CN112416579B - Time-sensitive multiparty data fusion excitation method - Google Patents
Time-sensitive multiparty data fusion excitation method Download PDFInfo
- Publication number
- CN112416579B CN112416579B CN202011243280.0A CN202011243280A CN112416579B CN 112416579 B CN112416579 B CN 112416579B CN 202011243280 A CN202011243280 A CN 202011243280A CN 112416579 B CN112416579 B CN 112416579B
- Authority
- CN
- China
- Prior art keywords
- winner
- relative
- participant
- target
- subtask
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a time-sensitive multiparty track data fusion excitation method, and belongs to the field of distributed computation. The method comprises the steps of initializing, determining allocation rules and price rules, firstly describing topological dependence relation of multiparty track data fusion by an AOE network, collecting target information from participants, then approximately solving social benefit maximization problem under time limit and deriving corresponding allocation results. The invention comprehensively determines the priority sequence of each subtask in the greedy flow by combining two different key values of gain increment and gain cost increment ratio and two different ordering modes of global comparison and long-chain priority. The payment result of the invention is determined by the distribution result and the Mierson theorem, thereby ensuring the dominant strategy excitation compatibility characteristic of the overall mechanism.
Description
Technical Field
The invention relates to the field of distributed computation, in particular to a time-sensitive multiparty data fusion excitation method.
Background
The motivation technology is a core method for fairly and reasonably exchanging and distributing resources, the main content of the motivation technology is reasonable mechanism design, and the motivation technology has been widely paid attention to academia and industry for a long time. The mechanism design originated in the 60 s of the last century. The concept of dominant policy incentive compliance was formally proposed by Hurwicz in 1972. The primary revenue source for search engines is keyword auctions, which use the generalized divalent auction (GSP) model. Whereas knapsack auctions, which are widely used in the field of resource allocation, are proposed by MuAlem and Nisan. On the other hand, the field of algorithm mechanism design based on approximation algorithm is also rapidly developing. Briest gives a computationally feasible and well-behaved knapsack auction mechanism based on the classical FPTAS approximation algorithm of the knapsack problem. Similarly, lehmann et al have realized a well-behaved mechanism using greedy algorithms based on aggregate coverage problems.
These incentive techniques are applied in many ways, including spectrum allocation, crowdsourcing, medical treatment, and the like. There are a large number of incentive mechanisms in the crowdsourcing field, but most of them are designed for mobile scenes, analyze incentive problems in data acquisition and sharing processes, and mainly focus on data sharing, which is inconsistent with multiparty data fusion computing scenes. In the federal learning field of multiparty calculation, the existing method mainly solves the algorithm performance and calculation safety of multiparty calculation, and most of the existing methods assume that all parties participating in calculation have enthusiasm for actively providing calculation and data, and have little research on an excitation mechanism.
In the big data age, distributed computing is an effective means of comprehensively utilizing data and computing resources. Multiparty computing is an important method for fully exploiting the advantages of each and creating greater value. The introduction of the incentive technology can guarantee that the participants truly express the demand degree of themselves for the calculation task under the assumption of a physical person. By combining a distribution mechanism of social benefit maximization, fair and effective task distribution can be ensured, the overall efficiency of the system is optimized, and the smooth proceeding of data fusion is effectively promoted.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: how to set a reasonable mechanism to stimulate participants to truly express demands, and at the same time maximize the total utility of the system as much as possible, namely effectively determine specific undertakers of data fusion calculation, so as to effectively promote smooth execution of data fusion.
The invention discloses a time-sensitive multiparty data fusion excitation method, which specifically comprises the following steps:
step 1: initializing:
modeling the multiparty data fusion Task by adopting an AOE network, and determining the internal topological relation of each sub-calculation Task;
each directed Edge < u, v > On the AOE network (Activity On Edge) represents a sub-calculation task, u and v represent two synchronous waiting nodes respectively, each synchronous waiting node in the AOE network can start all sub-tasks of the outgoing Edge after waiting for the sub-tasks of all incoming edges to finish, a source point and a sink point in the constructed topological graph represent an initial state and an end state respectively, the outgoing degree or the incoming degree of the intermediate node is not 0, namely, other intermediate nodes with the outgoing degree or the incoming degree of 0 are not existed in the topological graph except the source point and the sink point;
step 2: determining an allocation rule:
a heuristic algorithm based on greedy ideas is adopted to determine corresponding allocation rules allocation (bid):
for each sub-computing task, deleting all inferior participants in the competition list, namely those participants with lower real targets and higher computing time length, so that the real targets of the rest participants can be monotonically not reduced along with the computing time length;
initializing a reference solution, the winner of each subtask being its reference participant;
constructing a relative target, wherein the value of the relative target represents the benefit brought by replacing the reference participant; and ordering the relative target sequences;
there are two calculation methods for the relative target value:
mode one: total gain increment;
mode two: increment of income brought by increment of unit duration;
step 3: determining price rules based on the michelson theorem:
setting the allocation of participants to a vector of 0-1, i.e., the allocation of winners to 1, and the allocation of non-winners to 0;
a threshold target for winning allocated from 1 to 0 is calculated based on the milsen theory, the threshold target is set as the price to be paid by the winner, and the payments by non-winners are set to 0.
That is, the price rule formulation of the present invention relies on the allocation rule and the mylson Myerson theory. Under the condition that the allocation rule meets monotonicity, the corresponding price rule is directly calculated. Thereby ensuring the dominant strategic incentive compliance characteristics of the overall auction mechanism.
Further, in step 2, the sorting of the relative target sequence includes the following two methods:
global contrast:
directly sorting the relative targets of all non-reference participants according to the relative target values, wherein the higher the relative target value is, the higher the participant priority is;
long chain priority:
performing edge set division on the AOE, wherein each division subset represents a directed path;
the method comprises the steps that divided subsets with more edges are preferentially considered, relative targets of participants belonging to the same divided subsets are located in the same section, section ordering positions corresponding to the divided subsets with more edges are located at the front, and then the interior of the same section is arranged from large to small according to relative target values;
executing greedy flow according to the ordered relative target sequence: sequentially trying to replace a reference participant by each corresponding participant with a relative target from front to back, and running a longest path algorithm of a directed acyclic graph DAG to judge whether the time limit is met, if so, updating the solution, wherein the current participant becomes a winner of the subtask to which the current participant belongs; if not, continuing subsequent greedy; wherein winners of each subtask are updated at most once.
Further, in step 3, a critical target of the winner is calculated from 1 to 0 based on the milsen approach, and the price paid as the winner by the critical target is specifically:
1) If the winner is a non-reference participant,
deleting the relative targets of all participants of the subtask to which the winner belongs in the original relative target sequence; maintaining the other relative targets unchanged, and then operating a greedy flow in the allocation rule to obtain a series of new critical greedy results;
based on the obtained new critical greedy result, the current winner is sequentially replaced by the corresponding reference participant, the relative target which cannot be continuously winned by the current winner is found, and then the relative target is converted into a real target b ans The current price that the winner needs to pay is: max (b) ans ,b next ) Wherein b next Is the true target of the participant belonging to the same subtask as the current winner, and the relative target value is less than or equal to the maximum value of the winner;
2) If the winner is the reference participant,
since decreasing the current winner's target means increasing the relative targets of other participants of the same subtask at the same time, it is only necessary to continuously decrease the current winner's target, run the greedy algorithm in the allocation rule flow until the current subtask's reference participant is replaced, at which point the current winner's target is the price that it should pay.
In summary, due to the adoption of the technical scheme, the beneficial effects of the invention are as follows: the invention provides a time-sensitive multiparty data fusion excitation method, which provides support for task allocation problems. The method is particularly suitable for fusion processing of multi-party track data.
Drawings
Fig. 1 is an overall flow diagram of the present invention.
Fig. 2 is an example of trajectory data fusion.
Fig. 3 is an AOE task network used by the present invention.
FIG. 4 is an example of AOE network edge set partitioning.
Fig. 5 is an example of two sort strategies.
Detailed Description
The present invention will be described in further detail with reference to the embodiments and the accompanying drawings, for the purpose of making the objects, technical solutions and advantages of the present invention more apparent.
Referring to fig. 1, in the time-sensitive multi-party track data fusion excitation method of the present invention, firstly, an AOE network is used to describe the topological dependency relationship of multi-party track data fusion, and target information from participants is collected; namely, after the task information is released, receiving a participant submit target; and then carrying out distribution and payment processing, wherein the invention adopts approximate solution of the social benefit maximization problem under the time limit and derives the corresponding distribution result. Because of the NP difficulty of the problem, the invention comprehensively determines the priority order of each subtask in the greedy flow by combining two different ordering modes of global comparison and long-chain priority according to two different key values of gain increment and gain cost increment ratio. The payment result of the invention is determined by the distribution result and the Mierson theorem, thereby ensuring the dominant strategy excitation compatibility characteristic of the overall mechanism.
Referring to FIG. 2, for track dataFusion task T, which has a SubTask set SubTask, defines st= |subtask|. Because of the inherent structure of the geographical data distribution differences or data fusion sequences, the problem of synchronous waiting among the subtasks has a topological sequence relationship. To describe this relationship, the present invention models with an AOE network. Each SubTask i List of participants who have competing for the task i Each participant agent u ∈list i Relating to tasks SubTask i Is a secret value val of (1) i (i.e., true value) and the time required for the computing task to complete, timefiromagent i . Only one winner must be generated in each list to complete the calculation of the subtask. It is assumed here that all participants only participate in the competition of one subtask. The most important constraints are: after the winner of all subtasks is determined, the completion time finishedtime of the total task is less than or equal to a certain threshold timelimit.
Virtual tokens are used as currency in the mechanism. N participant agents in the mechanism can enter an auction link of the multiparty data fusion task T.
Participant agent i Possess a privacy value of 0val i INCENT-cost i Where INCENT is a fixed prize per unit of computational task awarded by the winner by the platform party, as determined by the multiparty data fusion task initiator. cost i Representing participant agents i And 0cost i INCENT, otherwise, participant agent i This auction should be exited to avoid damage to itself. I.e. val i Representing agent i Executing the real value brought by the corresponding multiparty data fusion task. Its val i For only participant agents i It is visible that the platfonn and the rest of the participants are not visible. In the mechanism, each participant will submit its own target b i A target vector bid is formed.
Allocation function allowanbid) is also a vector that represents the number of multiparty data fusion tasks that the mechanism allocates to the participants. agent i Is defined as a quasi-linear function utility i val i *allocation i -payment i . General social welfare SW medicineMeaning as
Which is the sum of the utility function of all participants and the platform side revenue. Namely: />
Wherein, allocation i Representing the assigned value, parameter i Representing platform side revenue, i.e., the price that the participant needs to pay.
In this embodiment, the specific flow steps of the auction mechanism are as follows:
s1: as shown in fig. 3, for the calculation task T, a Directed Acyclic Graph (DAG) d= (V, E) is constructed, set V being the set of all synchronization nodes. The directed edge < u, v > ∈E represents the progress of the subtask. The function t (e) related to the directed edge can be regarded as the weight of the directed edge in the DAG, and its value is the time period required for the subtask e to be calculated by the winner of the task.
S2: determining an allocation rule under a social benefit maximization optimization target by a heuristic algorithm, wherein the specific sub-flow is as follows:
s21: for any participant i, j, if val in any subtask competition list i ＜val j And timefiromagent i timefromagent j Then set allocation i 0, and to take the participant agent i And deleting from the competition list. Then list each competitor list i Each of the participants in (a) is incrementally ordered by timefiomagent key. At this time each list i Is referred to as the base participant for the subtask.
S22: initializing an allocation scheme S in which subtasks are SubTask i Is its base of reference participants.
S23: a determination is made as to whether the computing task is capable of being completed with a longest path algorithm.
The method specifically comprises the following steps: the total calculation time ftime=dp [ sink ] required for the allocation scheme S is calculated from the transfer equation dp [ i ] =max { dp [ k ] +t (< k, i >) | < k, i > ∈e }, dp [ source ] ] 0. Where dp [ i ] represents the length of the longest path ending with vertex i, and t (< k, i >) represents the distance of the directed edge between k, i. That is, first, dp [ i ] is accumulated based on dp [ i ] of each sub-task, and then dp [ sink ] is accumulated.
If ftime > finishedtime, the task cannot be completed within a given time limit, and error information is output. The algorithm terminates.
S24: initializing a relative target list as follows:
rlist＝{(relative i ,timefromagent i ) I1 is not less than i is not less than n, and agent i Not the reference participant },
relative target value relative i There are two calculation modes:
1) The relative target value is the overall benefit added by the replacement reference participant, relative i ＝bid i -bid base ，bid i Is a participant agent i Base is the participant agent i Reference participant, bid, of the subtask base Representing the true targets of the reference participants.
2) The relative target value is the ratio of the gain increment and the duration increment brought by the replacement reference participant, and relative i ＝(bid i -bid base 0/timefromagent i -timefromagent base )，bid i Is a participant agent i Base is the participant agent i Reference participant of the subtask, timefiomagent base Is the time required by the reference participant to complete the computing task.
The relative target list rlist is then ordered in the following two ways, which can be selected according to the specific situation. An example of two sort strategies is shown in fig. 5.
1) The idea of the first ordering method is to choose the relative in a global scope i The larger value participant replaces its reference participant. For a relative target list rlist, the first key relative in a binary group i And performing descending sorting.
2) The second way of ordering is to prioritize longer directed paths in the AOE network.
The specific method comprises the following steps: for AOE network d= (V,e) Dividing route by edge set 1 ＝{e 1 ,e 2 ,…e k },route 2 ＝{e k+1 ,e k+2 ,…e u }…route x ＝{e s ,e s+1 ,…e v Sequentially pass through route i The directed edge in (a) can obtain a directed path in the AOE network, and |route 1 |≥|route 2 |≥…|route x |，route 1 ∪route 2 ∪…∪route x =e. In addition, there is a function edgeind (route i ) =i, which represents the priority of the corresponding subset of edges. Assuming that the partitioning of the edge set is unique and deterministic (if there are directed paths of the same length, then the precedence order is manually specified). An example of this division is shown in fig. 4, where the same dotted-line type of directed edges belong to the same subset of divisions. The relative target list is then extended to exrlist= { (relative) i ,timefromagent i ,edgeind(round v ))|1≤i≤n,agent i Not the reference participant },
wherein route v Is a participant agent i SubTask of the SubTask m The subset of partitions where the corresponding directed edge e is located. Next, the exrlist is replaced by the first key word relative i And performing descending sorting. And then carrying out incremental stable sorting on the exrlist by the third keyword degeind. And finally deleting the third keyword edgeind in the ordered extension relative target list exrlist, and taking the third keyword edgeind as a new relative target list rlist.
S25: setting all subtasks to the unaccessed state.
S26: if the current rlist is null, determining an allocation result according to the winner information in the allocation scheme S.
S27: select the corresponding participant agent of the first relative target in the list rlist k . While the relative target is deleted from rlist.
S28: if the participant agent k SubTask where the SubTask is located u In the accessed state, the process returns to step S26 to continue execution.
S29: constructing a new allocation scheme S' =s, wherein SubTask u Winner of (2)Changing to participant agent k . It is judged whether the new allocation scheme S' can satisfy the time constraint with the same longest path algorithm as in step S23. If the calculation task of S 'can be completed in the finish time, let the current allocation scheme s=s', and simultaneously set SubTask u Is an accessed state. Otherwise, the process returns to step S26 to continue execution.
S3: the corresponding price rule is obtained by applying the Mierson theory on the basis of the distribution rule, and the specific flow is as follows:
s31: step S2 is executed to obtain allocation vector allocation. For each participant agent i If it is the winner, i.e. allocation i =1, step S32 is performed to calculate the price it needs to pay. Otherwise, it pays the price of the price party i =0, and step S33 is continued.
S32: assume that the original relative target sequence monotonically decreasing in step S24 is b=b 1 b 2 b 3 …b k (b i Actually the binary group described in step S24), to be associated therewith i All relative targets b belonging to the same competition list a ,b b ,b c ,…,
b u … is deleted to give a new original target sequence b'. Step S2 is performed again with sequence b' to obtain the DAG sequence gs=g corresponding to each winner when generated 1 ,G 2 ,G 3 …G u Each G i Is the DAG (mainly winner information on each side) corresponding to a new allocation scheme S' in step S29.
1) If agent i Not the reference participant. Assume agent i Is the relative standard of (2)
The relative positions in the DAG sequence GS are: g 1 ,G 2 ,…,/>
G j ,G j+1 ,…G u . Then traversing j.ltoreq.v.ltoreq.u in turn, in Directed Acyclic Graph (DAG) G v On the basis of (a), agent i The winner of the sub-task is set to obtain a new allocation scheme S ', and then the longest path algorithm same as that of the step S23 is used for judging whether the S' meets the time limit condition finishedtime. If not, the participant agent i The price to be paid is: />
Wherein bid i Is an agent i Is a true target of (a).
2) If agent i Is a reference participant. Hypothesis and agent i The remaining relative target of the same subtask is of order G in the DAG sequence GS 1 ,G 2 ,...,z 1 ,...,z 2 ,...,z k ,...G u . Continuously reducing bid i Until z 1 ,...z k The first time a new winner occurs, the decrease is Δb. If Δb > bid i Then the parent is i =0. Otherwise, the parent i ＝bid i -Δb。
S33: determining the price to be paid by each participant, i.e. the participant agent, based on the vector of parameters i Price of (1) is a parent i 。
While the invention has been described in terms of specific embodiments, any feature disclosed in this specification may be replaced by alternative features serving the equivalent or similar purpose, unless expressly stated otherwise; all of the features disclosed, or all of the steps in a method or process, except for mutually exclusive features and/or steps, may be combined in any manner.
Claims (3)
1. A time-sensitive multiparty data fusion incentive method comprising the steps of:
step 1: initializing:
modeling the multiparty data fusion Task by adopting an AOE network, and determining the internal topological relation of each sub-calculation Task;
each directed edge < u, v > on the AOE network represents a sub-calculation task, u and v represent two synchronous waiting nodes respectively, each synchronous waiting node in the AOE network can start all sub-tasks of outgoing edges after waiting for all sub-tasks of incoming edges to finish, a source point and a sink point in the constructed topological graph represent an initial state and an end state respectively, and the outgoing degree or the incoming degree of an intermediate node is not 0;
step 2: determining an allocation rule:
for each sub-calculation task, deleting the participants with low real targets and high calculation time length, so that the real targets of the rest participants do not decrease monotonically along with the calculation time length;
initializing a reference solution, the winner of each subtask being its reference participant;
constructing a relative target, wherein the value of the relative target represents the benefit brought by replacing the reference participant; and ordering the relative target sequences;
there are two calculation methods for the relative target value:
mode one: total gain increment;
mode two: increment of income brought by increment of unit duration;
step 3: determining price rules based on the michelson theorem:
setting the allocation of participants to a vector of 0-1, i.e., the allocation of winners to 1, and the allocation of non-winners to 0;
calculating a critical target of winning allocated from 1 to 0 based on the Mierson's lements, setting the critical target as a price to be paid by winning and setting a payment of non-winning to 0;
wherein, calculate the critical goal of the winner from 1 to 0 based on the Mierson's lemma, regard this critical goal as the price that winner need pay specifically:
1) If the winner is a non-reference participant,
deleting the relative targets of all participants of the subtask to which the winner belongs in the original relative target sequence; maintaining the other relative targets unchanged, and then operating a greedy flow in the allocation rule to obtain a series of new critical greedy results;
based on the obtained new critical greedy result, the current winner is sequentially replaced by the corresponding reference participant, the relative target which cannot be continuously winned by the current winner is found, and then the relative target is converted into a real target b ans The current price that the winner needs to pay is: max (b) ans ,b next ) Wherein b next Is the true target of the participant belonging to the same subtask as the current winner, and the relative target value is less than or equal to the maximum value of the winner;
2) If the winner is the reference participant,
since decreasing the current winner's target means increasing the relative targets of other participants of the same subtask at the same time, it is only necessary to continuously decrease the current winner's target, run the greedy algorithm in the allocation rule flow until the current subtask's reference participant is replaced, at which point the current winner's target is the price it pays.
2. The method of claim 1, wherein in step 2, the ordering of the relative target sequences uses global alignment or long chain prioritization;
wherein, the global contrast and the long chain priority are respectively;
global contrast:
directly sorting the relative targets of all non-reference participants according to the relative target values, wherein the higher the relative target value is, the higher the participant priority is;
long chain priority:
performing edge set division on the AOE, wherein each division subset represents a directed path;
the method comprises the steps that divided subsets with more edges are preferentially considered, relative targets of participants belonging to the same divided subsets are located in the same section, the section ordering positions corresponding to the divided subsets with more edges are more front, and then the interior of the same section is arranged from large to small according to relative target values;
executing greedy flow according to the ordered relative target sequence: sequentially replacing a reference participant with corresponding participants of each relative object from front to back, and running the longest path algorithm of the directed acyclic graph DAG to judge whether the time limit is met, if so, updating the solution, wherein the current participant becomes the winner of the subtask to which the current participant belongs; if not, continuing subsequent greedy; wherein winners of each subtask are updated at most once.
3. The method of any of claims 1-2, wherein the multiparty data is multiparty track data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011243280.0A CN112416579B (en) | 2020-11-10 | 2020-11-10 | Time-sensitive multiparty data fusion excitation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011243280.0A CN112416579B (en) | 2020-11-10 | 2020-11-10 | Time-sensitive multiparty data fusion excitation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112416579A CN112416579A (en) | 2021-02-26 |
| CN112416579B true CN112416579B (en) | 2023-05-19 |
Family
ID=74781004
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011243280.0A Active CN112416579B (en) | 2020-11-10 | 2020-11-10 | Time-sensitive multiparty data fusion excitation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112416579B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113886686A (en) * | 2021-09-24 | 2022-01-04 | 北京沃东天骏信息技术有限公司 | Method, apparatus, storage medium, and program product for recommending item information to user |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101107589A (en) * | 2005-01-25 | 2008-01-16 | 赛瑞斯特公司 | Attention economy for attention to messages, tasks and resources |
| CN102004641B (en) * | 2010-11-23 | 2013-10-02 | 武汉大学 | Topology potential-based software requirement case priority sorting method |
| CN105740339B (en) * | 2016-01-25 | 2019-03-01 | 河北中科恒运软件科技股份有限公司 | Civil administration big data merges management system |
| CN106155791B (en) * | 2016-06-30 | 2019-05-07 | 电子科技大学 | A kind of workflow task dispatching method under distributed environment |
| WO2018209323A1 (en) * | 2017-05-12 | 2018-11-15 | Intel Corporation | Dynamic access policy provisioning in a device fog |
| CN109582847B (en) * | 2018-11-23 | 2021-08-24 | 咪咕视讯科技有限公司 | Information processing method and device and storage medium |
| CN110378663A (en) * | 2019-04-19 | 2019-10-25 | 西北工业大学 | A kind of complicated movement crowdsourcing method for allocating tasks based on Greedy strategy |
-
2020
- 2020-11-10 CN CN202011243280.0A patent/CN112416579B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112416579A (en) | 2021-02-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109657930B (en) | Resource alliance construction method of cooperative task based on virtual constellation | |
| Brânzei et al. | Two approaches to the problem of sharing delay costs in joint projects | |
| CN111291973A (en) | Space crowdsourcing task allocation method based on alliance | |
| CN108415760B (en) | Crowd sourcing calculation online task allocation method based on mobile opportunity network | |
| CN112416579B (en) | Time-sensitive multiparty data fusion excitation method | |
| Du et al. | Adversarial deep learning for online resource allocation | |
| Baek et al. | Small profits and quick returns: An incentive mechanism design for crowdsourcing under continuous platform competition | |
| Yue et al. | A double auction-based approach for multi-user resource allocation in mobile edge computing | |
| Li et al. | Efficient adaptive matching for real-time city express delivery | |
| Tang et al. | Collaborative cloud-edge-end task offloading with task dependency based on deep reinforcement learning | |
| Sun et al. | An auction and league championship algorithm based resource allocation mechanism for distributed cloud | |
| Cui et al. | TCAM: A truthful combinatorial auction mechanism for crowdsourcing systems | |
| Travadi et al. | Welfare and fairness dynamics in federated learning: A client selection perspective | |
| Miller et al. | Dynamic trucking equilibrium through a freight exchange | |
| Maghsudi et al. | Distributed task management in cyber-physical systems: How to cooperate under uncertainty? | |
| Chen et al. | Delay-aware incentive mechanism for crowdsourcing with vehicles in smart cities | |
| Varakantham et al. | Mechanism design for strategic project scheduling | |
| Dall’Aglio et al. | Sometimes the computation of the shapley value is simple | |
| Sun et al. | An intelligent resource allocation mechanism in the cloud computing environment | |
| Ze'evi et al. | Socially motivated partial cooperation in multi-agent local search | |
| Zhu et al. | A privacy-preserving incentive mechanism in satellite-terrestrial crowdsensing | |
| Küçüksayacıgil et al. | A hybrid genetic algorithm application for a bi-objective, multi-project, multi-mode, resource-constrained project scheduling problem | |
| Wang et al. | Reinforcement Contract Design for Vehicular-Edge Computing Scheduling and Energy Trading Via Deep Q-Network With Hybrid Action Space | |
| Gao et al. | You Can Trade Your Experience in Distributed Multi-Agent Multi-Armed Bandits | |
| Wan et al. | Research on dynamic task allocation algorithm to improve user participation in the Witkey mode |
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 |