CN109872058A - Multimedia crowd sensing excitation method for machine learning system - Google Patents
Multimedia crowd sensing excitation method for machine learning system Download PDFInfo
- Publication number
- CN109872058A CN109872058A CN201910096432.XA CN201910096432A CN109872058A CN 109872058 A CN109872058 A CN 109872058A CN 201910096432 A CN201910096432 A CN 201910096432A CN 109872058 A CN109872058 A CN 109872058A
- Authority
- CN
- China
- Prior art keywords
- user
- task
- platform
- machine learning
- perception
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本发明提出了一种用于机器学习系统的多媒体群智感知激励方法，步骤如下：S1、机器学习系统根据数据训练需求发布图片采集任务；S2、移动感知平台获取机器学习系统发布的任务，并向平台用户发布感知任务请求；S3、用户制定感知计划并上传到平台；S4、移动感知平台选择获胜用户、计算获胜用户的报酬，并将获胜情况和报酬信息发送给用户；S5、获胜用户在规定时间内执行任务，将采集到的图片集上传到平台；S6、平台核实用户上传的数据，核实无误后支付报酬；S7、平台将所有图片集整理上传到机器学习系统作为机器学习系统的训练数据。本发明方法可以在较低成本的基础上为机器学习系统提供大量的高质量数据，帮助机器学习系统不断成长和优化。
The present invention provides a multimedia crowd-sensing incentive method for a machine learning system. The steps are as follows: S1. The machine learning system issues a picture collection task according to data training requirements; S2. The mobile perception platform acquires the task released by the machine learning system, and Issue a perception task request to platform users; S3, the user makes a perception plan and uploads it to the platform; S4, the mobile perception platform selects the winning user, calculates the winning user's reward, and sends the winning status and reward information to the user; S5, the winning user is in Perform the task within the specified time and upload the collected picture set to the platform; S6. The platform verifies the data uploaded by the user and pays the compensation after verification; S7. The platform organizes and uploads all the picture sets to the machine learning system as the training of the machine learning system data. The method of the invention can provide a large amount of high-quality data for the machine learning system on the basis of lower cost, and help the machine learning system to grow and optimize continuously.
Description
Technical field
The present invention relates to a kind of multimedia intelligent perception motivational techniques, belong to intelligent perception technical field.
Background technique
Machine learning model is a part important in artificial intelligence technology, the machine learning tentatively put up for one
Model, there are two types of the modes for promoting it to grow up: first is that optimization algorithm, by continuously improving algorithm come Optimized model frame；Two
It is that the extensive quality data collection of input is trained, provides endlessly data set support for the training study of model.It is existing
Some machine learning research focuses primarily upon the improvement of algorithm and ignores the important function that data set training grows up to model.One
A machine learning model tentatively built wants Fast Growth in a short time, it is necessary to come by the data set of magnanimity continuous
Training study.Studies have shown that still immature machine learning model once has a large amount of training data, one can be often defeated completely
A well-designed high level model for being based only upon low volume data and training, therefore the machine learning of " data " as forefront instantly
Key components, influence power can not be ignored.
Traditional data collection mode is obtained because the reasons such as coverage deficiency, high cost cost can not meet nowadays data
The demand taken, the mobile gunz cognition technology to come into being on the basis of crowdsourcing (Mobile Crowd Sensing, abbreviation MCS)
It can assist the collection work of completion large-scale data.MCS gives full play to the characteristics of There is strength in numbers, is joined by a large number of users
With can complete the collecting work of data in a very short period of time.It is a large amount of without spending without sending special staff
Time, compared to the smart machine specially disposed, MCS greatly reduces the cost of task and the maintenance cost of great number.
Most MCS system is voluntarily participated in based on user, lacks effective incentive mechanism and user is encouraged to participate in it
In, this often leads to perception task participant deficiency and submits the irregular of the quality of data.For a user, as participation
Perception task generally requires to pay through the nose, and in terms of time, energy, to acquire mobile device in some perceptions
Built-in sensing data, such as GPS, microphone, camera, this will threaten to privacy of user and disappear along with electricity
Consumption, flow consumption etc.；There are also part perception tasks to carry out pretreatment work in the mobile client of user, can occupy mobile phone
A large amount of computing resource.Therefore, if without reasonable incentive mechanism, perception task will be unable to complete on demand on time.In addition, sense
Know that task necessary not only for a large amount of data, also requires the quality of data, the data of high quality can help machine learning
System preferably learns, and improves the performance of system, improves the accuracy of data processing, so while improving user's participation rate
It is also required to guarantee the quality of data that user is finally uploaded to platform.
Current incentive mechanism mostly uses the remuneration payment incentive of money formula, and most important money formula energisation mode is auction
Mechanism, by using participant to the quotation of perception data as important measure index, lower participant's of selection payment cost
Collection is to complete perception task.Koutsopoulos I et al. seeks system using Bayesian game and user benefit is maximumlly equal
Weighing apparatus；Gao et al. devises the VCG auction of online user's selection based on Lyapunovv while providing long-term participate in
Mechanism.Quality of data demand is not all included in the design of incentive mechanism by above-mentioned mechanism, later, Krontiris et al. design
When mechanism user's participation rate and perception data quality both of these problems are considered using MAA model simultaneously, and is returned by different
Report pays that user is motivated to upload quality data；Wen et al. proposes the quality of data of the indoor locating system based on fingerprint
The incentive mechanism of driving, it is indicated that the remuneration for paying participant should be based on the quality of data rather than complete the required by task time, and use
Monotonically increasing function model indicates the mapping relations of financial value that the quality of data and server obtain.But these excitations at present
Method is assumed to a uniform variable to the measurement of the quality of data, is not bound with the shadow that specific application scenarios consider other factors
It rings.
Summary of the invention
The invention proposes a kind of multimedia intelligent perception motivational techniques for machine learning system, by machine learning system
The data conversion needed unite into the perception task of mobile awareness platform, perception task is issued in platform, is inhaled using auction mechanism
It quotes family and participates in task, comprehensively consider user and submit the factors such as quality, timeliness, coverage rate, the correlation of data, choose and win
User completes perception task.
In order to solve the above technical problems, present invention employs following technological means:
A kind of multimedia intelligent perception motivational techniques for machine learning system, comprising the following steps:
S1, machine learning system issue picture collection task according to data training demand；
S2, mobile awareness platform obtain the task of machine learning system publication, and ask to platform user publication perception task
It asks；
S3, user request to formulate perception plan and upload to platform according to perception task；
S4, mobile awareness platform select winning user according to the perception plan of historical information and user on platform, and count
The remuneration for calculating winning user, is sent to user for triumph situation and remuneration information；
S5, winning user execute task at the appointed time, and collected pictures are uploaded to platform；
S6, platform perceive plan according to user and verify the pictures that user uploads, payt after verification is errorless；
The pictures arrangement that all users acquire is uploaded to machine learning system as engineering by S7, mobile awareness platform
The training data of learning system.
Further, in the step 1 picture collection task include picture type, each type picture number, every
The picture gross mass of a type.
Further, the step S3 user requests to formulate perception plan according to perception task, perceives the specific of plan
Expression formula is as follows:
bidi=(ηi,bi) (1)
Wherein, bidiIndicate the perception plan of i-th of user, ηiIndicate that i-th of user plans the task-set executed, task
Concentrate includes that user plans the task type executed and task quantity, biIndicate that i-th user completes the quotation of task, i=1,
2 ..., m, m represent the number of participating user.
Further, the step S4 concrete operations are as follows:
S41, mobile awareness platform according to the corresponding perception task of picture collection term of reference of machine learning system and
Shi XingCorrelationAnd coverage rateWherein, T1Indicate job start time, T2Expression task is completed or task terminates
Time.
S42, platform calculate the acquisition quality of data q of user i according to historical informationi, participate in task timeliness (T1-T2)i、
Task dependencies αiWith task coverage rate βi。
S43, platform successively compare the data in S1 and S2, reject from the set U that all users form and are unsatisfactory for task
Defined timelinessCorrelationOr coverage rateUser, obtain new user's set N.
S44, the Mass Calculation user that data are acquired according to user collect the marginal social welfare of each user in N:
wj=λ qj|ηj|-bj (2)
Wherein, wjIndicate the marginal social welfare of j-th of user in set N, λ is to convert Task Quality to turning for money
Change coefficient, qjFor the acquisition quality of data of j-th of user in set N, | ηj| for j-th of user number of tasks to be executed in set N
Amount, bjIt indicates the quotation of j-th of user's completion task in set N, shares n user, j=1,2 ..., n in set N.
S45, marginal social welfare w in user's collection N is selectedj>=0 user is stored in first winning user set omega+, and will
Marginal social welfare wjThe user of < 0 is saved in set omega-In.
The difference for the gross mass that S46, the gross mass of computing platform release tasks and first winning user provide, is put down
The residual mass Q ' of the task of platform publication.
S47, set omega is successively chosen-The smallest user of middle limit social welfare efficiency ξ is stored in second batch winning user collection
S ' is closed, until the residual mass demand of each subtask is all satisfied in perception task, the calculating of marginal social welfare efficiency ξ
Formula is as follows:
Wherein, τlIndicate first of subtask, l=1,2 ..., L, L represents the total number of subtask, ηjIt indicates the in set N
J user plans the task-set executed, Q 'lIndicate the residual mass of first of subtask.
S48, set omega is set+In have S user, have K user, set omega in set S '+Triumph is collectively constituted with set S '
User set W, first winning user set omega+In each user remuneration calculation formula it is as follows:
p1=bs (4)
Wherein, bsFor set omega+In the perception of s-th winning user complete the quotation of task, s=1,2 ..., S in the works.
The remuneration calculation formula of each user in second batch winning user set S ' is as follows:
Wherein, qkFor the acquisition quality of data of k-th of winning user in set S ', | ηk| it is used for k-th of triumph in set S '
The task quantity that family plan executes, wkIndicate the marginal social welfare of the middle user k of set S ', ηkIndicate appointing for the middle user k of set S '
Business collection, k=1,2 ..., K.
S49, triumph situation and remuneration are sent to user, the user's remuneration that do not win is 0.
Further, the concrete operations of the step S42 are as follows:
A, the resolution ratio of the picture of platform is uploaded to according to user's history, calculates the average quality of the acquisition data of user i
qi, the acquisition quality of data of the higher user of history photo resolution is higher.
B, the time according to as defined in the time of the completion task of user in history and platform calculates user i and participates in the timely of task
Property (T1-T2)i。
C, the task dependencies α of user i is calculatedi:
αi=α (t-d)=[2sgn (t-d) f (d-t)+sgn (d-t)] mxy (6)
Wherein, α () is monotonic decreasing function, and value range is that 0 to 1, t is the time that historic task is added in user, and d is
Deadline as defined in historic task, sgn () are sign function, and f (t) is sigmoid function, mxyIt is platform to user's history
Acquire the marking of picture and the mission requirements degree of association.
D, the task coverage rate β of user i is calculatedi:
Wherein, r is user in user's history acquisition picture at a distance from target object.
Using following advantage can be obtained after the above technological means:
The invention proposes a kind of multimedia intelligent perception motivational techniques for machine learning system, mainly for machine
Data requirements is issued in mobile awareness platform, is adopted the case where needing a large amount of image datas in the training study stage by learning system
A large number of users is attracted to participate in the collecting work of image data with the incentive mechanism of auction, the method for the present invention is in selection winning user
In the process, the performance of other tasks participated in this platform history in conjunction with the user comprehensively considers the figure of user's acquisition
The factors such as the timeliness that the quality of piece, coverage rate, the correlation and task of picture and mission requirements are completed, it is ensured that selection can mention
The user of high higher quality data participates in task, further improves the matter of picture on the basis of guaranteeing picture collection quantity
Amount, avoids negative effect of the low quality image data to machine learning system.The method of the present invention can be in the base of lower cost
A large amount of quality data is provided for machine learning system on plinth, machine learning system is helped constantly to grow up and optimize.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of the multimedia intelligent perception motivational techniques for machine learning system of the present invention.
Fig. 2 is to choose winning user in a kind of multimedia intelligent perception motivational techniques for machine learning system of the present invention
Flow chart of steps.
Specific embodiment
Technical solution of the present invention is described further with reference to the accompanying drawing:
The invention proposes a kind of multimedia intelligent perception motivational techniques for machine learning system, as shown in Figure 1, tool
Body the following steps are included:
S1, machine learning system issue picture collection task according to data training demand；The method of the present invention is for meeting machine
The acquisition demand of image data in device learning system, for machine learning system to the different demands of picture, the picture of publication is adopted
Set task mainly includes the picture gross mass for the type of the picture and each type for needing to acquire, for example, flowers for identification
Machine learning system, picture collection task include rose, Jasmine and lily, and the gross mass of every kind of flowers is all 100.
S2, mobile awareness platform obtain the task of machine learning system publication, and the task publication of machine learning system is existed
On platform, corresponding perception task request is generated, the user on platform can check and sign up for the perception task.
S3, user request to formulate perception plan and upload to platform according to perception task；User checks perception task
Perception plan can be formulated after content according to own situation, perception plan is made of a binary group, and expression is as follows:
bidi=(ηi,bi) (8)
Wherein, bidiIndicate the perception plan of i-th of user, ηiIndicate that i-th of user plans the task-set executed, task
Concentrating includes that user plans the task type executed and task quantity, is opened for example, user's plan provides rose picture 10；biTable
Show that i-th of user completes the quotation of task, i.e. user it is expected the remuneration obtained, if shared m user participates in perception plan, i=
1,2,…,m。
S4, mobile awareness platform select winning user, Fig. 2 exhibition according to the perception plan of historical information and user on platform
The step of having shown selection winning user calculates the remuneration of winning user after determining winning user, by the triumph situation of user and
Remuneration information is sent to the user；Concrete operations are as follows:
S41, mobile awareness platform according to the corresponding perception task of picture collection term of reference of machine learning system and
Shi XingCorrelationAnd coverage rateWherein, T1Indicate job start time, T2Expression task is completed or task terminates
Time.
S42, platform calculate the acquisition quality of data q of user i according to historical informationi, participate in task timeliness (T1-T2)i、
Task dependencies αiWith task coverage rate βi, concrete operations are as follows:
A, the history picture that user uploads on platform is extracted, the resolution ratio of each history picture is obtained, user i is calculated and adopts
The average quality q of the image data of collectioni, the acquisition quality of data of the higher user of history photo resolution is higher.
B, the time according to as defined in the time of the completion task of user in history and platform calculates user i and participates in the timely of task
Property (T1-T2)i。
C, the task dependencies α of user i is calculatedi:
αi=α (t-d)=[2sgn (t-d) f (d-t)+sgn (d-t)] mxy (9)
Wherein, α () is monotonic decreasing function, and value range is that 0 to 1, t is the time that historic task is added in user, and d is
Deadline as defined in historic task, sgn () are sign function, and f (t) is sigmoid function, mxyIt is platform to user's history
Acquire the marking of picture and the mission requirements degree of association.
D, the task coverage rate β of user i is calculatedi:
Wherein, r is that for user at a distance from target object, user is same when uploading pictures in user's history acquisition picture
When need the relevant informations of uploading pictures, the distance etc. of picture is shot including user.
S43, platform successively compare the data in S1 and S2, if the timeliness (T of user i1-T2)i, correlation αi, covering
Rate βiIn have any one be less than term of reference timelinessCorrelationOr coverage rateThen the user is unsatisfactory for wanting
It asks.The timeliness for being unsatisfactory for term of reference is rejected from the set U that all users formCorrelationOr coverage rate
User, obtain new user's set N.
N user is shared in S44, set N, and each user in N is collected according to the Mass Calculation user that user acquires data
Marginal social welfare:
wj=λ qj|ηj|-bj (11)
Wherein, wjIndicate the marginal social welfare of j-th of user in set N, λ is to convert Task Quality to turning for money
Change coefficient, qjFor the acquisition quality of data of j-th of user in set N, | ηj| for j-th of user number of tasks to be executed in set N
Amount, bjIndicate the quotation of j-th of user's completion task in set N, j=1,2 ..., n.
S45, marginal social welfare w in user's collection N is selectedj>=0 user is stored in first winning user set omega+, and will
Marginal social welfare wjThe user of < 0 is saved in set omega-In.
The difference for the gross mass that S46, the gross mass of computing platform release tasks and first winning user provide, is put down
The residual mass Q ' of the task of platform publication.
S47, set omega is successively chosen-The smallest user of middle limit social welfare efficiency ξ is stored in second batch winning user collection
S ' is closed, according to the residual mass of the perception each subtask of schedule regeneration of the user, until all subtasks in perception task
Residual mass demand is all satisfied, and the calculation formula of marginal social welfare efficiency ξ is as follows:
Wherein, τlIndicate first of subtask, l=1,2 ..., L, L represents the total number of subtask, ηjIt indicates the in set N
J user plans the task-set executed, Q 'lIndicate the residual mass of first of subtask.
S48, set omega is set+In have S user, have K user, set omega in set S '+Triumph is collectively constituted with set S '
User set W, first winning user set omega+In each user remuneration calculation formula it is as follows:
p1=bs (13)
Wherein, bsFor set omega+In the perception of s-th winning user complete the quotation of task, s=1,2 ..., S in the works.
The remuneration calculation formula of each user in second batch winning user set S ' is as follows:
Wherein, qkFor the acquisition quality of data of k-th of winning user in set S ', | ηk| it is used for k-th of triumph in set S '
The task quantity that family plan executes, wkIndicate the marginal social welfare of the middle user k of set S ', ηkIndicate appointing for the middle user k of set S '
Business collection, k=1,2 ..., K.
S49, triumph situation and remuneration are sent to user, the user not won informs that it is not won, remuneration 0.
S5, winning user execute task at the appointed time, and collected pictures are uploaded to platform, when being more than regulation
Between the pictures that upload not adopt.
S6, platform perceive plan according to user and verify the pictures that user uploads, and only have a feeling of satisfaction and know plan, provide foot
The user of enough quality pictures can just obtain corresponding remuneration, and the user for failing to provide enough quality pictures at the appointed time can not
It obtains remuneration or can only obtain and upload the corresponding remuneration of quality.
The pictures arrangement that all users acquire is uploaded to machine learning system as engineering by S7, mobile awareness platform
The training data of learning system.
Embodiments of the present invention are explained in detail above in conjunction with attached drawing, but the invention is not limited to above-mentioned
Embodiment within the knowledge of a person skilled in the art can also be before not departing from present inventive concept
It puts and makes a variety of changes.
Claims (5)
1. a kind of multimedia intelligent perception motivational techniques for machine learning system, which comprises the following steps:
S1, machine learning system issue picture collection task according to data training demand；
S2, mobile awareness platform obtain the task of machine learning system publication, and request to platform user publication perception task；
S3, user request to formulate perception plan and upload to platform according to perception task；
S4, mobile awareness platform select winning user according to the perception plan of historical information and user on platform, and calculate and obtain
The remuneration for winning user, is sent to user for triumph situation and remuneration information；
S5, winning user execute task at the appointed time, and collected pictures are uploaded to platform；
S6, platform perceive plan according to user and verify the pictures that user uploads, payt after verification is errorless；
The pictures arrangement that all users acquire is uploaded to machine learning system as machine learning system by S7, mobile awareness platform
The training data of system.
2. a kind of multimedia intelligent perception motivational techniques for machine learning system according to claim 1, feature
It is, picture collection task includes the picture gross mass of picture type and each type in the step 1.
3. a kind of multimedia intelligent perception motivational techniques for machine learning system according to claim 1, feature
It is, the step S3 user requests to formulate perception plan according to perception task, and the expression for perceiving plan is as follows:
bidi=(ηi,bi)
Wherein, bidiIndicate the perception plan of i-th of user, ηiIndicate that i-th user plans the task-set executed, in task-set
Plan the task type executed and task quantity, b including useriIndicate that i-th user completes the quotation of task, i=1,2 ...,
M, m represent the number of participating user.
4. a kind of multimedia intelligent perception motivational techniques for machine learning system according to claim 1, feature
It is, the step S4 concrete operations are as follows:
S41, mobile awareness platform are according to the timeliness of the corresponding perception task of picture collection term of reference of machine learning systemCorrelationAnd coverage rateWherein, T1Indicate job start time, T2Expression task is completed or job end time；
S42, platform calculate the acquisition quality of data q of user i according to historical informationi, participate in task timeliness (T1-T2)i, task
Correlation αiWith task coverage rate βi；
S43, platform reject the timeliness for being unsatisfactory for term of reference from the set U that all users formCorrelationOr
Coverage rateUser, obtain new user's set N；
S44, the Mass Calculation user that data are acquired according to user collect the marginal social welfare of each user in N:
wj=λ qj|ηj|-bj
Wherein, wjIndicate the marginal social welfare of j-th of user in set N, λ is the transformation system for converting Task Quality to money
Number, qjFor the acquisition quality of data of j-th of user in set N, | ηj| it is j-th of user task quantity to be executed in set N,
bjIt indicates the quotation of j-th of user's completion task in set N, shares n user, j=1,2 ..., n in set N；
S45, marginal social welfare w in user's collection N is selectedj>=0 user is stored in first winning user set omega+, and will be marginal
Social welfare wjThe user of < 0 is saved in set omega-In；
The difference for the gross mass that S46, the gross mass of computing platform release tasks and first winning user provide obtains platform hair
The residual mass Q ' of the task of cloth；
S47, the middle marginal the smallest user's deposit second batch winning user set S ' of social welfare efficiency ξ of set omega-is successively chosen,
Until the residual mass demand of each subtask is all satisfied in perception task, the calculation formula of marginal social welfare efficiency ξ
It is as follows:
Wherein, τlIndicate first of subtask, l=1,2 ..., L, L represents the total number of subtask, ηjIt indicates in set N j-th
User plans the task-set executed, Q 'lIndicate the residual mass of first of subtask；
S48, set omega is set+In have S user, have K user, set omega in set S '+Winning user is collectively constituted with set S '
Set W, first winning user set omega+In each user remuneration calculation formula it is as follows:
p1=bs
Wherein, bsFor set omega+In the perception of s-th winning user complete the quotation of task, s=1,2 ..., S in the works；
The remuneration calculation formula of each user in second batch winning user set S ' is as follows:
Wherein, qkFor the acquisition quality of data of k-th of winning user in set S ', | ηk| for k-th of winning user meter in set S '
Draw the task quantity executed, wkIndicate the marginal social welfare of the middle user k of set S ', ηkIndicate the task of the middle user k of set S '
Collection, k=1,2 ..., K；
S49, triumph situation and remuneration are sent to user, the user's remuneration that do not win is 0.
5. a kind of multimedia intelligent perception motivational techniques for machine learning system according to claim 4, feature
It is, the concrete operations of the step S42 are as follows:
A, the resolution ratio of the picture of platform is uploaded to according to user's history, calculates the average quality q of the acquisition data of user ii；
B, the time according to as defined in the time of the completion task of user in history and platform calculates the timeliness that user i participates in task
(T1-T2)i；
C, the task dependencies α of user i is calculatedi:
αi=α (t-d)=[2sgn (t-d) f (d-t)+sgn (d-t)] mxy
Wherein, α () is monotonic decreasing function, and value range is that 0 to 1, t is the time that historic task is added in user, and d is history
The deadline of term of reference, sgn () are sign function, and f (t) is sigmoid function, mxyUser's history is acquired for platform
The marking of picture and the mission requirements degree of association；
D, the task coverage rate β of user i is calculatedi:
Wherein, r is user in user's history acquisition picture at a distance from target object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910096432.XA CN109872058B (en) | 2019-01-31 | 2019-01-31 | A Multimedia Crowdsensing Motivation Method for Machine Learning Systems |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910096432.XA CN109872058B (en) | 2019-01-31 | 2019-01-31 | A Multimedia Crowdsensing Motivation Method for Machine Learning Systems |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109872058A true CN109872058A (en) | 2019-06-11 |
| CN109872058B CN109872058B (en) | 2022-12-13 |
Family
ID=66918369
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910096432.XA Active CN109872058B (en) | 2019-01-31 | 2019-01-31 | A Multimedia Crowdsensing Motivation Method for Machine Learning Systems |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109872058B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111754000A (en) * | 2020-06-24 | 2020-10-09 | 清华大学 | Quality-aware edge intelligent federated learning method and system |
| CN112163751A (en) * | 2020-09-21 | 2021-01-01 | 百度在线网络技术（北京）有限公司 | Online learning incentive method, device, equipment and storage medium |
| CN112990268A (en) * | 2021-02-09 | 2021-06-18 | 大理大学 | Machine learning-oriented method and system for evaluating contribution degree of visual crowd sensing data |
| CN114780229A (en) * | 2021-01-05 | 2022-07-22 | 中国移动通信有限公司研究院 | A processing method, device and electronic device for a perception task |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107301509A (en) * | 2017-06-23 | 2017-10-27 | 武汉大学 | It is a kind of based on intelligent perception system towards the intelligent perception motivational techniques participated at random |
| CN108269129A (en) * | 2018-01-23 | 2018-07-10 | 重庆邮电大学 | User's motivational techniques in a kind of mobile gunz sensing network based on reverse auction |
| CN109242533A (en) * | 2018-08-04 | 2019-01-18 | 福州大学 | The online motivational techniques of car networking intelligent perception user based on Game Theory |
-
2019
- 2019-01-31 CN CN201910096432.XA patent/CN109872058B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107301509A (en) * | 2017-06-23 | 2017-10-27 | 武汉大学 | It is a kind of based on intelligent perception system towards the intelligent perception motivational techniques participated at random |
| CN108269129A (en) * | 2018-01-23 | 2018-07-10 | 重庆邮电大学 | User's motivational techniques in a kind of mobile gunz sensing network based on reverse auction |
| CN109242533A (en) * | 2018-08-04 | 2019-01-18 | 福州大学 | The online motivational techniques of car networking intelligent perception user based on Game Theory |
Non-Patent Citations (1)
| Title |
|---|
| 钟山: "移动群智感知中分阶式激励与惩罚机制研究", 《CNKI优秀硕士论文全文库》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111754000A (en) * | 2020-06-24 | 2020-10-09 | 清华大学 | Quality-aware edge intelligent federated learning method and system |
| CN111754000B (en) * | 2020-06-24 | 2022-10-14 | 清华大学 | Quality-aware edge intelligent federal learning method and system |
| CN112163751A (en) * | 2020-09-21 | 2021-01-01 | 百度在线网络技术（北京）有限公司 | Online learning incentive method, device, equipment and storage medium |
| CN112163751B (en) * | 2020-09-21 | 2024-04-09 | 百度在线网络技术（北京）有限公司 | Online learning excitation method, device, equipment and storage medium |
| CN114780229A (en) * | 2021-01-05 | 2022-07-22 | 中国移动通信有限公司研究院 | A processing method, device and electronic device for a perception task |
| CN114780229B (en) * | 2021-01-05 | 2025-05-13 | 中国移动通信有限公司研究院 | A method, device and electronic device for processing a perception task |
| CN112990268A (en) * | 2021-02-09 | 2021-06-18 | 大理大学 | Machine learning-oriented method and system for evaluating contribution degree of visual crowd sensing data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109872058B (en) | 2022-12-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109872058A (en) | Multimedia crowd sensing excitation method for machine learning system | |
| Padhi et al. | Competitive revenue management for fixing quota and price of hotel commodities under uncertainty | |
| CN110189174A (en) | Mobile crowd sensing excitation method based on data quality sensing | |
| CN109068288B (en) | Method and system for selecting mobile crowd sensing incentive mechanism based on multi-attribute user | |
| Nan et al. | A cross-space, multi-interaction-based dynamic incentive mechanism for mobile crowd sensing | |
| CN109410054B (en) | Block chain consensus method based on information sharing contribution value in autonomous domain mode | |
| CN108537691A (en) | A kind of region visit intelligent management system and method | |
| CN107784561A (en) | The implementation method of online incentive mechanism in a kind of mobile mass-rent system | |
| CN107146179B (en) | Training and team-assembling method | |
| CN109242533A (en) | The online motivational techniques of car networking intelligent perception user based on Game Theory | |
| CN103930914B (en) | Method and system for generating occupant's timetable | |
| Zhao et al. | The innovation economy calls for proactive growth of intellectual property by various innovation carriers–A China case | |
| Liang et al. | Operation-adversarial scenario generation | |
| Wang et al. | Reinforcement contract design for vehicular-edge computing scheduling and energy trading via deep Q-network with hybrid action space | |
| CN115936361A (en) | Online scheduling method and related device for federated learning demand response of cloud edge system | |
| Ren et al. | Efficient cross dynamic task assignment in spatial crowdsourcing | |
| CN115271092A (en) | A crowdfunding incentive method for indoor positioning federated learning | |
| Yang et al. | Analysis of offering behavior of generation-side integrated energy aggregator in electricity market: A Bayesian evolutionary approach | |
| Gu et al. | QoI-aware incentive for multimedia crowdsensing enabled learning system | |
| CN113139792B (en) | Specific material collection method based on crowd sensing technology | |
| Parkes et al. | Dynamic incentive mechanisms | |
| CN116523071A (en) | Group crowd funding game continuous excitation method for federal learning | |
| Li et al. | An efficient, fair, and robust image pricing mechanism for crowdsourced 3D reconstruction | |
| CN113128955B (en) | Electronic government affair processing method and system based on intelligent measurement | |
| CN115018086B (en) | Model training method based on federal learning and federal learning system |
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 |