CN114553869A - Method and device for determining resource contribution degree based on joint learning and electronic equipment - Google Patents
Method and device for determining resource contribution degree based on joint learning and electronic equipment Download PDFInfo
- Publication number
- CN114553869A CN114553869A CN202210129960.2A CN202210129960A CN114553869A CN 114553869 A CN114553869 A CN 114553869A CN 202210129960 A CN202210129960 A CN 202210129960A CN 114553869 A CN114553869 A CN 114553869A
- Authority
- CN
- China
- Prior art keywords
- value
- participant
- resource amount
- model
- determining
- 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
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000004891 communication Methods 0.000 claims abstract description 127
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 29
- 230000006870 function Effects 0.000 claims description 62
- 238000004364 calculation method Methods 0.000 claims description 41
- 238000005304 joining Methods 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 13
- 238000012614 Monte-Carlo sampling Methods 0.000 claims description 10
- OYFJQPXVCSSHAI-QFPUQLAESA-N enalapril maleate Chemical compound OC(=O)\C=C/C(O)=O.C([C@@H](C(=O)OCC)N[C@@H](C)C(=O)N1[C@@H](CCC1)C(O)=O)CC1=CC=CC=C1 OYFJQPXVCSSHAI-QFPUQLAESA-N 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000009826 distribution Methods 0.000 abstract description 8
- 230000008569 process Effects 0.000 description 8
- 238000003860 storage Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000007246 mechanism Effects 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 4
- 230000007423 decrease Effects 0.000 description 4
- 238000005259 measurement Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000007792 addition Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 238000013468 resource allocation Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000004931 aggregating effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003121 nonmonotonic effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000010355 oscillation Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000000638 stimulation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- 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
-
- 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
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
-
- 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
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Algebra (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Probability & Statistics with Applications (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Telephonic Communication Services (AREA)
Abstract
The disclosure relates to the technical field of joint learning, and provides a method and a device for determining resource contribution degree based on joint learning and electronic equipment. The method comprises the following steps: obtaining models uploaded by all participants in a joint learning framework; the uploaded model carries the communication turn of the model and the marginal contribution value of the participant corresponding to the communication turn; determining the resource amount of each participant based on the marginal contribution value; determining the allocation resource amount of each participant according to the resource amount of the participant and a Charapril value algorithm; and obtaining the contribution degree of each participant to the combined model based on the communication round and the distributed resource amount. The method and the system can realize incentive distribution of the joint learning through the contribution degree of the participators to the joint model so as to ensure the success and sustainability of the joint learning.
Description
Technical Field
The present disclosure relates to the field of joint learning technologies, and in particular, to a method and an apparatus for determining a resource contribution degree based on joint learning, and an electronic device.
Background
The joint learning organizes the co-building intelligence of a plurality of participants, and a reasonable distribution incentive mechanism needs to be designed to ensure the success and sustainability of cooperation. According to the IEEE P3652.1 joint learning architecture and application specification, an incentive mechanism generally needs to meet a plurality of constraint conditions such as individual reasonability, collective reasonability, fairness and the like, wherein fair distribution benefits are particularly important, so that distribution according to contribution size is a reasonable distribution method. Therefore, how to determine the contribution of each participant to the joint learning is a technical problem in the joint learning application.
Disclosure of Invention
In view of this, the embodiments of the present disclosure provide a method, an apparatus, and an electronic device for determining a resource contribution degree based on joint learning, so as to solve the problem in the prior art how to determine contributions of various participants to joint learning.
In a first aspect of the embodiments of the present disclosure, a method for determining a resource contribution degree based on joint learning is provided, including: obtaining models uploaded by all participants in a joint learning framework; the uploaded model carries the communication turn of the model and the marginal contribution value of the participant corresponding to the communication turn; determining the resource amount of each participant based on the marginal contribution value; determining the resource allocation amount of each participant according to the resource amount of the participant and a summer pril value algorithm; and obtaining the contribution degree of each participant to the combined model based on the communication round and the distributed resource amount.
In a second aspect of the embodiments of the present disclosure, an apparatus for determining a resource contribution degree based on joint learning is provided, including: the acquisition module is configured to acquire models uploaded by all participants in the joint learning architecture; the uploaded model carries the communication turn of the model and the marginal contribution value of the participant corresponding to the communication turn; a first determining module configured to determine an amount of resources of each participant based on the marginal contribution value; the computing module is configured to determine the distributed resource amount of each participant according to the resource amount of the participant and a Charapril value algorithm; and the second determining module is configured to obtain the contribution degree of each participant to the combined model based on the communication round and the distributed resource amount.
In a third aspect of the embodiments of the present disclosure, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program.
Compared with the prior art, the embodiment of the disclosure has the following beneficial effects: obtaining models uploaded by all participants in a joint learning framework; the uploaded model carries the communication turn of the model and the marginal contribution value of the participant corresponding to the communication turn; determining the resource amount of each participant based on the marginal contribution value; determining the allocation resource amount of each participant according to the resource amount of the participant and a Charapril value algorithm; based on communication rounds and resource allocation, the contribution degree of each participant to the joint model is obtained, so that the incentive allocation of the joint learning can be realized according to the contribution degree of the participants to the joint model, and the success and the sustainability of the joint learning are ensured.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings used in the embodiments or the prior art descriptions will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art without inventive efforts.
FIG. 1 is a scenario diagram of an application scenario of an embodiment of the present disclosure;
FIG. 2 is a flowchart illustrating a method for determining resource contribution based on joint learning according to an embodiment of the present disclosure;
fig. 3 is a schematic flow chart of a method of step S203 in fig. 2 according to an embodiment of the present disclosure;
fig. 4 is a schematic flowchart of another method of step S203 in fig. 2 according to an embodiment of the present disclosure;
fig. 5 is a schematic structural diagram of an apparatus for determining resource contribution degree based on joint learning according to an embodiment of the present disclosure;
fig. 6 is a schematic structural diagram of an electronic device provided in an embodiment of the present disclosure.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the disclosed embodiments. However, it will be apparent to one skilled in the art that the present disclosure may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present disclosure with unnecessary detail.
Joint learning refers to comprehensively utilizing multiple AI (Artificial Intelligence) technologies on the premise of ensuring data security and user privacy, jointly mining data values by combining multiple parties, and promoting new intelligent business states and modes based on joint modeling. The joint learning has at least the following characteristics:
(1) and the participating nodes control the weak centralized joint training mode of own data, so that the data privacy security in the co-creation intelligent process is ensured.
(2) Under different application scenes, a plurality of model aggregation optimization strategies are established by utilizing screening and/or combined AI algorithm and privacy protection calculation so as to obtain a high-level and high-quality model.
(3) On the premise of ensuring data security and user privacy, the method for improving the efficiency of the joint learning engine is obtained based on a plurality of model aggregation optimization strategies, wherein the efficiency method can improve the overall efficiency of the joint learning engine by solving the problems of information interaction, intelligent perception, abnormal processing mechanisms and the like under the conditions of parallel computing architectures and large-scale cross-domain networks.
(4) The requirements of the users of multiple parties in each scene are acquired, the real contribution degree of each joint participant is determined and reasonably evaluated through a mutual trust mechanism, and distribution stimulation is carried out.
Based on the mode, the AI technical ecology based on the joint learning can be established, the industrial data value is fully exerted, and the falling of scenes in the vertical field is promoted.
A method and an apparatus for determining resource contribution based on joint learning according to an embodiment of the present disclosure will be described in detail below with reference to the accompanying drawings.
Fig. 1 is an architecture diagram of joint learning according to an embodiment of the present disclosure. As shown in fig. 1, the architecture of joint learning may include a server (central node) 101, as well as a participant 102, a participant 103, and a participant 104.
In the joint learning process, a basic model may be built by the server 101, and the server 101 sends the model to the participants 102, 103, and 104 with which communication connections are established. The basic model may also be uploaded to the server 101 after any participant has established the model, and the server 101 sends the model to other participants with whom communication connection is established. The participating party 102, the participating party 103 and the participating party 104 construct models according to the downloaded basic structures and model parameters, perform model training by using local data to obtain updated model parameters, and upload the updated model parameters to the server 101 in an encrypted manner. Server 101 aggregates the model parameters sent by participants 102, 103, and 104 to obtain global model parameters, and passes the global model parameters back to participants 102, 103, and 104. And the participants 102, 103 and 104 iterate the respective models according to the received global model parameters until the models finally converge, thereby realizing the training of the models. In the joint learning process, data uploaded by the participants 102, 103 and 104 are model parameters, local data are not uploaded to the server 101, and all the participants can share the final model parameters, so that common modeling can be realized on the basis of ensuring data privacy. It should be noted that the number of the participants is not limited to three, but may be set according to needs, and the embodiment of the disclosure does not limit this.
Fig. 2 is a flowchart illustrating a method for determining a resource contribution degree based on joint learning according to an embodiment of the present disclosure. The method of determining resource contribution based on joint learning of fig. 2 may be performed by the server of fig. 1. As shown in fig. 2, the method for determining resource contribution degree based on joint learning includes:
s201, obtaining models uploaded by all participants in a joint learning framework; the uploaded model carries the communication turn of the model and the marginal contribution value of the participant corresponding to the communication turn;
s202, determining the resource amount of each participant based on the marginal contribution value;
s203, determining the distributed resource amount of each participant according to the resource amount of the participant and a Charapril value algorithm;
and S204, obtaining the contribution degree of each participant to the combined model based on the communication round and the distributed resource amount.
In particular, in conjunction with fig. 1, the data quantity held by each participant establishing a communication connection with the server 101 may not be the same, and thus it can be seen that each participant performs model training by using the held data, and the contribution corresponding to the resulting global model will be different. The resource amount of the participator can be the data amount used for the joint model training, and for the same participator, the resource amount used for the joint learning in each communication turn can be the same or different.
Specifically, the Shapley value is a solution to the cooperative game, suggested by the Nobel economics awarding Suapril. In the embodiment of the disclosure, joint learning can be abstracted to a cooperative game with participation of multiple parties, and if a utility function is defined as the accuracy or negative loss of a joint model, the smaller the loss is, the higher the model accuracy is, so that a charpy value applied in the joint learning can be used for contribution measurement and distribution. For example, the allocation of interest is fairly made by averaging the marginal contributions from joining federation S by participants i (equivalent to individuals) of joint learning, in detail, the value of xiapril takes into account all possible orders in which participants i join federation, N is the federation composed of full combinations, the number of participants is | N | ═ N, S is the sub-federation in case of a certain permutation of combinations, v () is the utility function, and | symbol represents the number of elements in the solution. [ v (S { U { i }) -v (S) ] represents the marginal utility of participant i after joining subunion S, | S |! (| N | - | S | -1) |! I N |! Indicates the probability (i.e., weight) of occurrence of the combination, where i, N, and S are positive integers, respectively.
According to the technical scheme provided by the embodiment of the disclosure, the amount of the resource distributed by the participants in each communication turn is measured by applying the value of the Charapril in the joint learning, so that the contribution degree of each participant to the joint model is accurately obtained, the joint learning is excited and distributed fairly, and the success and the sustainability of the joint learning are ensured.
In some embodiments, the amount of resources for each participant includes the amount of data used for joint model training.
Specifically, in each communication turn of joint learning, the number of participants participating in joint learning and the amount of resources (e.g., the amount of data used for model training) for model training of each participant in the communication turn are not constant. For example, in conjunction with fig. 1, only party 102 and party 103 may participate in joint learning in one communication round, while in the next communication round, party 102, party 103 and party 104 may participate in joint learning together, and the amount of data that party 102 and party 103 use to participate in joint model training in the two communication rounds may be the same or different. Therefore, the resource amount corresponding to each participant can be determined by the marginal contribution value carried in the model uploaded to the server 101 by the participant.
Specifically, the marginal contribution value of the participant in each communication turn may be calculated and measured according to the income generated by the participant joining the joint learning, or may be calculated and measured according to the loss generated by the participant leaving the joint learning. The method for performing the calculation and measurement according to the profit generated by the participant joining the joint learning may be to use a difference between the profit generated by the participant i joining the joint learning organization and the profit generated by the participant i joining the joint learning organization as a marginal contribution value of the participant, or may also use a charapril value of the data quantity of the training model of the participant as the marginal contribution value, which is not limited in the embodiment of the present disclosure.
In the practical application of joint learning, if the classical charpy value calculation method is directly applied, a large amount of calculation overhead is generated: firstly, a joint learning model needs to be trained aiming at a large number of different participators in a combined way; the second is the need to deduce the assessment accuracy or loss for these combined models. These computational loads will grow exponentially as the number of participants increases, requiring a significant amount of computational time to be spent. Therefore, the following embodiments will further optimize the method for determining the resource contribution degree, and save the calculation time while ensuring the calculation accuracy.
In some embodiments, the allocated resource amount comprises a sharpril value of the resource amount corresponding to the participant at each communication turn; determining the distributed resource amount of each participant according to the resource amount of the participant and a Charapril value algorithm, wherein the method comprises the following steps: determining all combination sequences of the participants for joining the alliances based on the number of the participants, wherein each combination sequence for joining the alliance is a sub-alliance; sampling all combination sequences based on a Monte Carlo sampling method, and determining a target sub-union for carrying out the calculation of the Sharply value; in each communication turn, calculating the marginal contribution of each participant in the combination sequence corresponding to the target sub-alliance based on a Shaapril value algorithm, wherein the marginal contribution is equal to a utility function value of the resource quantity of the participant in the current communication turn; under the condition that the absolute difference value between the marginal contribution of the global model of the current communication round and the global model is smaller than a preset first threshold value, the marginal contribution calculation of each participant in the combination sequence corresponding to the target sub-alliance of the current communication round is exited, and the xiapril value of each participant of the current communication round is the mean value of the marginal contribution calculated based on the Monte Carlo sampling method; and under the condition that the absolute difference value of the marginal contribution of the global model of the current communication round and the marginal contribution of the global model of the previous communication round is smaller than or equal to a preset second threshold value, finishing calculating the Sharple value of each participant, wherein the Sharple value of each participant of the current communication round is zero.
Specifically, referring to fig. 1, in each communication turn, the participant 102, the participant 103, and the participant 104 may join the federation of the joint learning in different orders to participate in the joint learning, and therefore, the charpy value of the resource amount corresponding to each participant may be calculated under the combined order of joining the federation of the participants in each communication turn. Obviously, if the number of participants is larger, the order of combination of these participants joining the federation in each communication turn will be larger, meaning a large number of computations will be performed. Therefore, in the embodiment of the present disclosure, a monte carlo sampling method is adopted to sample the combination sequence of the participants in each communication turn joining the federation, so that only the corresponding utility is calculated for the resource amount corresponding to each participant where the target combination sequence obtained by sampling is located, and thus the calculation time can be saved under the condition of ensuring the calculation accuracy.
Specifically, for the calculation of the value of sharp in the communication round, in the embodiment of the present disclosure, the absolute value of the difference between the utility of the global model in the current communication round and the global model is compared with a preset first threshold, and when the absolute value of the difference between the utility of the global model and the global model is smaller than the first threshold, the calculation of the value of sharp in the other combination sequences in the current communication round is cut off, so that the calculation time of the value of sharp in a single communication round can be effectively saved. The global model is a model structure or model parameters obtained by aggregating model parameters uploaded by each participant by the server under the current communication turn.
Specifically, for the calculation of the value of the sharp between the communication turns, in the embodiment of the present disclosure, an absolute value is obtained after a difference between the utility of the global model in the current communication turn and the utility of the global model corresponding to the previous communication turn is calculated, and is compared with a preset second threshold, and if the absolute value is smaller than or equal to the second threshold after the difference between the utility of the global model in the current communication turn and the utility of the global model corresponding to the previous communication turn is calculated, the calculation of the value of the sharp is ended, so that the calculation amount of the value of the sharp is reduced, and the calculation time is saved. Wherein the second threshold may be a convergence condition of the joint model.
For some of the foregoing embodiments, the second threshold is generally set empirically, so that in order to balance the convergence rate and the calculation accuracy, multiple trial and error are required, and in a practical application scenario, a satisfactory result is usually expected to be obtained by one calculation.
Further, in some embodiments of the foregoing, the method may further include: determining a marginal gain of the final combined model based on a summer pril value algorithm, wherein the marginal gain is equal to an absolute difference value of a utility function value of the final combined model and a utility function value of the initial model; based on the marginal gain, a second threshold is determined.
Specifically, let the marginal gain of the final joint model utility function relative to the initial model be ΔU＝|V(M(T))-V(M(0)) L where T is the total communication round, V (M)(T)) Represents the final joint model utility function, V (M)(0)) Representing the utility function of the initial model, a convergence parameter (i.e., a second threshold) λ ═ Δ may be setUThreshold, for example, threshold ═ 0.01. The least-calculated number of combined models parameter is set to int (2)nCoe), wherein coe is 0.8, thereby effectively reducing the calculation time.
According to the technical scheme provided by the embodiment of the disclosure, the pruning convergence parameter is adaptively adjusted through the final combined model utility function relative to the marginal gain of the initial model, so that manual intervention is reduced, and the condition that the model gain range exceeds 1.0 is suitable.
Further, in some embodiments of the foregoing, the method may further include: and in at least two continuous communication rounds, if the average value of the marginal gain is smaller than a second threshold value, finishing calculating the sharp value of each participant.
Specifically, in some embodiments described above, the assumption that the marginal gain of the utility function of the joint model decreases monotonically is implicit, but in an actual scenario, a non-monotonic decrease may occur. For example, when the utility function oscillation converges, the marginal gain of the (t) th communication round approaches zero, and a gain greater than the convergence parameter λ occurs in the (t +1) th communication round. If the method in some embodiments described above exits the iterative loop after reaching the convergence condition in the (t) th round, the information in the (t +1) th round is lost, so that the settlement result deviates more from the true value. For this reason, the present embodiment sets the convergence determination condition such that the average value of the margin gains of consecutive m rounds (m > -2) is smaller than the threshold λ, and determines that the loop is converged and exits the loop. According to the technical scheme provided by the embodiment of the disclosure, the performance of the computation method of the value of the Charapril on the actual scene real data set is more stable by enhancing the robustness of the truncation convergence condition among turns and capturing the condition of the decreasing fluctuation of the utility function.
In some embodiments, the determining the allocated resource amount of each participant according to the resource amount of the participant and a charapril value algorithm includes:
s301, determining all combination sequences of the participants for joining the alliance based on the number of the participants;
s302, calculating the weight of each combination sequence;
s303, in each communication turn, calculating an absolute difference value of a utility function value of the global model of the current communication turn and a utility function value of the global model of the previous communication turn;
s304, under the condition that the absolute difference value is larger than a preset second threshold value, acquiring a target combination sequence which is smaller than or equal to a preset weight threshold value in all combination sequences of the participants joining the alliance, and calculating the marginal contribution of each participant in the target combination sequence;
s305, under the condition that the absolute difference value is smaller than or equal to a preset second threshold value, quitting calculating the marginal contribution of the participant in the current communication turn;
s306, determining the charapril value of the resource amount corresponding to each communication turn of the participant based on the marginal contribution of the participant and a charapril value algorithm.
Specifically, monte carlo sampling is a statistical method, a large number of samples are required, the weight representing the occurrence probability of the sub-combination model in the xiapril value definition formula is naturally reflected in multiple samples, and the higher the probability, the more times the combination is sampled. However, the foregoing charpy value calculation method using monte carlo sampling in some embodiments does not consider the probability of occurrence of the sub-combination model (i.e., the sub-union in a certain permutation case), and this problem is increasingly highlighted when the actual number of samples is small. In practical application, it will happen that the average marginal gain calculated by sampling will deviate from the true value of the xiapril, thereby affecting the accuracy of calculating the value of the xiapril.
Specifically, defined in terms of the value of xiapril, the weight w ═ S |! (| N | - | S | -1) | N |! As S increases to N, the S increases from a large decrease to a larger increase in the S-N shape. And the contribution of the prunel value of the participant i made by the marginal gain of i in the case of this combination S is the weight w (v (S ueq { i }) -v (S)). If w is small or (v (S { i }) -v (S)) is small, the partial value will have a small effect on the charpy value of i and can be ignored to reduce the amount of computation. Therefore, in the embodiment of the present disclosure, in the calculation of the value of happril for each communication turn, a weight threshold value w _ threshold is set, only the gain of the utility function of the sub-combination model of w ═ w _ threshold is calculated, and the value of happril is substituted into the definition of happril to calculate the value of happril.
According to the technical scheme provided by the embodiment of the disclosure, the uncertainty of the iteration times of the loop of Monte Carlo sampling is replaced by the calculation of the determined times, and the preliminary measurement and calculation shows that when the calculated amount is equivalent, the average error of the computed value of the Charpy is lower; if the same precision is maintained, the calculation amount of the embodiment of the disclosure is less, so that the calculation time is saved under the condition of ensuring the same precision.
In some embodiments, the determining the allocated resource amount of each participant according to the resource amount of the participant and a charapril value algorithm includes:
s401, determining the full combination sequence of all participants in the alliance and the head combination sequence of part of the participants in the alliance based on the number of the participants;
s402, calculating the weight of each combination sequence;
s403, in each communication turn, calculating an absolute difference value between the utility function value of the current communication turn global model and the utility function value of the previous communication turn global model;
s404, under the condition that the absolute difference value is larger than a preset second threshold value, respectively calculating utility function values of the head combination sequence and utility function values of all the combination sequences, performing linear or nonlinear function interpolation by using the utility function values of the head combination sequence and the utility function values of all the combination sequences, and estimating utility function values of the rest part of participants added in the rest combination sequences of the alliance;
s405, determining a Sharpril value of the resource amount corresponding to each communication turn of the participant based on the utility function value and the Sharpril algorithm.
Specifically, in some embodiments described above, the sharp approximation algorithm that relies on a first threshold for intra-round truncation of communications sets the first threshold to ε if | vN-vj-1|＝|V(M(t+1))-Agg(M(t),{ΔC})|<E, the inner loop is exited, i.e. the marginal contribution of the participants in the sub-combination situation after j, where j is a positive integer less than N, Agg (M)(t),{ΔC}) represents the global model obtained at communication round (t +1), i.e. M(t+1)＝Agg(M(t),{ΔC}). Therefore, if the marginal contribution in these cases (for example, far below the deep neural network model derivation time) can be estimated with low computational complexity and involved in the computation of the xiapril value, it is advantageous to further reduce the computation error of the xiapril value.
Specifically, in a typical joint model training process, as the number of participants in the sub-federation S increases, the marginal contribution of the newly added participant i decreases within one communication turn. First, a utility function v (S) of a head sub-combination S composed of a small number of participants is calculated, | S | ═ 1,2 … K, K < | N |, K being an integer. Calculating a utility function v (N) of the total combination, applying a linear (e.g. y ═ ax + b) or non-linear (e.g. softsign, tanh, etc.) interpolation function, making an estimate of the remaining v (S) | S | ═ k +1, k +2 … | N | -1, and using the estimate to participate in the calculation of the xiapril value. For example, in the embodiment shown in fig. 1, a total of 3 participants 102, 103 and 104, i.e., N is 3 and k is 1, when v (1) is calculated, v (1), v (1,2,3) (i.e., a full combinational model) can be interpolated by a linear or non-linear interpolation function to estimate v (1,2), and the estimated value is used to participate in the calculation of the xiapril value.
According to the technical scheme provided by the embodiment of the disclosure, the marginal contribution value smaller than the cut-off threshold value (namely the first threshold value) in the communication round is estimated by a low calculation overhead method and participates in the calculation of the sharp value, and the preliminary measurement and calculation of practical application show that the calculation error of the sharp value can be reduced on the premise of not obviously improving the calculation amount.
In some embodiments, obtaining the contribution of each participant to the federated model based on the communication round and the amount of allocated resources includes: and calculating the sum of the summer pril values of the resource quantities corresponding to the communication rounds of the participants to obtain the contribution of the participants to the combined model.
Specifically, in conjunction with some of the above embodiments, the sharpril value of the resource amount corresponding to each participant in each communication turn may be calculated, and then the sharpril values of the participants in each communication turn may be summed, so as to obtain the contribution of the participant to the joined model. In practical applications, the incentive mechanism for the participants can be determined according to the contribution degree, so that the joint learning can be successfully carried out and the sustainability is maintained.
All the above optional technical solutions may be combined arbitrarily to form optional embodiments of the present application, and are not described herein again.
The following are embodiments of the disclosed apparatus that may be used to perform embodiments of the disclosed methods. For details not disclosed in the embodiments of the apparatus of the present disclosure, refer to the embodiments of the method of the present disclosure.
Fig. 5 is a schematic diagram of an apparatus for determining a resource contribution degree based on joint learning according to an embodiment of the present disclosure. As shown in fig. 5, the apparatus for determining resource contribution degree based on joint learning includes:
an obtaining module 501 configured to obtain models uploaded by each participant in the joint learning architecture; the uploaded model carries the communication turn of the model and the marginal contribution value of the participant corresponding to the communication turn;
a first determining module 502 configured to determine an amount of resources of each participant based on the marginal contribution value;
a calculating module 503 configured to determine the allocated resource amount of each participant according to the resource amount of the participant and a charapril value algorithm;
and a second determining module 504 configured to obtain the contribution degree of each participant to the federated model based on the communication round and the allocated resource amount.
In some embodiments, the amount of resources for each participant includes the amount of data used for joint model training.
In some embodiments, the allocated resource amount comprises a sharpril value of the resource amount corresponding to the participant at each communication turn; the calculation module 503 in fig. 5 is configured to determine all combination orders of joining the alliances of the participants based on the number of the participants, wherein each combination order of joining the alliance is a sub-alliance; sampling all combination sequences based on a Monte Carlo sampling method, and determining a target sub-union for carrying out the calculation of the Sharply value; in each communication turn, calculating the marginal contribution of each participant in the combination sequence corresponding to the target sub-alliance based on a Shaapril value algorithm, wherein the marginal contribution is equal to a utility function value of the resource quantity of the participant in the current communication turn; under the condition that the absolute difference value between the marginal contribution of the global model of the current communication round and the global model is smaller than a preset first threshold value, the marginal contribution calculation of each participant in the combination sequence corresponding to the target sub-alliance of the current communication round is exited, and the xiapril value of each participant of the current communication round is the mean value of the marginal contribution calculated based on the Monte Carlo sampling method; and under the condition that the absolute difference value of the marginal contribution of the global model of the current communication round and the marginal contribution of the global model of the previous communication round is smaller than or equal to a preset second threshold value, finishing calculating the Sharple value of each participant, wherein the Sharple value of each participant of the current communication round is zero.
In some embodiments, the apparatus for determining resource contribution degree based on joint learning in fig. 5 further includes:
a gain calculation module 505 configured to determine a marginal gain of the final combined model based on a charpril value algorithm, wherein the marginal gain is equal to an absolute difference of a utility function value of the final combined model and a utility function value of the initial model;
a threshold adjustment module 506 configured to determine a second threshold based on the marginal gain.
In some embodiments, the apparatus for determining resource contribution degree based on joint learning in fig. 5 further includes:
and a convergence determining module 507 configured to end the calculation of the sharp value of each participant if the average value of the marginal gain is smaller than the second threshold value in at least two consecutive communication rounds.
In some embodiments, the allocated resource amount comprises a sharpril value of the resource amount corresponding to the participant at each communication turn; the calculation module 503 in fig. 5 is configured to determine the overall combination order in which the participants join the federation based on the number of participants; calculating the weight of each combination sequence; in each communication turn, calculating the absolute difference value of the utility function value of the global model of the current communication turn and the utility function value of the global model of the previous communication turn; under the condition that the absolute difference value is larger than a preset second threshold value, acquiring a target combination sequence which is smaller than or equal to a preset weight threshold value in all combination sequences of the participants joining the alliance, and calculating the marginal contribution of each participant in the target combination sequence; under the condition that the absolute difference value is smaller than or equal to a preset second threshold value, quitting calculating the marginal contribution of the participants in the current communication turn; and determining the charapril value of the resource amount corresponding to each communication turn of the participant based on the marginal contribution of the participant and a charapril value algorithm.
In some embodiments, the allocated resource amount comprises a sharpril value of the resource amount corresponding to the participant at each communication turn; the calculation module 503 in fig. 5 is configured to determine a full combination order in which all participants join the federation, and a head combination order in which some participants join the federation, based on the number of participants; calculating the weight of each combination sequence; in each communication turn, calculating the absolute difference value of the utility function value of the global model of the current communication turn and the utility function value of the global model of the previous communication turn; under the condition that the absolute difference value is larger than a preset second threshold value, respectively calculating utility function values of the head combination sequence and utility function values of all the combination sequences, performing linear or nonlinear function interpolation by using the utility function values of the head combination sequence and the utility function values of all the combination sequences, and estimating utility function values of the rest of participants in the remaining combination sequences added in the alliance; and determining the Sharpril value of the resource amount corresponding to each communication turn of the participant based on the utility function value and the Sharpril algorithm.
In some embodiments, the second determining module 504 in fig. 5 is configured to calculate a sum of the charpril values of the resource amounts corresponding to the participants in the respective communication rounds, and obtain the contribution of the participants to the federated model.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure.
Fig. 6 is a schematic diagram of an electronic device 6 provided by an embodiment of the disclosure. The electronic device in fig. 6 may be the server in fig. 1. As shown in fig. 6, the electronic apparatus 6 of this embodiment includes: a processor 601, a memory 602, and a computer program 603 stored in the memory 602 and executable on the processor 601. The steps in the various method embodiments described above are implemented when the computer program 603 is executed by the processor 601. Alternatively, the processor 601 realizes the functions of each module/unit in the above-described apparatus embodiments when executing the computer program 603.
Illustratively, the computer program 603 may be partitioned into one or more modules/units, which are stored in the memory 602 and executed by the processor 601 to accomplish the present disclosure. One or more modules/units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution of the computer program 603 in the electronic device 6.
The electronic device 6 may be a desktop computer, a notebook, a palm computer, a cloud server, or other electronic devices. The electronic device 6 may include, but is not limited to, a processor 601 and a memory 602. Those skilled in the art will appreciate that fig. 6 is merely an example of an electronic device 6, and does not constitute a limitation of the electronic device 6, and may include more or fewer components than shown, or combine certain components, or different components, e.g., the electronic device may also include input-output devices, network access devices, buses, etc.
The Processor 601 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The storage 602 may be an internal storage unit of the electronic device 6, for example, a hard disk or a memory of the electronic device 6. The memory 602 may also be an external storage device of the electronic device 6, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like provided on the electronic device 6. Further, the memory 602 may also include both internal storage units of the electronic device 6 and external storage devices. The memory 602 is used for storing computer programs and other programs and data required by the electronic device. The memory 602 may also be used to temporarily store data that has been output or is to be output.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules, so as to perform all or part of the functions described above. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.
In the embodiments provided in the present disclosure, it should be understood that the disclosed apparatus/electronic device and method may be implemented in other ways. For example, the above-described apparatus/electronic device embodiments are merely illustrative, and for example, a module or a unit may be divided into only one logical function, and may be implemented in other ways, and multiple units or components may be combined or integrated into another system, or some features may be omitted or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present disclosure may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, the present disclosure may implement all or part of the flow of the method in the above embodiments, and may also be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of the above methods and embodiments. The computer program may comprise computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, and the like. It should be noted that the computer readable medium may contain suitable additions or additions that may be required in accordance with legislative and patent practices within the jurisdiction, for example, in some jurisdictions, computer readable media may not include electrical carrier signals or telecommunications signals in accordance with legislative and patent practices.
The above examples are only intended to illustrate the technical solutions of the present disclosure, not to limit them; although the present disclosure has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present disclosure, and are intended to be included within the scope of the present disclosure.
Claims (10)
1. A method for determining resource contribution degree based on joint learning is characterized by comprising the following steps:
obtaining models uploaded by all participants in a joint learning framework; the uploaded model carries a communication turn of the model and a marginal contribution value of a participant corresponding to the communication turn;
determining the resource amount of each participant based on the marginal contribution value;
determining the allocation resource amount of each participant according to the resource amount of the participant and a Sharply value algorithm;
and obtaining the contribution degree of each participant to the combined model based on the communication round and the distributed resource amount.
2. The method of claim 1, wherein the amount of resources of each participant comprises an amount of data used for joint model training.
3. The method of claim 1, wherein the allocated resource amount comprises a Shaapril value of the resource amount corresponding to the participant in each communication turn;
the determining the allocated resource amount of each participant according to the resource amount of the participant and the sharp value algorithm includes:
determining all combination orders of the participants for joining the alliances based on the number of the participants, wherein each combination order of joining the alliance is a sub-alliance;
sampling all the combination sequences based on a Monte Carlo sampling method, and determining a target sub-union for calculating a Sharply value;
in each communication turn, calculating the marginal contribution of each participant to join the target sub-alliance in the combination sequence corresponding to the target sub-alliance based on a Shaapril value algorithm, wherein the marginal contribution is equal to a utility function value of the resource quantity of the participant in the current communication turn;
under the condition that the absolute difference value between the marginal contribution of the global model of the current communication round and the global model is smaller than a preset first threshold value, the marginal contribution of each participant in the combination sequence corresponding to the target sub-alliance of the current communication round is calculated, and the Charpy value of each participant in the current communication round is the mean value of the marginal contributions calculated based on the Monte Carlo sampling method;
and under the condition that the absolute difference value of the marginal contribution of the global model of the current communication round and the marginal contribution of the global model of the previous communication round is smaller than or equal to a preset second threshold value, finishing calculating the Sharpril value of each participant, wherein the Sharpril value of each participant in the current communication round is zero.
4. The method of claim 3, further comprising:
determining a marginal gain of the final combined model based on the charpy value algorithm, wherein the marginal gain is equal to an absolute difference value of a utility function value of the final combined model and a utility function value of the initial model;
determining the second threshold based on the marginal gain.
5. The method of claim 4, further comprising: and in at least two continuous communication rounds, if the average value of the marginal gain is smaller than the second threshold value, finishing calculating the xiapril value of each participant.
6. The method of claim 1, wherein the allocated resource amount comprises a Shaapril value of the resource amount corresponding to the participant in each communication turn;
the determining the allocated resource amount of each participant according to the resource amount of the participant and the sharp value algorithm includes:
determining a total combination order in which the participants join a federation based on the number of participants;
calculating a weight for each of the combination orders;
in each communication turn, calculating the absolute difference value of the utility function value of the global model of the current communication turn and the utility function value of the global model of the previous communication turn;
under the condition that the absolute difference value is larger than a preset second threshold value, acquiring a target combination sequence which is smaller than or equal to a preset weight threshold value in all combination sequences of the participants joining the alliance, and calculating the marginal contribution of each participant in the target combination sequence;
under the condition that the absolute difference value is smaller than or equal to a preset second threshold value, quitting calculating the marginal contribution of the participant in the current communication turn;
and determining the summer pril value of the resource amount corresponding to each communication turn of the participant based on the marginal contribution of the participant and a summer pril value algorithm.
7. The method of claim 1, wherein the allocated resource amount comprises a Shaapril value of the resource amount corresponding to the participant in each communication turn;
the determining the allocated resource amount of each participant according to the resource amount of the participant and the sharp value algorithm includes:
determining a full combination order in which all of the participants join the federation, and a head combination order in which some of the participants join the federation, based on the number of the participants;
calculating a weight for each of the combination orders;
in each communication turn, calculating the absolute difference value of the utility function value of the global model of the current communication turn and the utility function value of the global model of the previous communication turn;
under the condition that the absolute difference value is larger than a preset second threshold value, respectively calculating utility function values of the head combination sequence and utility function values of all the combination sequences, performing linear or nonlinear function interpolation by using the utility function values of the head combination sequence and the utility function values of all the combination sequences, and estimating utility function values of the rest of the residual combination sequences of the participants who join in the alliance;
and determining the Sharpril value of the resource amount corresponding to each communication turn of the participant based on the utility function value and the Sharpril algorithm.
8. The method according to claim 3, 6 or 7, wherein the obtaining the contribution degree of each participant to the federated model based on the communication round and the allocated resource amount comprises:
and calculating the sum of the summer pril values of the resource amount corresponding to each communication turn of the participant to obtain the contribution of the participant to the combined model.
9. An apparatus for determining resource contribution based on joint learning, comprising:
the acquisition module is configured to acquire models uploaded by all participants in the joint learning architecture; the uploaded model carries a communication turn of the model and a marginal contribution value of a participant corresponding to the communication turn;
a first determination module configured to determine an amount of resources of the participants based on the marginal contribution value;
a calculation module configured to determine an allocated resource amount of each of the participants according to the resource amount of the participant and a sharp value algorithm;
and the second determining module is configured to obtain the contribution degree of each participant to the combined model based on the communication round and the distributed resource amount.
10. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1 to 8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210129960.2A CN114553869B (en) | 2022-02-11 | 2022-02-11 | Method and device for determining resource contribution degree based on joint learning and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210129960.2A CN114553869B (en) | 2022-02-11 | 2022-02-11 | Method and device for determining resource contribution degree based on joint learning and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114553869A true CN114553869A (en) | 2022-05-27 |
| CN114553869B CN114553869B (en) | 2024-09-13 |
Family
ID=81672757
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210129960.2A Active CN114553869B (en) | 2022-02-11 | 2022-02-11 | Method and device for determining resource contribution degree based on joint learning and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114553869B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116205313A (en) * | 2023-04-27 | 2023-06-02 | 数字浙江技术运营有限公司 | Federal learning participant selection method and device and electronic equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101528A (en) * | 2020-09-17 | 2020-12-18 | 上海交通大学 | Terminal contribution measurement method based on back propagation |
| CN112506753A (en) * | 2020-12-14 | 2021-03-16 | 德清阿尔法创新研究院 | Efficient contribution evaluation method in federated learning scene |
| CN112926897A (en) * | 2021-04-12 | 2021-06-08 | 平安科技（深圳）有限公司 | Client contribution calculation method and device based on federal learning |
| CN113159190A (en) * | 2021-04-25 | 2021-07-23 | 平安科技（深圳）有限公司 | Federal incentive distribution method and device, computer equipment and storage medium |
| WO2021189906A1 (en) * | 2020-10-20 | 2021-09-30 | 平安科技（深圳）有限公司 | Target detection method and apparatus based on federated learning, and device and storage medium |
| CN113947213A (en) * | 2021-10-19 | 2022-01-18 | 中国电信股份有限公司 | Method, device, storage medium and equipment for measuring contribution of federal learning participants |
-
2022
- 2022-02-11 CN CN202210129960.2A patent/CN114553869B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112101528A (en) * | 2020-09-17 | 2020-12-18 | 上海交通大学 | Terminal contribution measurement method based on back propagation |
| WO2021189906A1 (en) * | 2020-10-20 | 2021-09-30 | 平安科技（深圳）有限公司 | Target detection method and apparatus based on federated learning, and device and storage medium |
| CN112506753A (en) * | 2020-12-14 | 2021-03-16 | 德清阿尔法创新研究院 | Efficient contribution evaluation method in federated learning scene |
| CN112926897A (en) * | 2021-04-12 | 2021-06-08 | 平安科技（深圳）有限公司 | Client contribution calculation method and device based on federal learning |
| CN113159190A (en) * | 2021-04-25 | 2021-07-23 | 平安科技（深圳）有限公司 | Federal incentive distribution method and device, computer equipment and storage medium |
| CN113947213A (en) * | 2021-10-19 | 2022-01-18 | 中国电信股份有限公司 | Method, device, storage medium and equipment for measuring contribution of federal learning participants |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116205313A (en) * | 2023-04-27 | 2023-06-02 | 数字浙江技术运营有限公司 | Federal learning participant selection method and device and electronic equipment |
| CN116205313B (en) * | 2023-04-27 | 2023-08-11 | 数字浙江技术运营有限公司 | Federal learning participant selection method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114553869B (en) | 2024-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111030861B (en) | Edge calculation distributed model training method, terminal and network side equipment | |
| CN114116705B (en) | Method and device for determining contribution value of participants in joint learning | |
| CN114169412B (en) | Federal learning model training method for large-scale industry chain privacy calculation | |
| CN114116707A (en) | Method and device for determining contribution degree of participants in joint learning | |
| CN112818588B (en) | Optimal power flow calculation method, device and storage medium of power system | |
| CN114398634A (en) | Federal learning participant weight calculation method based on information entropy | |
| CN114553869B (en) | Method and device for determining resource contribution degree based on joint learning and electronic equipment | |
| CN118211268A (en) | Heterogeneous federal learning privacy protection method and system based on diffusion model | |
| CN114492847B (en) | Efficient personalized federal learning system and method | |
| CN116776997A (en) | Federal learning model construction method under non-independent co-distribution environment | |
| CN113592541B (en) | Load aggregation business offer optimizing method, device and terminal equipment | |
| CN116402366A (en) | Data contribution evaluation method and device based on joint learning | |
| CN116340959A (en) | Breakpoint privacy protection-oriented method, device, equipment and medium | |
| CN114298319A (en) | Method and device for determining joint learning contribution value, electronic equipment and storage medium | |
| CN114154714A (en) | Time series data prediction method, time series data prediction device, computer equipment and medium | |
| CN113887746A (en) | Method and device for reducing communication pressure based on joint learning | |
| CN116050557A (en) | Power load prediction method, device, computer equipment and medium | |
| CN115564055A (en) | Asynchronous joint learning training method and device, computer equipment and storage medium | |
| CN110162400B (en) | Method and system for realizing cooperation of intelligent agents in MAS system in complex network environment | |
| CN116432010A (en) | Training method and device for joint learning model | |
| WO2023082787A1 (en) | Method and apparatus for determining contribution degree of participant in federated learning, and federated learning training method and apparatus | |
| CN113887744A (en) | Data feature extraction method and device based on joint learning | |
| CN115271042A (en) | Model training method and device based on sample sampling time | |
| CN114298320A (en) | Method and device for calculating contribution value of joint learning, electronic equipment and storage medium | |
| CN113887745A (en) | Data heterogeneous joint learning 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |