CN109145937A - A kind of method and device of model training - Google Patents
A kind of method and device of model training Download PDFInfo
- Publication number
- CN109145937A CN109145937A CN201810664585.5A CN201810664585A CN109145937A CN 109145937 A CN109145937 A CN 109145937A CN 201810664585 A CN201810664585 A CN 201810664585A CN 109145937 A CN109145937 A CN 109145937A
- Authority
- CN
- China
- Prior art keywords
- sample data
- cluster labels
- cluster
- data
- class categories
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The embodiment of the invention provides a kind of method and devices of model training, the method comprise the steps that obtaining the sample data to be trained in specified class categories；Feature extraction is carried out to the sample data to be trained, obtains the corresponding characteristic information of the specified class categories；The corresponding characteristic information of the specified class categories is clustered, multiple cluster labels are obtained；Data balancing processing is carried out to the corresponding sample data of the cluster labels；Using data balancing treated sample data as target sample data；Using the target sample data, training designated model.The present invention can refine the label in existing class categories by above-mentioned unsupervised method, realize that the sample in classification is balanced, balanced sample data is provided for model, the model of the available optimization of model training is carried out according to the sample data of the equilibrium, data are carried out using the model of the optimization and predict available more accurate prediction result, improve the accuracy rate of model prediction.
Description
Technical field
The present invention relates to modeling technique fields, method, a kind of dress of model training more particularly to a kind of model training
It sets, a kind of model training systems and one or more machine readable media.
Background technique
Image classification is to be distinguished different classes of image according to the semantic information of image, is important in computer vision
Basic problem and image detection, image segmentation, object tracking, behavioural analysis etc. other high-rise visual tasks basis.
In recent years, deep learning has obtained answering extensively in related fieldss such as video image, speech recognition, natural language processings
With.An important branch of the convolutional neural networks (CNN) as deep learning, due to its superpower capability of fitting and end-to-end
Global optimization ability so that image classification task application convolutional neural networks after, precision of prediction is substantially improved.
Although current image classification model is provided with certain classification capacity to image, a large amount of predictions are still suffered from
How the sample of mistake advanced optimizes image classification model as a problem to be solved.
Summary of the invention
The embodiment of the present invention is the technical problem to be solved is that a kind of method of model training is provided, to solve existing mould
There are the sample of a large amount of prediction errors, the lower problems of predictablity rate for type.
Correspondingly, the embodiment of the invention also provides a kind of devices of model training, a kind of model training systems and one
A or multiple machine readable medias, to guarantee the implementation and application of the above method.
To solve the above-mentioned problems, the invention discloses a kind of methods of model training, which comprises
Obtain the sample data to be trained in specified class categories；
Feature extraction is carried out to the sample data to be trained, obtains the corresponding feature letter of the specified class categories
Breath；
The corresponding characteristic information of the specified class categories is clustered, multiple cluster labels are obtained；
Data balancing processing is carried out to the corresponding sample data of the cluster labels；
Using data balancing treated sample data as target sample data；
Using the target sample data, training designated model.
Preferably, before the progress data balancing processing to the cluster labels corresponding sample data, further includes:
The cluster labels for meeting preset condition are filtered out from the multiple cluster labels.
It is preferably, described that the cluster labels for meeting preset condition are filtered out from the multiple cluster labels, comprising:
From the sample data of the specified class categories, the number of the corresponding sample data of each cluster labels is counted respectively
Amount；
It calculates in the specified class categories, the average of samples of each cluster labels；
The quantity for abandoning sample data is lower than the cluster labels of the average of samples, obtains the cluster labels of reservation.
It is preferably, described that data balancing processing is carried out to the corresponding sample data of the cluster labels, comprising:
The quantity of sample data in more each candidate cluster label, choose at least one cluster labels of minimum number into
Row data enhancing processing.
Preferably, described that the corresponding characteristic information of the specified class categories is clustered, multiple cluster labels are obtained,
Include:
Set the quantity of cluster centre；
According to the quantity of the cluster centre, the characteristic information is gathered by the way of cosine cosine distance
Class obtains multiple cluster labels.
The invention also discloses a kind of device of model training, described device includes:
Sample data obtains module, for obtaining the sample data to be trained in specified class categories；
Characteristic extracting module obtains the specified classification for carrying out feature extraction to the sample data to be trained
The corresponding characteristic information of classification；
Cluster module obtains multiple clusters marks for clustering to the corresponding characteristic information of the specified class categories
Label；
Balance processing module, for carrying out data balancing processing to the corresponding sample data of the cluster labels；
Model training module, for using data balancing treated sample data as target sample data, and using institute
State target sample data, training designated model.
Preferably, described device further include:
Screening module, for filtering out the cluster labels for meeting preset condition from the multiple cluster labels.
Preferably, the screening module includes:
Statistic submodule, for counting each cluster labels pair respectively from the sample data of the specified class categories
The quantity for the sample data answered；
Average of samples computational submodule, for calculating in the specified class categories, the sample of each cluster labels is flat
Mean；
Sample abandons submodule, and the quantity for abandoning sample data is lower than the cluster labels of the average of samples, obtains
The cluster labels that must retain.
Preferably, the balance processing module is also used to:
The quantity of sample data in more each candidate cluster label, choose at least one cluster labels of minimum number into
Row data enhancing processing.
Preferably, the cluster module is also used to:
Set the quantity of cluster centre；
According to the quantity of the cluster centre, the characteristic information is gathered by the way of cosine cosine distance
Class obtains multiple cluster labels.
The invention also discloses a kind of model training systems, comprising:
One or more processors；With
One or more machine readable medias of instruction are stored thereon with, are executed when by one or more of processors
When, so that the method that the electronic equipment executes a kind of above-mentioned model training.
The invention also discloses one or more machine readable medias, are stored thereon with instruction, when by one or more
When managing device execution, so that the method that the processor executes a kind of above-mentioned model training.
Compared with prior art, the embodiment of the present invention includes following advantages:
In embodiments of the present invention, when obtain in specified class categories after the sample data trained, can be to this
Sample data carries out feature extraction, obtains the specified corresponding characteristic information of class categories, then clusters, obtain to characteristic information
Data balancing processing is carried out to multiple cluster labels, and to the sample data in cluster labels.Pass through above-mentioned unsupervised method
The label in existing class categories is refined, realizes that the sample in classification is balanced, balanced sample data is provided for model, according to this
Balanced sample data carries out the model of the available optimization of model training, and using the model of the optimization to carry out data prediction can be with
More accurate prediction result is obtained, the accuracy rate of model prediction is improved.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of the embodiment of the method for model training of the embodiment of the present invention；
Fig. 2 is a kind of structural block diagram of the Installation practice of model training of the embodiment of the present invention.
Specific embodiment
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Referring to Fig.1, a kind of step flow chart of the embodiment of the method for model training of the embodiment of the present invention is shown, specifically
It may include steps of:
Step 101, the sample data to be trained in specified class categories is obtained；
In one embodiment, specified class categories can be the class categories determining by preset disaggregated model,
And/or the specified class categories are also possible to through artificially specified class categories, the embodiment of the present invention to this with no restriction.
In one embodiment, which can be the class categories of coarseness, for example, a differentiation figure
As the inside is either with or without the scene of people, specified class categories may include " someone " and " no one " the two class categories.
The embodiment of the present invention can be applied to the processing of the sample data in classification.It, can for each specified class categories
To obtain all sample datas in the specified class categories, i.e. all samples for being identified as specified class categories in acquisition training set
Notebook data.For example, being identified as all sample datas of " someone " classification.
As an example, which may include image data.
Step 102, feature extraction is carried out to the sample data to be trained, it is corresponding obtains the specified class categories
Characteristic information；
Obtain in specified class categories after the sample data trained, can be by a disaggregated model, such as convolution mind
It through network model, specifies the sample data to be trained in class categories to carry out feature extraction this, obtains the specified classification class
Not corresponding characteristic information.
In one embodiment, for i-th of specified class categories, the characteristic information (i.e. characteristic set) extracted can
With with featureiIt indicates.
Step 103, the corresponding characteristic information of the specified class categories is clustered, obtains multiple cluster labels；
After obtaining the specified corresponding characteristic information of class categories, this feature information can be clustered, to obtain
Multiple cluster labels.
In one embodiment, the characteristic information that K-means specifies class categories to each can be used
featureiIt is clustered.
Wherein, K-means algorithm is hard clustering algorithm, is the clustering algorithm based on distance, using distance as similar
The evaluation index of property, that is, think that the distance of two objects is closer, similarity is bigger.The algorithm thinks that cluster is by apart from close
Object composition, therefore obtaining compact and independent cluster as final goal.The input of K-MEANS algorithm are as follows: in cluster
Heart number k and the corresponding characteristic information of specified class categories；Its output is: meeting multiple cluster marks of variance minimum sandards
Label.
Specifically, in one embodiment, step 103 may include following sub-step:
Sub-step S11 sets the quantity of cluster centre；
In one embodiment, the quantity K of cluster centre can be selected according to the priori knowledge of data.For mistake
The quantity K of cluster centre can be set greater than the value of priori knowledge by the purpose of noise filtering data.For example, being identified as " has
There are 3 subclassifications, respectively " man ", " woman " and " child " in the specified class categories of people ", then it can be by cluster centre
Quantity K is set greater than the value of numerical value 3, for example, K is set as numerical value 5.
It should be noted that the setting of the quantity of cluster centre can be set by receiving the input value of user, it can also
To set after obtaining priori knowledge automatically, the embodiment of the present invention to this with no restriction.
Sub-step S12 believes the feature by the way of cosine cosine distance according to the quantity of the cluster centre
Breath is clustered, and multiple cluster labels are obtained.
In the present embodiment, in order to match with the feature space that disaggregated model extracts, the distance measure of K-means algorithm can
It to use cosine distance, is clustered according to cosine similarity, finally obtains multiple cluster labels j.
In a preferred embodiment of an embodiment of the present invention, can also include the following steps:
The cluster labels for meeting preset condition are filtered out from the multiple cluster labels；
After obtaining multiple cluster labels in specified class categories, the cluster mark for not meeting preset condition can be deleted
Label retain the cluster labels for meeting preset condition.
In a preferred embodiment of an embodiment of the present invention, it is above-mentioned filtered out from the multiple cluster labels meet it is pre-
If the cluster labels of condition can further include following sub-step:
Sub-step S21 counts the corresponding sample of each cluster labels from the sample data of the specified class categories respectively
The quantity of notebook data；
In one embodiment, specify the quantity of the corresponding sample data of cluster labels j in class categories i that can indicate
For Counti,j.For example, cluster labels are the number of samples of " man " in this specified class categories of statistics " someone ".
Sub-step S22 is calculated in the specified class categories, the average of samples of each cluster labels；
In one embodiment, in the average of samples of cluster labels=specified class categories all samples quantity
(CountiThe quantity of)/cluster labels.
Sub-step S23, the quantity for abandoning sample data are lower than the cluster labels of the average of samples, obtain the poly- of reservation
Class label.
In one embodiment, a threshold tau can be set in advance, theoretical selective value can be that (K is poly- to τ=1/K
The quantity of class label).If Counti,j<τ*Counti, i.e., the sample size of some cluster labels be less than average sample quantity,
Then abandon cluster labels j.After having traversed all cluster labels, the cluster labels finally retained are obtained.
Step 104, data balancing processing is carried out to the corresponding sample data of the cluster labels；
For remaining cluster labels, data balancing can be carried out using data enhancing or other data balancing algorithms
Processing, to reach data balancing in specified class categories.
In a preferred embodiment of an embodiment of the present invention, step 104 can further include following sub-step:
The quantity of sample data in more each candidate cluster label, choose at least one cluster labels of minimum number into
Row data enhancing processing.
For example, under the specified class categories of " someone ", altogether there are three cluster labels, respectively " man ", " woman " and
" child ", wherein " child " this cluster labels cause the corresponding sample data of this cluster labels seldom for some reason,
That is the cluster labels of " child " are unbalanced, eventually lead to model and are deteriorated to the recognition capability of " child ", it is thus possible to increase " small
Sample data in this cluster labels of child ", so that data balancing.
Step 105, using data balancing treated sample data as target sample data, and the target sample is used
Data, training designated model.
Specifically, carrying out data balancing, treated that sample data can be used as final training data (i.e. target sample
Data), can be with re -training model according to the target sample data, therefore obtained model is sample in specified class categories
Balanced Optimized model, the accuracy for carrying out data prediction using the Optimized model are higher.
In embodiments of the present invention, when obtain in specified class categories after the sample data trained, can be to this
Sample data carries out feature extraction, obtains the specified corresponding characteristic information of class categories, then clusters, obtain to characteristic information
Data balancing processing is carried out to multiple cluster labels, and to the sample data in cluster labels.Pass through above-mentioned unsupervised method
The label in existing class categories is refined, realizes that the sample in classification is balanced, balanced sample data is provided for model, according to this
Balanced sample data carries out the model of the available optimization of model training, and using the model of the optimization to carry out data prediction can be with
More accurate prediction result is obtained, the accuracy rate of model prediction is improved.
It should be noted that for simple description, therefore, it is stated as a series of action groups for embodiment of the method
It closes, but those skilled in the art should understand that, embodiment of that present invention are not limited by the describe sequence of actions, because according to
According to the embodiment of the present invention, some steps may be performed in other sequences or simultaneously.Secondly, those skilled in the art also should
Know, the embodiments described in the specification are all preferred embodiments, and the related movement not necessarily present invention is implemented
Necessary to example.
Referring to Fig. 2, a kind of structural block diagram of the Installation practice of model training of the embodiment of the present invention is shown, specifically may be used
To include following module:
Sample data obtains module 201, for obtaining the sample data to be trained in specified class categories；
Characteristic extracting module 202 obtains described specified point for carrying out feature extraction to the sample data to be trained
The corresponding characteristic information of class classification；
Cluster module 203 obtains multiple clusters for clustering to the corresponding characteristic information of the specified class categories
Label；
Balance processing module 204, for carrying out data balancing processing to the corresponding sample data of the cluster labels；
Model training module 205 for using data balancing treated sample data as target sample data, and uses
The target sample data, training designated model.
In a preferred embodiment of an embodiment of the present invention, described device can also include following module:
Screening module, for filtering out the cluster labels for meeting preset condition from the multiple cluster labels.
In a preferred embodiment of an embodiment of the present invention, the screening module can further include following submodule
Block:
Statistic submodule, for counting each cluster labels pair respectively from the sample data of the specified class categories
The quantity for the sample data answered；
Average of samples computational submodule, for calculating in the specified class categories, the sample of each cluster labels is flat
Mean；
Sample abandons submodule, and the quantity for abandoning sample data is lower than the cluster labels of the average of samples, obtains
The cluster labels that must retain.
In a preferred embodiment of an embodiment of the present invention, the balance processing module 204 is also used to:
The quantity of sample data in more each candidate cluster label, choose at least one cluster labels of minimum number into
Row data enhancing processing.
In a preferred embodiment of an embodiment of the present invention, the cluster module 203 is also used to:
Set the quantity of cluster centre；
According to the quantity of the cluster centre, the characteristic information is gathered by the way of cosine cosine distance
Class obtains multiple cluster labels.
For device embodiment, since it is substantially similar to above method embodiment, so be described relatively simple,
The relevent part can refer to the partial explaination of embodiments of method.
The embodiment of the invention also provides a kind of model training systems, comprising:
One or more processors；With
One or more machine readable medias of instruction are stored thereon with, are executed when by one or more of processors
When, so that the method that the electronic equipment executes a kind of above-mentioned model training.
The embodiment of the invention also provides one or more machine readable medias, are stored thereon with instruction, when by one or
When multiple processors execute, so that the method that the processor executes a kind of above-mentioned model training.
All the embodiments in this specification are described in a progressive manner, the highlights of each of the examples are with
The difference of other embodiments, the same or similar parts between the embodiments can be referred to each other.
It should be understood by those skilled in the art that, the embodiment of the embodiment of the present invention can provide as method, apparatus or calculate
Machine program product.Therefore, the embodiment of the present invention can be used complete hardware embodiment, complete software embodiment or combine software and
The form of the embodiment of hardware aspect.Moreover, the embodiment of the present invention can be used one or more wherein include computer can
With in the computer-usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) of program code
The form of the computer program product of implementation.
The embodiment of the present invention be referring to according to the method for the embodiment of the present invention, terminal device (system) and computer program
The flowchart and/or the block diagram of product describes.It should be understood that flowchart and/or the block diagram can be realized by computer program instructions
In each flow and/or block and flowchart and/or the block diagram in process and/or box combination.It can provide these
Computer program instructions are set to general purpose computer, special purpose computer, Embedded Processor or other programmable data processing terminals
Standby processor is to generate a machine, so that being held by the processor of computer or other programmable data processing terminal devices
Capable instruction generates for realizing in one or more flows of the flowchart and/or one or more blocks of the block diagram
The device of specified function.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing terminal devices
In computer-readable memory operate in a specific manner, so that instruction stored in the computer readable memory generates packet
The manufacture of command device is included, which realizes in one side of one or more flows of the flowchart and/or block diagram
The function of being specified in frame or multiple boxes.
These computer program instructions can also be loaded into computer or other programmable data processing terminal devices, so that
Series of operation steps are executed on computer or other programmable terminal equipments to generate computer implemented processing, thus
The instruction executed on computer or other programmable terminal equipments is provided for realizing in one or more flows of the flowchart
And/or in one or more blocks of the block diagram specify function the step of.
Although the preferred embodiment of the embodiment of the present invention has been described, once a person skilled in the art knows bases
This creative concept, then additional changes and modifications can be made to these embodiments.So the following claims are intended to be interpreted as
Including preferred embodiment and fall into all change and modification of range of embodiment of the invention.
Finally, it is to be noted that, herein, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that process, method, article or terminal device including a series of elements not only wrap
Those elements are included, but also including other elements that are not explicitly listed, or further includes for this process, method, article
Or the element that terminal device is intrinsic.In the absence of more restrictions, being wanted by what sentence "including a ..." limited
Element, it is not excluded that there is also other identical elements in process, method, article or the terminal device for including the element.
A kind of method and device of model training provided by the present invention is described in detail above, it is used herein
A specific example illustrates the principle and implementation of the invention, and the above embodiments are only used to help understand
Method and its core concept of the invention；At the same time, for those skilled in the art is having according to the thought of the present invention
There will be changes in body embodiment and application range, in conclusion the content of the present specification should not be construed as to the present invention
Limitation.
Claims (12)
1. a kind of method of model training, which is characterized in that the described method includes:
Obtain the sample data to be trained in specified class categories；
Feature extraction is carried out to the sample data to be trained, obtains the corresponding characteristic information of the specified class categories；
The corresponding characteristic information of the specified class categories is clustered, multiple cluster labels are obtained；
Data balancing processing is carried out to the corresponding sample data of the cluster labels；
Using data balancing treated sample data as target sample data；
Using the target sample data, training designated model.
2. the method according to claim 1, wherein it is described to the corresponding sample data of the cluster labels into
Before row data equilibrium treatment, further includes:
The cluster labels for meeting preset condition are filtered out from the multiple cluster labels.
3. according to the method described in claim 2, it is characterized in that, it is described filtered out from the multiple cluster labels meet it is pre-
If the cluster labels of condition, comprising:
From the sample data of the specified class categories, the quantity of the corresponding sample data of each cluster labels is counted respectively；
It calculates in the specified class categories, the average of samples of each cluster labels；
The quantity for abandoning sample data is lower than the cluster labels of the average of samples, obtains the cluster labels of reservation.
4. according to the method described in claim 3, it is characterized in that, described carry out the corresponding sample data of the cluster labels
Data balancing processing, comprising:
The quantity of sample data in more each candidate cluster label, at least one cluster labels for choosing minimum number are counted
It is handled according to enhancing.
5. method according to claim 1-4, which is characterized in that described corresponding to the specified class categories
Characteristic information is clustered, and multiple cluster labels are obtained, comprising:
Set the quantity of cluster centre；
According to the quantity of the cluster centre, the characteristic information is clustered by the way of cosine cosine distance, is obtained
To multiple cluster labels.
6. a kind of device of model training, which is characterized in that described device includes:
Sample data obtains module, for obtaining the sample data to be trained in specified class categories；
Characteristic extracting module obtains the specified class categories for carrying out feature extraction to the sample data to be trained
Corresponding characteristic information；
Cluster module obtains multiple cluster labels for clustering to the corresponding characteristic information of the specified class categories；
Balance processing module, for carrying out data balancing processing to the corresponding sample data of the cluster labels；
Model training module, for data balancing treated sample data to be used as to target sample data, and the use mesh
Standard specimen notebook data, training designated model.
7. device according to claim 6, which is characterized in that further include:
Screening module, for filtering out the cluster labels for meeting preset condition from the multiple cluster labels.
8. device according to claim 7, which is characterized in that the screening module includes:
Statistic submodule, for it is corresponding to count each cluster labels respectively from the sample data of the specified class categories
The quantity of sample data；
Average of samples computational submodule, for calculating in the specified class categories, the average of samples of each cluster labels；
Sample abandons submodule, and the quantity for abandoning sample data is lower than the cluster labels of the average of samples, is protected
The cluster labels stayed.
9. device according to claim 8, which is characterized in that the balance processing module is also used to:
The quantity of sample data in more each candidate cluster label, at least one cluster labels for choosing minimum number are counted
It is handled according to enhancing.
10. according to the described in any item devices of claim 6-9, which is characterized in that the cluster module is also used to:
Set the quantity of cluster centre；
According to the quantity of the cluster centre, the characteristic information is clustered by the way of cosine cosine distance, is obtained
To multiple cluster labels.
11. a kind of model training systems characterized by comprising
One or more processors；With
One or more machine readable medias of instruction are stored thereon with, when being executed by one or more of processors, are made
The method for obtaining a kind of model training that the electronic equipment is executed such as claim 1-5 one or more.
12. one or more machine readable medias, are stored thereon with instruction, when executed by one or more processors, so that
The method that the processor executes a kind of model training such as claim 1-5 one or more.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810664585.5A CN109145937A (en) | 2018-06-25 | 2018-06-25 | A kind of method and device of model training |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810664585.5A CN109145937A (en) | 2018-06-25 | 2018-06-25 | A kind of method and device of model training |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109145937A true CN109145937A (en) | 2019-01-04 |
Family
ID=64802349
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810664585.5A Pending CN109145937A (en) | 2018-06-25 | 2018-06-25 | A kind of method and device of model training |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109145937A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109934281A (en) * | 2019-03-08 | 2019-06-25 | 电子科技大学 | A kind of unsupervised training method of two sorter networks |
| CN110147851A (en) * | 2019-05-29 | 2019-08-20 | 北京达佳互联信息技术有限公司 | Method for screening images, device, computer equipment and storage medium |
| CN110570469A (en) * | 2019-08-16 | 2019-12-13 | 广州威尔森信息科技有限公司 | intelligent identification method for angle position of automobile picture |
| CN110689066A (en) * | 2019-09-24 | 2020-01-14 | 成都考拉悠然科技有限公司 | Training method combining face recognition data equalization and enhancement |
| CN111191122A (en) * | 2019-12-20 | 2020-05-22 | 重庆邮电大学 | Learning resource recommendation system based on user portrait |
| CN111553173A (en) * | 2020-04-23 | 2020-08-18 | 苏州思必驰信息科技有限公司 | Natural language generation training method and device |
| CN111611388A (en) * | 2020-05-29 | 2020-09-01 | 北京学之途网络科技有限公司 | Account classification method, device and equipment |
| CN111611457A (en) * | 2020-05-20 | 2020-09-01 | 北京金山云网络技术有限公司 | Page classification method, device, equipment and storage medium |
| CN111783869A (en) * | 2020-06-29 | 2020-10-16 | 杭州海康威视数字技术股份有限公司 | Training data screening method and device, electronic equipment and storage medium |
| CN111797876A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Data classification method and device, storage medium and electronic equipment |
| CN111860671A (en) * | 2020-07-28 | 2020-10-30 | 中山大学 | Classification model training method and device, terminal equipment and readable storage medium |
| CN111970584A (en) * | 2020-07-08 | 2020-11-20 | 国网宁夏电力有限公司电力科学研究院 | Method, device and equipment for processing data and storage medium |
| CN112749724A (en) * | 2019-10-31 | 2021-05-04 | 阿里巴巴集团控股有限公司 | Method and equipment for training classifier and predicting application performance expansibility |
| CN112861512A (en) * | 2021-02-05 | 2021-05-28 | 北京百度网讯科技有限公司 | Data processing method, device, equipment and storage medium |
| CN113128536A (en) * | 2019-12-31 | 2021-07-16 | 奇安信科技集团股份有限公司 | Unsupervised learning method, system, computer device and readable storage medium |
| CN113408282A (en) * | 2021-08-06 | 2021-09-17 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for topic model training and topic prediction |
| WO2021189830A1 (en) * | 2020-03-26 | 2021-09-30 | 平安科技（深圳）有限公司 | Sample data optimization method, apparatus and device, and storage medium |
| CN113487320A (en) * | 2021-06-28 | 2021-10-08 | 深圳索信达数据技术有限公司 | Fraud transaction detection method, device, computer equipment and storage medium |
| CN114077860A (en) * | 2020-08-18 | 2022-02-22 | 鸿富锦精密电子(天津)有限公司 | Method and system for sorting parts before assembly, electronic device and storage medium |
| WO2023166747A1 (en) * | 2022-03-04 | 2023-09-07 | 日本電信電話株式会社 | Training data generation device, training data generation method, and program |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103048273A (en) * | 2012-11-09 | 2013-04-17 | 江苏大学 | Fruit near infrared spectrum sorting method based on fuzzy clustering |
| CN103617429A (en) * | 2013-12-16 | 2014-03-05 | 苏州大学 | Sorting method and system for active learning |
| US8787692B1 (en) * | 2011-04-08 | 2014-07-22 | Google Inc. | Image compression using exemplar dictionary based on hierarchical clustering |
| CN106599935A (en) * | 2016-12-29 | 2017-04-26 | 重庆邮电大学 | Three-decision unbalanced data oversampling method based on Spark big data platform |
| CN106682684A (en) * | 2016-11-23 | 2017-05-17 | 天津津航计算技术研究所 | K-means clustering-based target recognition method |
| CN107194430A (en) * | 2017-05-27 | 2017-09-22 | 北京三快在线科技有限公司 | A kind of screening sample method and device, electronic equipment |
-
2018
- 2018-06-25 CN CN201810664585.5A patent/CN109145937A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8787692B1 (en) * | 2011-04-08 | 2014-07-22 | Google Inc. | Image compression using exemplar dictionary based on hierarchical clustering |
| CN103048273A (en) * | 2012-11-09 | 2013-04-17 | 江苏大学 | Fruit near infrared spectrum sorting method based on fuzzy clustering |
| CN103617429A (en) * | 2013-12-16 | 2014-03-05 | 苏州大学 | Sorting method and system for active learning |
| CN106682684A (en) * | 2016-11-23 | 2017-05-17 | 天津津航计算技术研究所 | K-means clustering-based target recognition method |
| CN106599935A (en) * | 2016-12-29 | 2017-04-26 | 重庆邮电大学 | Three-decision unbalanced data oversampling method based on Spark big data platform |
| CN107194430A (en) * | 2017-05-27 | 2017-09-22 | 北京三快在线科技有限公司 | A kind of screening sample method and device, electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 张良均: "《MATLAB数据分析与挖掘实战》", 30 June 2015, 机械工业出版社 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109934281A (en) * | 2019-03-08 | 2019-06-25 | 电子科技大学 | A kind of unsupervised training method of two sorter networks |
| CN111797876A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Data classification method and device, storage medium and electronic equipment |
| CN110147851A (en) * | 2019-05-29 | 2019-08-20 | 北京达佳互联信息技术有限公司 | Method for screening images, device, computer equipment and storage medium |
| CN110147851B (en) * | 2019-05-29 | 2022-04-01 | 北京达佳互联信息技术有限公司 | Image screening method and device, computer equipment and storage medium |
| CN110570469A (en) * | 2019-08-16 | 2019-12-13 | 广州威尔森信息科技有限公司 | intelligent identification method for angle position of automobile picture |
| CN110570469B (en) * | 2019-08-16 | 2020-08-25 | 广州威尔森信息科技有限公司 | Intelligent identification method for angle position of automobile picture |
| CN110689066A (en) * | 2019-09-24 | 2020-01-14 | 成都考拉悠然科技有限公司 | Training method combining face recognition data equalization and enhancement |
| CN112749724A (en) * | 2019-10-31 | 2021-05-04 | 阿里巴巴集团控股有限公司 | Method and equipment for training classifier and predicting application performance expansibility |
| CN111191122A (en) * | 2019-12-20 | 2020-05-22 | 重庆邮电大学 | Learning resource recommendation system based on user portrait |
| CN113128536A (en) * | 2019-12-31 | 2021-07-16 | 奇安信科技集团股份有限公司 | Unsupervised learning method, system, computer device and readable storage medium |
| WO2021189830A1 (en) * | 2020-03-26 | 2021-09-30 | 平安科技（深圳）有限公司 | Sample data optimization method, apparatus and device, and storage medium |
| CN111553173A (en) * | 2020-04-23 | 2020-08-18 | 苏州思必驰信息科技有限公司 | Natural language generation training method and device |
| CN111553173B (en) * | 2020-04-23 | 2023-09-15 | 思必驰科技股份有限公司 | Natural language generation training method and device |
| CN111611457A (en) * | 2020-05-20 | 2020-09-01 | 北京金山云网络技术有限公司 | Page classification method, device, equipment and storage medium |
| CN111611457B (en) * | 2020-05-20 | 2024-01-02 | 北京金山云网络技术有限公司 | Page classification method, device, equipment and storage medium |
| CN111611388A (en) * | 2020-05-29 | 2020-09-01 | 北京学之途网络科技有限公司 | Account classification method, device and equipment |
| CN111783869A (en) * | 2020-06-29 | 2020-10-16 | 杭州海康威视数字技术股份有限公司 | Training data screening method and device, electronic equipment and storage medium |
| CN111970584A (en) * | 2020-07-08 | 2020-11-20 | 国网宁夏电力有限公司电力科学研究院 | Method, device and equipment for processing data and storage medium |
| CN111860671A (en) * | 2020-07-28 | 2020-10-30 | 中山大学 | Classification model training method and device, terminal equipment and readable storage medium |
| CN114077860A (en) * | 2020-08-18 | 2022-02-22 | 鸿富锦精密电子(天津)有限公司 | Method and system for sorting parts before assembly, electronic device and storage medium |
| CN112861512A (en) * | 2021-02-05 | 2021-05-28 | 北京百度网讯科技有限公司 | Data processing method, device, equipment and storage medium |
| CN113487320A (en) * | 2021-06-28 | 2021-10-08 | 深圳索信达数据技术有限公司 | Fraud transaction detection method, device, computer equipment and storage medium |
| CN113408282B (en) * | 2021-08-06 | 2021-11-09 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for topic model training and topic prediction |
| CN113408282A (en) * | 2021-08-06 | 2021-09-17 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for topic model training and topic prediction |
| WO2023166747A1 (en) * | 2022-03-04 | 2023-09-07 | 日本電信電話株式会社 | Training data generation device, training data generation method, and program |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109145937A (en) | A kind of method and device of model training | |
| CN111523621B (en) | Image recognition method and device, computer equipment and storage medium | |
| CN110942072B (en) | Quality score based on quality assessment, detection model training and detection method and device | |
| CN109190646B (en) | A kind of data predication method neural network based, device and nerve network system | |
| CN110633745A (en) | Image classification training method and device based on artificial intelligence and storage medium | |
| CN104899579A (en) | Face recognition method and face recognition device | |
| CN113591527A (en) | Object track identification method and device, electronic equipment and storage medium | |
| CN110298297A (en) | Flame identification method and device | |
| CN104778230B (en) | A kind of training of video data segmentation model, video data cutting method and device | |
| CN113656582B (en) | Training method of neural network model, image retrieval method, device and medium | |
| CN110245679A (en) | Image clustering method, device, electronic equipment and computer readable storage medium | |
| CN113837308B (en) | Knowledge distillation-based model training method and device and electronic equipment | |
| CN108197177A (en) | Monitoring method, device, storage medium and the computer equipment of business object | |
| CN111783712A (en) | Video processing method, device, equipment and medium | |
| CN109214412A (en) | A kind of training method and device of disaggregated model | |
| CN111160959B (en) | User click conversion prediction method and device | |
| CN115062709A (en) | Model optimization method, device, equipment, storage medium and program product | |
| CN112712068B (en) | Key point detection method and device, electronic equipment and storage medium | |
| CN113806501B (en) | Training method of intention recognition model, intention recognition method and equipment | |
| CN115567371B (en) | Abnormity detection method, device, equipment and readable storage medium | |
| CN112884730B (en) | Cooperative significance object detection method and system | |
| CN115546554A (en) | Sensitive image identification method, device, equipment and computer readable storage medium | |
| CN115129671A (en) | Log detection method, log detection device and computer-readable storage medium | |
| CN115052154A (en) | Model training and video coding method, device, equipment and storage medium | |
| CN110188798B (en) | Object classification method and model training method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190104 |
|
| RJ01 | Rejection of invention patent application after publication |