Repeatable labeling method, system and device for fuzzy information - Google Patents
Repeatable labeling method, system and device for fuzzy information Download PDFInfo
- Publication number
- CN115859059B CN115859059B CN202211026598.2A CN202211026598A CN115859059B CN 115859059 B CN115859059 B CN 115859059B CN 202211026598 A CN202211026598 A CN 202211026598A CN 115859059 B CN115859059 B CN 115859059B
- Authority
- CN
- China
- Prior art keywords
- labeling
- fuzzy membership
- fuzzy
- tensor
- sample
- 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
- 238000002372 labelling Methods 0.000 title claims abstract description 140
- 238000000034 method Methods 0.000 claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims abstract description 24
- 238000012937 correction Methods 0.000 claims abstract description 15
- 230000010354 integration Effects 0.000 claims abstract description 4
- 238000005315 distribution function Methods 0.000 claims description 17
- 238000004364 calculation method Methods 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a repeatable labeling method system and device for fuzzy information, wherein the method comprises the following steps: acquiring an annotation task and determining an annotation mode, a sample to be annotated and an annotation main body; based on the labeling mode, labeling the sample to be labeled by the labeling main body to obtain corresponding fuzzy membership degrees, and integrating the fuzzy membership degrees corresponding to the labeling main bodies to obtain an initial fuzzy membership degree tensor; estimating fuzzy membership distribution and labeling main body preference, and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix. The system comprises: the system comprises an acquisition module, a labeling module, an integration module and a correction module. The apparatus includes a memory and a processor for performing the repeatable labeling method of ambiguous information described above. By using the method and the device, the fuzzy annotation data can be corrected from the repeated annotation of the fuzzy information, and the influence of the missing data on the annotation quality is reduced. The method can be widely applied to the field of data information labeling.
Description
Technical Field
The invention relates to the field of information labeling, in particular to a repeatable labeling method, a repeatable labeling system and a repeatable labeling device for fuzzy information.
Background
Currently, fuzzy information crowdsourcing labeling works generally involve several problems: 1. the fuzzy information can be repeatedly marked by a plurality of marking personnel on a single sample, and the marking results of the samples cannot be unified; 2. the labeling standards among the samples are difficult to be consistent because the labeling personnel involved in the same batch of samples to be labeled are numerous, and the personal labeling preference difference and the labeling knowledge difference are caused; based on the above reasons, a great amount of repeatable labeling data of heterogeneous unconstrained fuzzy information with different preference information and missing information is generated, and a fuzzy data processing method is urgently needed to correct fuzzy labeling from repeated fuzzy information labeling and reduce the influence of partial data missing in the fuzzy information.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a repeatable labeling method, a repeatable labeling system and a repeatable labeling device for fuzzy information, which can correct fuzzy labeling data from repeated labeling of fuzzy information and reduce the influence of missing data on labeling quality.
The first technical scheme adopted by the invention is as follows: a repeatable labeling method of fuzzy information comprises the following steps:
acquiring an annotation task and determining an annotation mode, a sample to be annotated and an annotation main body;
based on the labeling mode, labeling the sample to be labeled by the labeling main body to obtain a corresponding fuzzy membership degree;
integrating the fuzzy membership degrees corresponding to the labeling subjects to obtain an initial fuzzy membership degree tensor;
estimating fuzzy membership distribution and labeling main body preference, and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix.
Further, the sample to be marked is a repeatable marking sample, the marking main body is marking expert with different preference, and the marking constraint condition of the fuzzy membership degree is an unconstrained marking condition.
Further, the step of integrating the fuzzy membership degrees corresponding to the plurality of labeling subjects to obtain an initial fuzzy membership degree tensor specifically includes:
constructing fuzzy membership matrixes of different labeling experts according to fuzzy membership corresponding to a plurality of labeling experts;
summarizing the fuzzy membership matrixes of all labeling experts to form an initial fuzzy membership tensor.
Further, the step of estimating the fuzzy membership distribution and labeling the subject preference and correcting the initial fuzzy membership tensor to obtain the final fuzzy membership matrix specifically includes:
based on the law of large numbers, predicting a sample to be marked and the fuzzy membership to obey a normal distribution function to obtain a fuzzy membership distribution function;
obtaining a maximum value and a minimum value of the fuzzy membership degree according to the initial fuzzy membership degree tensor;
predicting and labeling preference coefficients of the expert according to the maximum value and the minimum value of the fuzzy membership degree to obtain expert preference coefficient predicted values;
correcting the fuzzy membership tensor according to the fuzzy membership distribution function and the expert preference coefficient predicted value to obtain a final fuzzy membership matrix.
Further, the fuzzy membership obeys a normal distribution, and the formula is as follows:
further, the step of correcting the fuzzy membership tensor according to the fuzzy membership distribution function and the expert preference coefficient pre-estimation value to obtain a final fuzzy membership matrix specifically comprises the following steps:
determining a distribution interval according to the fuzzy membership distribution function;
calculating the average value of expert preference coefficient predicted values;
combining the average value of expert preference coefficient predicted values with the distribution interval to obtain correction data;
and correcting the fuzzy membership tensor according to the correction data to obtain a final fuzzy membership matrix.
The second technical scheme adopted by the invention is as follows: a repeatable annotation system of ambiguous information comprising:
the acquisition module is used for acquiring the labeling task and determining a labeling mode, a sample to be labeled and a labeling main body;
the labeling module is used for labeling the sample to be labeled based on the labeling mode, so as to obtain the corresponding fuzzy membership degree;
the integration module is used for integrating the fuzzy membership degrees corresponding to the labeling subjects to obtain an initial fuzzy membership degree tensor;
and the correction module is used for estimating the fuzzy membership distribution and labeling the main body preference and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix.
The third technical scheme adopted by the invention is as follows: a repeatable labeling device for ambiguous information, comprising:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement a repeatable method of labeling ambiguous information as described above.
The method, the system and the device have the beneficial effects that: according to the fuzzy label correction method, a fuzzy membership degree labeling tensor is established by designing a fuzzy information heterogeneous unconstrained repeatable labeling method, fuzzy membership degree distribution is estimated according to fuzzy information labeling data, preference information of different labeling subjects is estimated, and fuzzy labels are corrected from heterogeneous unconstrained repeatable labels, so that the aims of correcting fuzzy labels from repeated labeling of fuzzy information and reducing influence of missing data on labeling quality are achieved.
Drawings
FIG. 1 is a flow chart of steps of a repeatable labeling method of fuzzy information according to the present invention;
FIG. 2 is a block diagram of a repeatable annotation system of fuzzy information according to the present invention;
fig. 3 is a schematic diagram of a labeling process in an application scenario according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of a correction process in an application scenario according to an embodiment of the present invention.
Detailed Description
The invention 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.
Referring to fig. 1, 3 and 4, the present invention provides a repeatable labeling method of fuzzy information, the method comprising the steps of:
s1, acquiring an annotation task and determining an annotation mode, a sample to be annotated and an annotation main body;
specifically, the sample to be marked is a repeatable marking sample, the marking main body is marking expert with different preferences, and the marking constraint condition of the membership to be fuzzy is an unconstrained marking condition. Expert preference is represented by preference coefficient ρ r (r=1, 2, 3..r.) denotes the R-th expert preference, the higher the expert preference coefficient, the higher the fuzzy membership of the expert label, indicating that the fuzzy membership of the expert label is lower when the sample is in a large relationship with the labeling mode.
The labeling service demander will generally provide a labeling standard (labeling mode)And (5) giving the labeling expert for labeling the sample to be labeled.
X＝{x 1 ,x 2 ,...,x n The sample x is a set of samples to be marked j Represents the j-th sample in X,representing sample x j (j=1, 2,., n) for the labeling mode +.>Is a fuzzy membership of (1). If->Then sample x is called j About labeling mode->Is of fuzzy membership of (2)/>Strong constraints are satisfied (corresponding labeling constraints are referred to as strong constraint labeling conditions), otherwise weak constraints are satisfied (corresponding labeling constraints are referred to as weak constraint labeling conditions). If fuzzy membership->Without the need to satisfy strong or weak constraints, it is said that it satisfies unconstrained (the corresponding labeling constraint is an unconstrained labeling condition).
When the r expert marks part of the sample, sample x j (j=1, 2,., n) relates to the annotation modeFuzzy membership->Form the fuzzy membership matrix of>Wherein->Sample x representing the r expert annotation j About labeling mode->Fuzzy membership->Contains the r-th expert preference coefficient ρ r And labeling mode->Is a piece of information of (a).
Summarizing fuzzy membership matrix of R expertsInitial fuzzy membership tensor composing c×n×r dimension>
S2, labeling the sample to be labeled based on a labeling mode by a labeling main body to obtain a corresponding fuzzy membership degree;
s3, integrating the fuzzy membership degrees corresponding to the labeling subjects to obtain an initial fuzzy membership degree tensor;
s3.1, constructing fuzzy membership matrixes of different labeling experts according to fuzzy membership corresponding to a plurality of labeling experts;
and S3.2, summarizing fuzzy membership matrixes of all labeling experts to form an initial fuzzy membership tensor.
Specifically, in the labeling stage, the r-th expert determines the sample x empirically j And annotation modeLabeling the relation of sample x j About labeling mode->Fuzzy membership->Sample x marked by the r th expert j Concerning annotation mode Can constitute a vectorAfter labeling a portion of the sample by the r-th expert,fuzzy membership matrix capable of forming r expert label +.>
Summarizing fuzzy membership matrix of R expertsComposition of initial fuzzy membership tensor +.>
S4, estimating fuzzy membership distribution and labeling main body preference, and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix.
Fuzzy membership matrix of R experts obtained through initial fuzzy membership tensor labeling processThere are two types of missing values: the first missing value is generated by a sample which is not marked by an expert, and the fuzzy membership degree of the sample with respect to all marking modes is the missing value; the second missing value is generated by human factors or hardware loss, and the fuzzy membership of the sample with respect to the part of the labeling mode is the missing value. The first missing value is called the full missing value and the second missing value is called the partial missing value. Sample x is shown below 2 Is a complete missing value, and sample x j About labeling mode->When the fuzzy membership of (2) is a partial deletion value, < ->NA is the missing value:
the fuzzy membership correction flow is divided into three stages: estimating fuzzy membership distribution of each sample; estimating expert preference; and step three, correcting the fuzzy membership degree by using the fuzzy membership degree distribution and the expert preference predicted value.
S4.1, estimating a sample to be marked and the fuzzy membership degree to obey a normal distribution function based on a large number law to obtain a fuzzy membership degree distribution function;
sample x after initial fuzzy membership tensor labeling process j R experts in sample x j Concerning annotation modeIs +.>Based on the law of large numbers>Obeys normal distribution. Sample x j About labeling mode->The normal distribution formula of (2) is
Wherein:
wherein R is r Sample x noted for R experts j Concerning annotation modeFuzzy membership degreeIs satisfied by->R is the number of R r R is not more than. Initial fuzzy membership tensorFuzzy membership in->Obeys a normal distribution functionThus, the initial fuzzy membership tensor +.>Obeys by->And (3) forming a normal distribution function matrix:
s4.2, obtaining a maximum value and a minimum value of the fuzzy membership degree according to the initial fuzzy membership degree tensor;
s4.3, predicting and labeling preference coefficients of the expert according to the maximum value and the minimum value of the fuzzy membership degree to obtain expert preference coefficient predicted values;
in particular, the method comprises the steps of,the kth sample marked for the kth expert,/->Andsample->About labeling mode->Is of fuzzy membership of (2)And the maximum and minimum of (a) are defined.
K (r) is the number of samples marked by the r expert, the preference ρ of the r expert r With fuzzy membership degreeMaximum value of>And minimum->And (5) correlation. Obtaining a preference coefficient predictive value of the r-th expert according to the following formula>c is the number of annotation modes.
Thus, if expertAll marked with sample x j ，/>To mark sample x j H (x) j ) The h expert in the experts, sample x j Average expert preference coefficient ∈ ->Is that
And S4.4, correcting the fuzzy membership tensor according to the fuzzy membership distribution function and the expert preference coefficient predicted value to obtain a final fuzzy membership matrix.
S4.4.1, determining a distribution interval according to the fuzzy membership distribution function;
s4.4.2, calculating the average value of expert preference coefficient predicted values;
s4.4.3, combining the average value of expert preference coefficient predicted values with a distribution interval to obtain correction data;
s4.4.4, correcting the fuzzy membership tensor according to the correction data to obtain a final fuzzy membership matrix.
Most of fuzzy membership degree3 sigma with values all in the fuzzy membership distribution function ij Interval (mu) ij -3σ ij ,μ ij +3σ ij ) In, sample x j Average expert preference coefficient ∈ ->Binding regionInterval (mu) ij -3σ ij ,μ ij +3σ ij ) Designing a label correction method of repeatable fuzzy membership of the following formula to obtain a final fuzzy membership matrix +.>
Where θ is the threshold, sample x j At the position ofIn about annotation mode->Is>Is that
In particular, if R experts are in sample x j The fuzzy membership degree of all labeling modes is the complete missing value, and thenIf R experts are in sample x j About mode->The fuzzy membership degree of (2) is a partial deletion value, let +.>
Compared with the existing labeling method, the method can be more suitable for large-scale fuzzy information labeling data, can more effectively process the missing problem and can solve the repeatable fuzzy information label correction problem.
As shown in fig. 2, a repeatable labeling system for fuzzy information includes:
the acquisition module is used for acquiring the labeling task and determining a labeling mode, a sample to be labeled and a labeling main body;
the labeling module is used for labeling the sample to be labeled based on the labeling mode, so as to obtain the corresponding fuzzy membership degree;
the integration module is used for integrating the fuzzy membership degrees corresponding to the labeling subjects to obtain an initial fuzzy membership degree tensor;
and the correction module is used for estimating the fuzzy membership distribution and labeling the main body preference and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix.
The content in the method embodiment is applicable to the system embodiment, the functions specifically realized by the system embodiment are the same as those of the method embodiment, and the achieved beneficial effects are the same as those of the method embodiment.
A repeatable labeling device for fuzzy information comprises:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement a repeatable method of labeling ambiguous information as described above.
The content in the method embodiment is applicable to the embodiment of the device, and the functions specifically realized by the embodiment of the device are the same as those of the method embodiment, and the obtained beneficial effects are the same as those of the method embodiment.
A storage medium having stored therein instructions executable by a processor, characterized by: the processor-executable instructions, when executed by the processor, are for implementing a repeatable method of labeling ambiguous information as described above.
The content in the method embodiment is applicable to the storage medium embodiment, and functions specifically implemented by the storage medium embodiment are the same as those of the method embodiment, and the achieved beneficial effects are the same as those of the method embodiment.
While the preferred embodiment of the present invention has been described in detail, the invention 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 invention, and these modifications and substitutions are intended to be included in the scope of the present invention as defined in the appended claims.
Claims (7)
1. The repeatable labeling method of the fuzzy information is characterized by comprising the following steps of:
acquiring an annotation task and determining an annotation mode, a sample to be annotated and an annotation main body;
based on the labeling mode, labeling the sample to be labeled by the labeling main body to obtain a corresponding fuzzy membership degree;
integrating the fuzzy membership degrees corresponding to the labeling subjects to obtain an initial fuzzy membership degree tensor;
estimating fuzzy membership distribution and labeling main body preference, and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix;
the step of estimating the fuzzy membership distribution, labeling the main body preference and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix specifically comprises the following steps:
based on the law of large numbers, predicting a sample to be marked and the fuzzy membership to obey a normal distribution function to obtain a fuzzy membership distribution function;
obtaining a maximum value and a minimum value of the fuzzy membership degree according to the initial fuzzy membership degree tensor;
predicting and labeling preference coefficients of the expert according to the maximum value and the minimum value of the fuzzy membership degree to obtain expert preference coefficient predicted values;
correcting the fuzzy membership tensor according to the fuzzy membership distribution function and the expert preference coefficient predicted value to obtain a final fuzzy membership matrix;
preference coefficient predictive value of the r-th expertThe calculation formula of (2) is as follows:
wherein c is the number of marking modes, r k The kth sample noted for the kth expert,represents the maximum value of fuzzy membership, +.>And (3) representing the minimum value of the fuzzy membership degree, wherein K (r) is the number of samples marked by the r-th expert.
2. The repeatable labeling method of fuzzy information according to claim 1, wherein the sample to be labeled is a repeatable labeling sample, the labeling subjects are labeling experts with different preferences, and the labeling constraint condition of the fuzzy membership degree is an unconstrained labeling condition.
3. The repeatable labeling method of fuzzy information according to claim 2, wherein the step of integrating fuzzy membership degrees corresponding to a plurality of labeling subjects to obtain an initial fuzzy membership degree tensor comprises the following steps:
constructing fuzzy membership matrixes of different labeling experts according to fuzzy membership corresponding to a plurality of labeling experts;
summarizing the fuzzy membership matrixes of all labeling experts to form an initial fuzzy membership tensor.
4. A method for repeatable labeling of fuzzy information according to claim 3, wherein said fuzzy membership obeys a normal distribution, formulated as follows:
wherein,representing an initial fuzzy membership tensor ρ r R=1, 2, 3..r, R represents the R-th expert preference, R represents the total number of experts,/-in>Representing the annotation mode.
5. The method for repeatable labeling of fuzzy information of claim 4, wherein said step of correcting fuzzy membership tensors based on fuzzy membership distribution function and expert preference coefficient predictive value to obtain final fuzzy membership matrix comprises:
determining a distribution interval according to the fuzzy membership distribution function;
calculating the average value of expert preference coefficient predicted values;
combining the average value of expert preference coefficient predicted values with the distribution interval to obtain correction data;
and correcting the fuzzy membership tensor according to the correction data to obtain a final fuzzy membership matrix.
6. A system for repeatable annotation of ambiguous information as claimed in claim 1 comprising:
the acquisition module is used for acquiring the labeling task and determining a labeling mode, a sample to be labeled and a labeling main body;
the labeling module is used for labeling the sample to be labeled based on the labeling mode, so as to obtain the corresponding fuzzy membership degree;
the integration module is used for integrating the fuzzy membership degrees corresponding to the labeling subjects to obtain an initial fuzzy membership degree tensor;
and the correction module is used for estimating the fuzzy membership distribution and labeling the main body preference and correcting the initial fuzzy membership tensor to obtain a final fuzzy membership matrix.
7. A repeatable marking device for fuzzy information, comprising:
at least one processor;
at least one memory for storing at least one program;
when the at least one program is executed by the at least one processor, the at least one processor is caused to implement a method of repeatable labeling of ambiguous information as claimed in any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211026598.2A CN115859059B (en) | 2022-08-25 | 2022-08-25 | Repeatable labeling method, system and device for fuzzy information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211026598.2A CN115859059B (en) | 2022-08-25 | 2022-08-25 | Repeatable labeling method, system and device for fuzzy information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115859059A CN115859059A (en) | 2023-03-28 |
| CN115859059B true CN115859059B (en) | 2024-03-22 |
Family
ID=85660630
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211026598.2A Active CN115859059B (en) | 2022-08-25 | 2022-08-25 | Repeatable labeling method, system and device for fuzzy information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115859059B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104199852A (en) * | 2014-08-12 | 2014-12-10 | 上海交通大学 | Label propagation community structure mining method based on node membership degree |
| CN106600605A (en) * | 2016-12-14 | 2017-04-26 | 陕西科技大学 | Unsupervised fast image segmentation algorithm |
| CN109389608A (en) * | 2018-10-19 | 2019-02-26 | 山东大学 | There is the fuzzy clustering image partition method of noise immunity using plane as cluster centre |
| CN109495296A (en) * | 2018-11-02 | 2019-03-19 | 国网四川省电力公司电力科学研究院 | Intelligent substation communication network state evaluation method based on clustering and neural network |
| CN112560921A (en) * | 2020-12-10 | 2021-03-26 | 百维金科(上海)信息科技有限公司 | Internet financial platform application fraud detection method based on fuzzy C-mean |
| CN113177358A (en) * | 2021-04-30 | 2021-07-27 | 燕山大学 | Soft measurement method for cement quality based on fuzzy fine-grained feature extraction |
| CN113688258A (en) * | 2021-08-20 | 2021-11-23 | 广东工业大学 | Information recommendation method and system based on flexible multidimensional clustering |
| CN114266321A (en) * | 2021-12-31 | 2022-04-01 | 广东泰迪智能科技股份有限公司 | Weak supervision fuzzy clustering algorithm based on unconstrained prior information mode |
| CN114330574A (en) * | 2021-12-31 | 2022-04-12 | 广东泰迪智能科技股份有限公司 | Fuzzy labeling method for pattern recognition |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070112695A1 (en) * | 2004-12-30 | 2007-05-17 | Yan Wang | Hierarchical fuzzy neural network classification |
| CN112580483B (en) * | 2020-12-15 | 2023-12-01 | 海南大学 | Remote sensing image feature discretization method based on rough fuzzy model |
| CN115392376B (en) * | 2022-08-25 | 2024-02-02 | 广东工业大学 | Heterogeneous fuzzy membership matrix labeling method, system and device |
-
2022
- 2022-08-25 CN CN202211026598.2A patent/CN115859059B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104199852A (en) * | 2014-08-12 | 2014-12-10 | 上海交通大学 | Label propagation community structure mining method based on node membership degree |
| CN106600605A (en) * | 2016-12-14 | 2017-04-26 | 陕西科技大学 | Unsupervised fast image segmentation algorithm |
| CN109389608A (en) * | 2018-10-19 | 2019-02-26 | 山东大学 | There is the fuzzy clustering image partition method of noise immunity using plane as cluster centre |
| CN109495296A (en) * | 2018-11-02 | 2019-03-19 | 国网四川省电力公司电力科学研究院 | Intelligent substation communication network state evaluation method based on clustering and neural network |
| CN112560921A (en) * | 2020-12-10 | 2021-03-26 | 百维金科(上海)信息科技有限公司 | Internet financial platform application fraud detection method based on fuzzy C-mean |
| CN113177358A (en) * | 2021-04-30 | 2021-07-27 | 燕山大学 | Soft measurement method for cement quality based on fuzzy fine-grained feature extraction |
| CN113688258A (en) * | 2021-08-20 | 2021-11-23 | 广东工业大学 | Information recommendation method and system based on flexible multidimensional clustering |
| CN114266321A (en) * | 2021-12-31 | 2022-04-01 | 广东泰迪智能科技股份有限公司 | Weak supervision fuzzy clustering algorithm based on unconstrained prior information mode |
| CN114330574A (en) * | 2021-12-31 | 2022-04-12 | 广东泰迪智能科技股份有限公司 | Fuzzy labeling method for pattern recognition |
Non-Patent Citations (2)
| Title |
|---|
| Kernel extreme learning machine based on fuzzy set theory for multi-label classification;Kongsorot Y等;《International Journal of Machine Learning and Cybernetics》;第 979-989页 * |
| 基于交叉熵测度的成对约束模糊核聚类算法;徐圣兵;林上钧;钟国祥;;应用科技;47(第01期);第80-87页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115859059A (en) | 2023-03-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112232407B (en) | Neural network model training method and device for pathological image samples | |
| CN112446441B (en) | Model training data screening method, device, equipment and storage medium | |
| CN115392376B (en) | Heterogeneous fuzzy membership matrix labeling method, system and device | |
| CN110826494A (en) | Method and device for evaluating quality of labeled data, computer equipment and storage medium | |
| CN113255849B (en) | Label noisy image learning method based on double active queries | |
| CN111199186A (en) | Image quality scoring model training method, device, equipment and storage medium | |
| CN115859059B (en) | Repeatable labeling method, system and device for fuzzy information | |
| US20050080979A1 (en) | Optimization with unknown objective function | |
| CN114782355A (en) | Gastric cancer digital pathological section detection method based on improved VGG16 network | |
| US20020181767A1 (en) | Classification of photos with sepia tones | |
| CN108597602B (en) | Label error correction method for skin medical data | |
| WO2023201772A1 (en) | Cross-domain remote sensing image semantic segmentation method based on adaptation and self-training in iteration domain | |
| CN114943741A (en) | Visual SLAM method based on target detection and geometric probability in dynamic scene | |
| CN109089125B (en) | DCT coefficient distribution characteristic model optimization method in video coding | |
| CN115357769B (en) | Government affair data labeling system and method | |
| CN116563642B (en) | Image classification model credible training and image classification method, device and equipment | |
| US20040225478A1 (en) | EDR direction estimating method, system, and program, and memory medium for storing the program | |
| CN116415423B (en) | Computer simulation data processing system and method based on big data analysis | |
| Fan et al. | Local linear smoothers in regression function estimation | |
| CN116128426A (en) | Automatic auditing method for crowd-sourced labeling task of digital cultural heritage data | |
| CN117831046A (en) | Data annotation frame correction method | |
| CN116543015A (en) | Method and system for constructing cell lineage tree based on deep learning | |
| CN117745101A (en) | Computing method, computing system and terminal for industry recommendation technology service | |
| CN115100455A (en) | Image recognition model generalization capability enhancement method and system and computer readable storage medium | |
| CN113936168A (en) | Image classification method and system based on online domain adaptive deep learning |
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 |