WO2022110975A1 - 联邦学习参与者选择方法、装置、设备和存储介质 - Google Patents

联邦学习参与者选择方法、装置、设备和存储介质 Download PDF

Info

Publication number
WO2022110975A1
WO2022110975A1 PCT/CN2021/117779 CN2021117779W WO2022110975A1 WO 2022110975 A1 WO2022110975 A1 WO 2022110975A1 CN 2021117779 W CN2021117779 W CN 2021117779W WO 2022110975 A1 WO2022110975 A1 WO 2022110975A1
Authority
WO
WIPO (PCT)
Prior art keywords
factor
participant
service
data quality
participants
Prior art date
Application number
PCT/CN2021/117779
Other languages
English (en)
French (fr)
Inventor
郭华
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Priority to JP2023532158A priority Critical patent/JP2023550806A/ja
Priority to EP21896479.9A priority patent/EP4250188A4/en
Priority to US18/254,474 priority patent/US20230394365A1/en
Publication of WO2022110975A1 publication Critical patent/WO2022110975A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • 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
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • 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
    • H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
    • H04L41/5003—Managing SLA; Interaction between SLA and QoS
    • H04L41/5009—Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W24/00—Supervisory, monitoring or testing arrangements
    • H04W24/02—Arrangements for optimising operational condition

Definitions

  • the present application relates to the field of communication technologies, and in particular, to a participant selection method, apparatus, device and storage medium.
  • Federated learning is a solution to the privacy and security of data sharing in the field of artificial intelligence.
  • the quality and communication effectiveness of participants in federated learning have a great impact on the results of federated learning.
  • FIG 1 is a schematic diagram of the overall architecture of the federated learning system in the existing communication field; as shown in Figure 1, the Operation and Maintenance Center (OMC) acts as the coordinator or server of federated learning, and the base station acts as a participant or client end.
  • OMC Operation and Maintenance Center
  • the selection of participants can use the historical data of the OMC to avoid the communication overhead caused by frequent interaction with the base station and the impact on the normal service of the base station.
  • the federated learning participant selection method, apparatus, device and storage medium provided in this application.
  • an embodiment of the present application provides a method for selecting a federated learning participant, including: acquiring a plurality of participants to be selected; respectively determining a data quality factor, a business factor, and a stability factor of each of the participants to be selected; The selected participant is determined based on the data quality factor, service factor and stability factor of a plurality of the participants to be selected.
  • an embodiment of the present application provides an apparatus for selecting a federated learning participant, which includes: an acquisition module configured to acquire a plurality of participants to be selected; a factor determination module configured to separately determine each of the participants to be selected data quality factors, business factors and stability factors of the participants; the participant selection module is configured to determine the selected participants based on the data quality factors, business factors and stability factors of a plurality of the participants to be selected.
  • embodiments of the present application provide a federated learning participant selection device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are selected by the one or more programs or multiple processors to execute, so that the one or more processors implement the method according to any one of the embodiments of the present application.
  • an embodiment of the present application provides a storage medium, where a computer program is stored in the storage medium, and when the computer program is executed by a processor, the method according to any one of the embodiments of the present application is implemented.
  • FIG. 1 is a schematic diagram of the overall architecture of a federated learning system in the existing communication field
  • FIG. 2 is a flowchart of a method for selecting a federated learning participant provided by an embodiment of the present application
  • FIG. 3 is a schematic structural diagram of a federated learning participant selection device provided by an embodiment of the present application.
  • FIG. 4 is a schematic structural diagram of a federated learning participant selection device provided by an embodiment of the present application.
  • steps shown in the flowcharts of the figures may be performed in a computer system, such as a set of computer-executable instructions. Also, although a logical order is shown in the flowcharts, in some cases, steps shown or described may be performed in an order different from that herein.
  • GSM Global System of Mobile communication
  • CDMA Code Division Multiple Access
  • Wideband Code Division Multiple Access Wideband Code Division Multiple Access
  • WCDMA Wideband Code Division Multiple Access
  • GPRS General Packet Radio Service
  • LTE Long Term Evolution
  • LIE-A Advanced long term evolution, Advanced Long Term Evolution
  • UMTS Universal Mobile Telecommunication System
  • 5G fifth generation mobile communication technology
  • the base station may be a device capable of communicating with a user terminal.
  • the base station can be any device with wireless transceiver function. Including but not limited to: base station NodeB, evolved base station eNodeB, base station in 5G communication system, base station in future communication system, access node in WiFi system, wireless relay node, wireless backhaul node, etc.
  • the base station may also be a wireless controller in a cloud radio access network (cloud radio access network, CRAN) scenario; the base station may also be a small cell, a transmission reference point (transmission reference point, TRP), etc., which are not limited in the embodiments of the present application.
  • a 5G base station is used as an example for description.
  • the present application provides a method for selecting a federated learning participant. As shown in FIG. 2 , the method for selecting a participant provided by this embodiment mainly includes steps S11 , S12 and S13 .
  • the participant refers to a base station participating in federated learning, which may also be referred to as a client.
  • Participants to be selected refer to the selected participants who want to participate in federated learning.
  • Obtaining the to-be-selected participants may be to select a plurality of to-be-selected participants to communicate with the OMC.
  • determining the data quality factor, service factor and stability factor of the participant to be selected includes: determining the data quality factor, service factor and stability factor of the participant to be selected based on a federated learning type.
  • the federated learning type includes one or more of the following: KPI degradation detection, cell weight optimization, and optical module failure prediction.
  • the data quality factor, service factor and stability factor in different scenarios can be determined according to the type of federated learning.
  • the data quality factor, service factor and stability factor of the to-be-selected participant are determined based on the federated learning type, including:
  • a stability factor is determined based on the time in service of the participant to be selected.
  • the QoS alarm is selected as the data quality factor data
  • the longitude and latitude of the base station is selected as the service factor value
  • the service time of the base station is used as the stability factor available.
  • which base stations are selected to participate in the optimization is based on cell performance data or MR data.
  • determining the selected participant based on the data quality factors, business factors and stability factors of a plurality of the participants to be selected includes: satisfying the data quality factor, business factor and stability factor to an optimal function The participant to be selected with the value is determined as the selected participant.
  • the optimal function is:
  • n refers to the total number of participants to be selected
  • i is the ith participant to be selected
  • data is the data quality factor
  • value is the business factor
  • available is the stability factor
  • a is the first threshold, which is determined by each to-be-selected participant. The value of the selected participant is determined
  • b is the second threshold, determined by the data value of each participant to be selected
  • c is the third threshold, determined by the available value of each participant to be selected.
  • the minimum value after the accumulation of the variance of each participant's service factor value is set as a
  • the accumulated value of data of each base station is set as b
  • the accumulated value of available of each base station is set as c.
  • the values of a, b, and c are set according to the accuracy requirements of the algorithm. After setting an initial value, it can be adjusted with the progress of the algorithm.
  • each base station is brought into the optimization function f(data, value, available), and as long as the value of the formula is satisfied, the base station is selected.
  • the selected participant is determined based on the data quality factor, service factor and stability factor of a plurality of the to-be-selected participants
  • the method includes: determining the selection parameters of each participant to be selected based on the data quality factors, business factors and stability factors of a plurality of the participants to be selected; The participant to be selected is determined as the selected participant.
  • determining the data quality factor, business factor and stability factor of the participant to be selected based on the federated learning type includes: based on the to-be-selected participant The performance data integrity parameter of the participant determines the data quality factor; the service factor is determined based on the QoS alarm parameter of the participant to be selected; the stability factor is determined based on the service time of the participant to be selected.
  • the performance data integrity is selected as the data quality factor data
  • the QoS alarm is selected as the service factor value
  • the time when the base station is in service is used as the stability factor available.
  • Time refers to the total time of performance data collection
  • Time (supplementary collection) is the length of time for supplementary collection of performance data.
  • the KPI degradation detection is taken as an example. It is necessary to select sites where KPI degradation often occurs, so as to analyze more degradation reasons. Therefore, a QoS alarm is selected as the service factor value.
  • the selection parameters of each participant to be selected are determined based on the data quality factors, service factors and stability factors of the plurality of participants to be selected, including: for each participant to be selected, the data The quality factor, business factor and stability factor are respectively normalized by linear normalization method; the normalized data quality factor, business factor and stability factor are first weighted to obtain the selection of the participants to be selected. parameter.
  • the above data quality factor data, business factor value and stability factor available are not in one dimension, so normalization processing is required.
  • the above-mentioned linear normalization method is used for normalization respectively, and the normalized data quality factor Data1, business factor Value1 and stability factor Available1 are obtained.
  • the data of the above-mentioned dimensions can be accumulated.
  • f(data, value, available) A1 ⁇ Data1+B1 ⁇ Value1+C1 ⁇ Available1, where the weight factors A1, B1, and C1 can be set according to experience , but A+B+C should be 1.
  • the calculations are performed according to the algorithm described above for each base station.
  • the final results are sorted in descending order, and the top L base stations are determined as the selected base stations.
  • L is a positive integer, which can be set according to the actual situation.
  • determining the data quality factor, service factor and stability factor of the participant to be selected based on the federated learning type includes: based on the to-be-selected participant The data quality factor is determined based on the performance data integrity parameter of the selected participant; the service factor is determined based on the quality parameter of the optical module link of the to-be-selected participant; and the stability factor is determined based on the service time of the to-be-selected participant.
  • the performance data integrity is selected as the data quality factor data
  • the service time of the optical module link is selected to represent the service factor value
  • the service time of the base station is used as the stability factor available.
  • the integrity of the performance data is selected as the data quality factor.
  • determining the selection parameters of each participant to be selected based on the data quality factor, service factor and stability factor of a plurality of the participants to be selected including: for each participant to be selected, the data The quality factor, the service factor and the stability factor are subjected to a second weighting process to obtain the selection parameters of the participants to be selected.
  • a, b, and c are adjusted as weight factors according to empirical settings.
  • the calculations are performed according to the algorithm described above for each base station.
  • the final results are sorted in descending order, and the top L base stations are determined as the selected base stations.
  • L is a positive integer, which can be set according to the actual situation.
  • KPI degradation is used to detect the management of federated learning participants, and this embodiment requires that a cell with relatively obvious KPI degradation be selected as a participant.
  • Step 1 Eliminate base stations that are in peak business hours, sleep or energy-saving states.
  • Step 2 Select performance data integrity as the data quality factor data.
  • Time refers to the total time of performance data collection
  • Time (supplementary collection) is the length of time for supplementary collection of performance data.
  • Step 3 Select QoS alarm as the service factor value.
  • QoS alarms are selected as the service factor value.
  • Step 5 The above-mentioned data quality factor data, business factor value and stability factor available are respectively normalized by using a linear normalization method to obtain normalized data quality factor Data1, business factor Value1 and stability factor Available1.
  • Step 6 After normalization, the data of the above-mentioned dimensions can be accumulated.
  • Step 7 Several base stations can be randomly selected, or each base station can be calculated according to the above algorithm. The final results are sorted in descending order, and the top L base stations are determined as the selected base stations. Among them, L is a positive integer, which can be set according to the actual situation.
  • cell weight optimization is taken as an example for description.
  • Cell weight optimization provides automatic adjustment of cell broadcast weights. Users create a task for an area and set appropriate parameters. The system automatically adjusts the weights through a series of steps of measuring, optimizing and calculating the optimal weight group, evaluating and restoring the weights.
  • Step 1 Eliminate base stations that are in peak business hours, sleep or energy-saving states.
  • Step 2 Select the QoS alarm as the data quality factor data.
  • which base stations are selected to participate in the optimization is based on cell performance data or MR data, and cells with poor performance participate in the optimization, so QoS alarms are selected as the data quality factor, and the ratio of QoS alarm occurrences is used as the data quality factor data
  • the value of , that is, data Num (QoS alarm)/Num (alarm).
  • Step 3 Select the longitude and latitude of the base station as the service factor value.
  • Step 5 For the service factor value, the latitude and longitude of each base station needs to be the closest. Therefore, the minimum value after the accumulation of the variance of the service factor value of each base station is set to a, the threshold value of data accumulation of each base station is set to b, and the available base station is accumulated. The threshold is set to c, then take f() as the optimal function as follows:
  • n refers to the total number of participants to be selected
  • i is the ith participant to be selected
  • data is the data quality factor
  • value is the business factor
  • available is the stability factor
  • a is the first threshold, determined by each The value of the participant to be selected is determined
  • b is the second threshold, determined by the data value of each participant to be selected
  • c is the third threshold, determined by the available value of each participant to be selected.
  • the values of a, b, and c are set according to the accuracy requirements of the algorithm. After setting an initial value, it can be adjusted with the progress of the algorithm.
  • Step 6 Randomly select several base stations, and bring the value of each base station into f(data, value, available). As long as the value of the formula is satisfied, the base station is selected.
  • optical module failure prediction mainly uses data such as bit error data of the optical port of the optical module for prediction.
  • Step 1 Eliminate base stations that are in peak business hours, sleep or energy-saving states
  • Step 2 Select performance data integrity as the data quality factor data.
  • the integrity of the performance data is selected as the data quality factor.
  • Step 3 Select the service time of the optical module link to represent the service factor value.
  • Step 6 Randomly select several base stations, or calculate each base station according to the above algorithm. The final results are sorted in descending order, and the top L base stations are determined as the selected base stations. Among them, L is a positive integer, which can be set according to the actual situation.
  • the present application provides a participant selection apparatus.
  • the participant selection apparatus provided by this embodiment mainly includes an acquisition module 31 , a factor determination module 32 and a participant selection module 33 .
  • an acquisition module 31 configured to acquire a plurality of participants to be selected
  • the factor determination module 32 is configured to determine the data quality factor, service factor and stability factor of each of the to-be-selected participants respectively;
  • the participant selection module 33 is configured to determine the selected participant based on the data quality factor, service factor and stability factor of the plurality of participants to be selected.
  • the participant selection device includes: acquiring a plurality of participants to be selected, respectively determining the data quality factor, service factor and stability factor of each participant to be selected, and based on the data quality of the plurality of participants to be selected Factors, business factors and stability factors determine the selected participants.
  • data quality factors, business factors, and stability factors are introduced to select participants, and the results are quantified to facilitate the selection of participants. Quality federated learning participants and improve the effect of federated learning.
  • determining the data quality factor, service factor and stability factor of the participant to be selected includes: determining the data quality factor, service factor and stability factor of the participant to be selected based on a federated learning type.
  • the federated learning type includes one or more of the following: KPI degradation detection, cell weight optimization, and optical module failure prediction.
  • the data quality factor, service factor and stability factor of the to-be-selected participant are determined based on the federated learning type, including:
  • a stability factor is determined based on the time in service of the participant to be selected.
  • determining the selected participant based on the data quality factors, business factors and stability factors of a plurality of the participants to be selected includes: satisfying the data quality factor, business factor and stability factor to an optimal function The participant to be selected with the value is determined as the selected participant.
  • the optimal function is:
  • n refers to the total number of participants to be selected
  • i is the ith participant to be selected
  • data is the data quality factor
  • value is the business factor
  • available is the stability factor
  • a is the first threshold, which is determined by each to-be-selected participant. The value of the selected participant is determined
  • b is the second threshold, determined by the data value of each participant to be selected
  • c is the third threshold, determined by the available value of each participant to be selected.
  • the selected participant is determined based on the data quality factor, service factor and stability factor of a plurality of the to-be-selected participants
  • the method includes: determining the selection parameters of each participant to be selected based on the data quality factors, business factors and stability factors of a plurality of the participants to be selected; The participant to be selected is determined as the selected participant.
  • determining the data quality factor, business factor and stability factor of the participant to be selected based on the federated learning type includes: based on the to-be-selected participant The performance data integrity parameter of the participant determines the data quality factor; the service factor is determined based on the QoS alarm parameter of the participant to be selected; the stability factor is determined based on the service time of the participant to be selected.
  • the selection parameters of each participant to be selected are determined based on the data quality factors, service factors and stability factors of the plurality of participants to be selected, including: for each participant to be selected, the data The quality factor, business factor and stability factor are respectively normalized by linear normalization method; the normalized data quality factor, business factor and stability factor are first weighted to obtain the selection of the participants to be selected. parameter.
  • determining the data quality factor, service factor and stability factor of the participant to be selected based on the federated learning type includes: based on the to-be-selected participant The data quality factor is determined based on the performance data integrity parameter of the selected participant; the service factor is determined based on the quality parameter of the optical module link of the to-be-selected participant; and the stability factor is determined based on the service time of the to-be-selected participant.
  • determining the selection parameters of each participant to be selected based on the data quality factor, service factor and stability factor of a plurality of the participants to be selected including: for each participant to be selected, the data The quality factor, the service factor and the stability factor are subjected to a second weighting process to obtain the selection parameters of the participants to be selected.
  • the participant selection apparatus provided in this embodiment can execute the participant selection method provided by any embodiment of the present application, and has corresponding functional modules and beneficial effects for performing the method.
  • the participant selection method provided by any embodiment of this application can execute the participant selection method provided by any embodiment of this application.
  • the included units and modules are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be realized; in addition, The specific names of the functional units are only for the convenience of distinguishing from each other, and are not used to limit the protection scope of the present application.
  • FIG. 4 is a schematic structural diagram of a device for selecting a Lianbang study participant provided by an embodiment of the present application.
  • the device includes a processor 41 , memory 42, input device 43, output device 44 and communication device 45; the number of processors 41 in the device can be one or more, and one processor 41 is taken as an example in FIG. 4; the processor 41, memory 42 in the device , the input device 43 and the output device 44 may be connected through a bus or other means, and the connection through a bus is taken as an example in FIG. 4 .
  • the memory 42 can be used to store software programs, computer-executable programs, and modules, such as program instructions/modules corresponding to the method for selecting Lianbang learning participants in the embodiments of the present application (for example, Lianbang Learning).
  • the participant selects the acquisition module 31, the factor determination module 32 and the participant selection module 33) in the device.
  • the processor 41 executes various functional applications and data processing of the device by running the software programs, instructions, and modules stored in the memory 42, ie, implements any method provided by the embodiments of the present application.
  • the memory 42 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the device, and the like. Additionally, memory 42 may include high speed random access memory, and may also include nonvolatile memory, such as at least one magnetic disk storage device, flash memory device, or other nonvolatile solid state storage device. In some instances, memory 42 may further include memory located remotely from processor 41, which may be connected to the device through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
  • the input device 43 may be configured to receive input numerical or character information, and to generate key signal input related to user settings and function control of the device.
  • the output device 44 may include a display device such as a display screen.
  • the communication device 45 may include a receiver and a transmitter.
  • the communication device 45 is configured to transmit and receive information according to the control of the processor 41 .
  • the embodiments of the present application further provide a storage medium containing computer-executable instructions, the computer-executable instructions, when executed by a computer processor, are used to execute a method for selecting a federal learning participant, comprising: :
  • the selected participant is determined based on the data quality factor, service factor and stability factor of a plurality of the participants to be selected.
  • a storage medium containing computer-executable instructions provided by the embodiments of the present application, the computer-executable instructions of which are not limited to the above-mentioned method operations, and can also execute the participant selection method provided by any embodiment of the present application. related operations.
  • the method, device, device, and storage medium for selecting federated learning participants include: acquiring multiple participants to be selected, and separately determining the data quality factor, service factor, and stability factor of each participant to be selected, based on The data quality factors, business factors and stability factors of multiple participants to be selected determine the selected participants.
  • the data quality factors, business factors, and stability factors are introduced to select participants, and the results are quantified as It is convenient to select participants, select high-quality federated learning participants, and improve the effect of federated learning.
  • the present application can be implemented by means of software and necessary general-purpose hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation manner .
  • the technical solutions of the present application can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products can be stored in a computer-readable storage medium, such as a floppy disk of a computer , read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including several instructions to make a computer device (which can be a personal computer , server, or network device, etc.) to execute the methods described in the various embodiments of this application.
  • a computer-readable storage medium such as a floppy disk of a computer , read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc.
  • user terminal encompasses any suitable type of wireless user equipment, such as a mobile telephone, portable data processing device, portable web browser or vehicle mounted mobile station.
  • the various embodiments of the present application may be implemented in hardware or special purpose circuits, software, logic, or any combination thereof.
  • some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor or other computing device, although the application is not limited thereto.
  • Embodiments of the present application may be implemented by the execution of computer program instructions by a data processor of a mobile device, eg in a processor entity, or by hardware, or by a combination of software and hardware.
  • the computer program instructions may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state setting data, or source code written in any combination of one or more programming languages or object code.
  • ISA instruction set architecture
  • the block diagrams of any logic flow in the figures of the present application may represent program steps, or may represent interconnected logic circuits, modules and functions, or may represent a combination of program steps and logic circuits, modules and functions.
  • Computer programs can be stored on memory.
  • the memory may be of any type suitable for the local technical environment and may be implemented using any suitable data storage technology, such as but not limited to read only memory (ROM), random access memory (RAM), optical memory devices and systems (Digital Versatile Discs). DVD or CD disc) etc.
  • Computer-readable media may include non-transitory storage media.
  • the data processor may be of any type suitable for the local technical environment, such as, but not limited to, a general purpose computer, special purpose computer, microprocessor, digital signal processor (DSP), application specific integrated circuit (ASIC), programmable logic device (FGPA) and processors based on multi-core processor architectures.
  • DSP digital signal processor
  • ASIC application specific integrated circuit
  • FGPA programmable logic device

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Databases & Information Systems (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Telephonic Communication Services (AREA)

Abstract

一种联邦学习参与者选择方法、装置、设备和存储介质，该联邦学习参与者选择方法包括：获取多个待选择参与者(S11)，分别确定每个待选择参与者的数据质量因子，业务因子和稳定性因子(S12)，基于多个待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者(S13)。

Description

联邦学习参与者选择方法、装置、设备和存储介质
相关申请的交叉引用
本申请基于申请号为202011345470.3、申请日为2020年11月25日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本申请涉及通信技术领域，具体涉及一种参与者选择方法、装置、设备和存储介质。
背景技术
联邦学习是为了解决人工智能领域数据共享的私密性和安全性的一种解决方案。联邦学习中参与者的质量和通信效果，对联邦学习的结果影响很大。
图1是现有的通信领域中联邦学习系统的总体架构的示意图；如图1所示，操作维护中心(Operation and Maintenance Center，OMC)作为联邦学习的协调者或服务器，基站作为参与者或客户端。在该架构中，参与者的选择可以利用OMC的历史数据，避免与基站频繁交互带来的通信开销，以及对基站正常业务的影响。
发明内容
本申请提供的联邦学习参与者选择方法、装置、设备和存储介质。
第一方面，本申请实施例提供一种联邦学习参与者选择方法，包括：获取多个待选择参与者；分别确定每个所述待选择参与者的数据质量因子，业务因子和稳定性因子；基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者。
第二方面，本申请实施例提供一种联邦学习参与者选择装置，其包括：获取模块，被配置为获取多个待选择参与者；因子确定模块，被配置为分别确定每个所述待选择参与者的数据质量因子，业务因子和稳定性因子；参与者选择模块，被配置为基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者。
第三方面，本申请实施例提供一种联邦学习参与者选择设备，包括：一个或多个处理器；存储器，用于存储一个或多个程序；当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如本申请实施例提供的任一项所述的方法。
第四方面，本申请实施例提供一种存储介质，所述存储介质存储有计算机程序，所述计算机程序被处理器执行时实现如本申请实施例提供的任一项所述的方法。
关于本申请的以上实施例和其他方面以及其实现方式，在附图说明、具体实施方式和权利要求中提供更多说明。
附图说明
图1是现有的通信领域中联邦学习系统的总体架构的示意图；
图2是本申请实施例提供的一种联邦学习参与者选择方法的流程图；
图3是本申请实施例提供的一种联邦学习参与者选择装置的结构示意图；
图4是本申请实施例提供的一种联邦学习参与者选择设备的结构示意图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚明白，下文中将结合附图对本申请的实施例进行详细说明。需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互任意组合。
在附图的流程图示出的步骤可以在诸如一组计算机可执行指令的计算机系统中执行。并且，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
本申请的技术方案可以应用于各种通信系统，例如：全球移动通讯(Global System of Mobile communication，GSM)系统、码分多址(Code Division Multiple Access，CDMA)系统、宽带码分多址(Wideband Code Division Multiple Access，WCDMA)系统、通用分组无线业务(General Packet Radio Service，GPRS)、长期演进(Long Term Evolution，LTE)系统、LIE-A(Advanced long term evolution，先进的长期演进)系统、通用移动通信系统(Universal Mobile Telecommunication System，UMTS)、以及第五代移动通信技术(5th generation wireless systems，5G)系统等，本申请实施例并不限定。在本申请中以5G系统为例进行说明。
本申请实施例中，基站可以是能和用户终端进行通信的设备。基站可以是任意一种具有无线收发功能的设备。包括但不限于：基站NodeB、演进型基站eNodeB、5G通信系统中的基站、未来通信系统中的基站、WiFi系统中的接入节点、无线中继节点、无线回传节点等。基站还可以是云无线接入网络(cloud radio access network，CRAN)场景下的无线控制器；基站还可以是小站，传输节点(transmission reference point，TRP)等，本申请实施例并不限定。在本申请中以5G基站为例进行说明。
在一个实施例中，本申请提供一种联邦学习参与者选择方法，如图2所示，本实施例提供的参与者选择方法主要包括步骤S11、S12和S13。
S11、获取多个待选择参与者；
S12、分别确定每个所述待选择参与者的数据质量因子，业务因子和稳定性因子；
S13、基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者。
在本实施例中，所述参与者是指参与联邦学习的基站，也可以称为客户端。待选择参与者是指想要参加联邦学习，被选择的参与者。
获取待选择参与者可以是选取与OMC进行通信的多个待选择参与者。
在一个实施方式中，确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子。
在一个实施方式中，所述联邦学习类型包括如下一个或多个：KPI劣化检测，小区权重优化，光模块故障预测。
在本实施例中，可以根据联邦学习的类型来确定不同场景下的数据质量因子，业务因子和稳定性因子。
在一个实施方式中，在所述联邦学习类型是小区权重优化的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：
基于所述待选择参与者的QoS告警参数确定数据质量因子；
基于所述待选择参与者的经纬度确定业务因子；
基于所述待选择参与者的在服时间确定稳定性因子。
在本实施例中，选择QoS告警作为数据质量因子data，选择基站的经纬度作为业务因子value，将基站在服的时间作为稳定度因子available。
本实施例中是挑选哪些基站参与优化。哪些小区参加优化是根据小区性能数据或是MR数据来的。小区性能差的基站参与优化，因此选择QoS告警作为数据质量因子，将QoS告警出现的比率作为数据质量因子data的取值，即data＝Num(QoS告警)/Num(告警)。
本实施例中，天线权值优化跟基站的地理位置相关，一个区域的基站地理位置要有相关性。因此，选择基站的经纬度作为业务因子value，即value＝[atitude,longitude]；atitude是指纬度值，longitude是指经度值。
将基站在服的时间作为稳定度因子available，即available＝time(在服)/time。
在一个实施方式中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，包括：将数据质量因子，业务因子和稳定性因子满足最优函数取值的待选择参与者确定为被选中的参与者。
在一个实施方式中，所述最优函数为：
Figure PCTCN2021117779-appb-000001
Figure PCTCN2021117779-appb-000002
其中，n是指待选择参与者的总个数，i是第i个待选择参与者，data是数据质量因子，value是业务因子，available是稳定性因子，a是第一阈值，由各个待选择参与者的value值确定，b是第二阈值，由各个待选择参与者的data值确定，c是第三阈值，由各个待选择参与者的available值确定。
在本实施例中，各参与者业务因子value值方差的累加之后的最小值设为a，将各基站data累加值设置为b，将各基站available累加值设置为c。a,b,c的取值根据对该算法的精度要求设置，设置一个初始值之后可以随着算法进度调整。
在一些实例中，将每个基站的取值带入优化函数f(data,value,available)，只要满足该公式取值，则该基站被选中。
在一个实施方式中，在所述联邦学习类型是KPI劣化检测或光模块故障预测的情况下，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，包括：基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数；将所述选取参数进行排序，将排在前L个的所述待选择参与者确定为被选中的参与者。
在一个实施方式中，在所述联邦学习类型是KPI劣化检测的情况下，基于联邦学习类型 确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：基于所述待选择参与者的性能数据完整性参数确定数据质量因子；基于所述待选择参与者的QoS告警参数确定业务因子；基于所述待选择参与者的在服时间确定稳定性因子。
在本实施例中，选择性能数据完整性作为数据质量因子data，选择QoS告警作为业务因子value，将基站在服的时间作为稳定度因子available。
本实施例中，上述数据质量因子的计算方法为计算正常性能数据采集时间比例，即data＝1-Time(补采)/Time。Time是指性能数据采集总时长，Time(补采)是补充采集性能数据的时长。
本实施例中，KPI劣化检测为例，需要挑选出经常出现KPI劣化的站点，这样才能更多分析出劣化原因，因此选择QoS告警作为业务因子value。将QoS告警出现的比率作为该业务因子value取值，即value＝Num(QoS告警)/Num(告警)；Num(告警)是指告警的总数，Num(QoS告警)是指QoS告警的次数。
本实施例中，将基站在服的时间作为稳定度因子available，即available＝time(在服)/time；time(在服)基站在服的时间。
在一个实施方式中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数，包括：针对每个待选择的参与者，对数据质量因子，业务因子和稳定性因子分别采用线性标准化方法进行归一化处理；将归一化处理后的数据质量因子，业务因子和稳定性因子进行第一加权处理，得到待选择参与者的选取参数。
将上述数据质量因子data，业务因子value和稳定度因子available不在一个维度内，因此需要进行归一化处理。分别对上述因此采用线性标准化方法进行归一化处理，得到归一化的的数据质量因子Data1，业务因子Value1和稳定度因子Available1。
归一化之后，将上述几个维度的数据可以进行累加处理。
由于需要考虑这几个因子在该算法中的权重，因此取f(data,value,available)＝A1×Data1+B1×Value1+C1×Available1，其中的权重因子A1、B1、C1可以根据经验设置，但A+B+C应为1。
在一些实例中，对每个基站都按照上述算法进行计算。并对最终得到的结果按照从大到小排序，将排在前L个的基站确定为被选中的基站。其中，L为正整数，可以根据实际情况进行设置。
在一个实施方式中，在所述联邦学习类型是光模块故障预测的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：基于所述待选择参与者的性能数据完整性参数确定数据质量因子；基于所述待选择参与者的光模块链路的质量参数确定业务因子；基于所述待选择参与者的在服时间确定稳定性因子。
在本实施例中，选择性能数据完整性作为数据质量因子data，选择光模块链路的在服时间表示业务因子value，将基站在服的时间作为稳定度因子available。
在本实施例中，由于算法主要利用的误码率数据是通过性能数据上报的，因此选择性能数据完整性作为数据质量因子。该因子的计算方法为计算正常性能数据采集时间比例，即data＝1-Time(补采)/Time；
在本实施例中，预测所需的其他数据要从光模块获取，因此需要保证光模块链路的稳定性。用光模块链路的质量作为业务因子，该因子使用光模块链路的在服时间表示，即value＝time(光模块在服)/time。
在本实施例中，将基站在服的时间作为稳定度因子available，即available＝time(在服)/time。
在一个实施方式中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数，包括：针对每个待选择的参与者，将数据质量因子，业务因子和稳定性因子进行第二加权处理，得到待选择参与者的选取参数。
本实施例中，上述各因子的维度相同，不需要进行归一化处理，因此取f(data,value,available)＝a2×data+b2×value+c2×available即可。其中a,b,c作为权重因子根据经验设置调整。
在一些实例中，对每个基站都按照上述算法进行计算。并对最终得到的结果按照从大到小排序，将排在前L个的基站确定为被选中的基站。其中，L为正整数，可以根据实际情况进行设置。
在一个应用性实例中，以KPI劣化检测联邦学习参与者管理，本实施例要求挑选KPI劣化较为明显的小区作为参与者。
步骤一：剔除处于业务高峰期、睡眠或节能状态的基站。
步骤二：选择性能数据完整性作为数据质量因子data。
该数据质量因子的计算方法为计算正常性能数据采集时间比例，即data＝1-Time(补采)/Time。Time是指性能数据采集总时长，Time(补采)是补充采集性能数据的时长。
步骤三：选择QoS告警作为业务因子value。
KPI劣化检测为例，需要挑选出经常出现KPI劣化的站点，这样才能更多分析出劣化原因，因此选择QoS告警作为业务因子value。将QoS告警出现的比率作为该业务因子value取值，即value＝Num(QoS告警)/Num(告警)；Num(告警)是指告警的总数，Num(QoS告警)是指QoS告警的次数。
步骤四：将基站在服的时间作为稳定度因子available，即available＝time(在服)/time；time(在服)基站在服的时间。
步骤五：将上述数据质量因子data，业务因子value和稳定度因子available分别采用线性标准化方法进行归一化处理，得到归一化的数据质量因子Data1，业务因子Value1和稳定度因子Available1。
Figure PCTCN2021117779-appb-000003
Figure PCTCN2021117779-appb-000004
Figure PCTCN2021117779-appb-000005
步骤六：归一化之后，将上述几个维度的数据可以进行累加处理。
还需要考虑这几个因子在该算法中的权重，因此取f(data,value,available)＝A1×Data1+B1×Value1+C1×Available1，其中的权重因子A1、B1、C1可以根据经验设置，但A+B+C应为1。
步骤七：可以随机挑选若干基站，也可以对每个基站都按照上述算法进行计算。并对最终得到的结果按照从大到小排序，将排在前L个的基站确定为被选中的基站。其中，L为正 整数，可以根据实际情况进行设置。
在一个应用性实例中，小区权值优化为例进行说明。小区权值优化提供小区广播权值的自动调整，用户为一个区域建一个任务，设置合适的参数。系统通过测量、优化计算最优权值组、评估、恢复权值一系列步骤进行权值的自动调整。
步骤一：剔除处于业务高峰期、睡眠或节能状态的基站。
步骤二：选择QoS告警作为数据质量因子data。
本实施例中是挑选哪些基站参与优化，是根据小区性能数据或是MR数据来的，小区性能差的参与优化，因此选择QoS告警作为数据质量因子，将QoS告警出现的比率作为数据质量因子data的取值，即data＝Num(QoS告警)/Num(告警)。
步骤三：选择基站的经纬度作为业务因子value。
本实施例中，天线权值优化为例，跟基站的地理位置相关，一个区域的基站地理位置要有相关性。因此，选择基站的经纬度作为业务因子value，即value＝[atitude,longitude]；atitude是指纬度值，longitude是指经度值。
步骤四：将基站在服的时间作为稳定度因子available，即available＝time(在服)/time。
步骤五：对于业务因子value需要满足各基站的经纬度最接近，因此各基站业务因子value值方差的累加之后的最小值设为a，将各基站data累加的阈值设置为b，将各基站available累加的阈值设置为c，那么取f()为最优函数如下：
Figure PCTCN2021117779-appb-000006
其中其中，n是指待选择参与者的总个数，i是第i个待选择参与者，data是数据质量因子，value是业务因子，available是稳定性因子，a是第一阈值，由各个待选择参与者的value值确定，b是第二阈值，由各个待选择参与者的data值确定，c是第三阈值，由各个待选择参与者的available值确定。
a,b,c的取值根据对该算法的精度要求设置，设置一个初始值之后可以随着算法进度调整。
步骤六：随机挑选若干基站，将每个基站的取值带入f(data,value,available)，只要满足该公式取值，则该基站被选中。
在一个应用性实例中，光模块故障预测。该算法主要利用光模块的光口误码数据等数据进行预测。
步骤一：剔除处于业务高峰期、睡眠或节能状态的基站；
步骤二：选择性能数据完整性作为数据质量因子data。
由于算法主要利用的误码率数据是通过性能数据上报的，因此选择性能数据完整性作为 数据质量因子。该因子的计算方法为计算正常性能数据采集时间比例，即data＝1-Time(补采)/Time；
步骤三：选择光模块链路的在服时间表示业务因子value。
预测所需的其他数据要从光模块获取，因此需要保证光模块链路的稳定性。用光模块链路的质量作为业务因子，该因子使用光模块链路的在服时间表示，即value＝time(光模块在服)/time。
步骤四：将基站在服的时间作为稳定度因子available，即available＝time(在服)/time。
步骤五：上述各因子的维度相同，因此取f(data,value,available)＝a2×data+b2×value+c2×available即可。其中a2,b2,c2作为权重因子根据经验设置调整。
步骤六：随机挑选若干基站，也可以对每个基站都按照上述算法进行计算。并对最终得到的结果按照从大到小排序，将排在前L个的基站确定为被选中的基站。其中，L为正整数，可以根据实际情况进行设置。
在一个实施例中，本申请提供一种参与者选择装置，如图3所示，本实施例提供的参与者选择装置主要包括获取模块31、因子确定模块32和参与者选择模块33。
获取模块31，被配置为获取多个待选择参与者；
因子确定模块32，被配置为分别确定每个所述待选择参与者的数据质量因子，业务因子和稳定性因子；
参与者选择模块33，被配置为基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者。
本实施例提供的参与者选择装置，包括：获取多个待选择参与者，分别确定每个待选择参与者的数据质量因子，业务因子和稳定性因子，基于多个待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，本实施例中通过引入数据质量因子、业务因子、稳定性因子来挑选参与者，并将结果量化以方便挑选参与者，以选出高质量联邦学习参与者，提高联邦学习的效果。
在一个实施方式中，确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子。
在一个实施方式中，所述联邦学习类型包括如下一个或多个：KPI劣化检测，小区权重优化，光模块故障预测。
在一个实施方式中，在所述联邦学习类型是小区权重优化的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：
基于所述待选择参与者的QoS告警参数确定数据质量因子；
基于所述待选择参与者的经纬度确定业务因子；
基于所述待选择参与者的在服时间确定稳定性因子。
在一个实施方式中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，包括：将数据质量因子，业务因子和稳定性因子满足最优函数取值的待选择参与者确定为被选中的参与者。
在一个实施方式中，所述最优函数为：
Figure PCTCN2021117779-appb-000007
Figure PCTCN2021117779-appb-000008
Figure PCTCN2021117779-appb-000009
其中，n是指待选择参与者的总个数，i是第i个待选择参与者，data是数据质量因子，value是业务因子，available是稳定性因子，a是第一阈值，由各个待选择参与者的value值确定，b是第二阈值，由各个待选择参与者的data值确定，c是第三阈值，由各个待选择参与者的available值确定。
在一个实施方式中，在所述联邦学习类型是KPI劣化检测或光模块故障预测的情况下，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，包括：基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数；将所述选取参数进行排序，将排在前L个的所述待选择参与者确定为被选中的参与者。
在一个实施方式中，在所述联邦学习类型是KPI劣化检测的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：基于所述待选择参与者的性能数据完整性参数确定数据质量因子；基于所述待选择参与者的QoS告警参数确定业务因子；基于所述待选择参与者的在服时间确定稳定性因子。
在一个实施方式中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数，包括：针对每个待选择的参与者，对数据质量因子，业务因子和稳定性因子分别采用线性标准化方法进行归一化处理；将归一化处理后的数据质量因子，业务因子和稳定性因子进行第一加权处理，得到待选择参与者的选取参数。
在一个实施方式中，在所述联邦学习类型是光模块故障预测的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：基于所述待选择参与者的性能数据完整性参数确定数据质量因子；基于所述待选择参与者的光模块链路的质量参数确定业务因子；基于所述待选择参与者的在服时间确定稳定性因子。
在一个实施方式中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数，包括：针对每个待选择的参与者，将数据质量因子，业务因子和稳定性因子进行第二加权处理，得到待选择参与者的选取参数。
本实施例中提供的参与者选择装置可执行本申请任意实施例所提供的参与者选择方法，具备执行该方法相应的功能模块和有益效果。未在本实施例中详尽描述的技术细节，可参见本申请任意实施例所提供的参与者选择方法。
值得注意的是，上述参与者选择装置的实施例中，所包括的各个单元和模块只是按照功能逻辑进行划分的，但并不局限于上述的划分，只要能够实现相应的功能即可；另外，各功能单元的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。
本申请实施例还提供一种连邦学习参与者选择设备，图4是本申请实施例提供的一种连邦学习参与者选择设备的结构示意图，如图4所示，该设备包括处理器41、存储器42、输入装置43、输出装置44和通信装置45；设备中处理器41的数量可以是一个或多个，图4中以一个处理器41为例；设备中的处理器41、存储器42、输入装置43和输出装置44可以通过 总线或其他方式连接，图4中以通过总线连接为例。
存储器42作为一种计算机可读存储介质，可用于存储软件程序、计算机可执行程序以及模块，如本申请实施例中的连邦学习参与者选择方法对应的程序指令/模块(例如，连邦学习参与者选择装置中的获取模块31、因子确定模块32和参与者选择模块33)。处理器41通过运行存储在存储器42中的软件程序、指令以及模块，从而执行设备的各种功能应用以及数据处理，即实现本申请实施例提供的任一方法。
存储器42可主要包括存储程序区和存储数据区，其中，存储程序区可存储操作系统、至少一个功能所需的应用程序；存储数据区可存储根据设备的使用所创建的数据等。此外，存储器42可以包括高速随机存取存储器，还可以包括非易失性存储器，例如至少一个磁盘存储器件、闪存器件、或其他非易失性固态存储器件。在一些实例中，存储器42可进一步包括相对于处理器41远程设置的存储器，这些远程存储器可以通过网络连接至设备。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
输入装置43可设置为接收输入的数字或字符信息，以及产生与设备的用户设置以及功能控制有关的键信号输入。输出装置44可包括显示屏等显示设备。
通信装置45可以包括接收器和发送器。通信装置45设置为根据处理器41的控制进行信息收发通信。
在一个实施方式中，本申请实施例还提供一种包含计算机可执行指令的存储介质，所述计算机可执行指令在由计算机处理器执行时用于执行一种连邦学习参与者选择方法，包括：
获取多个待选择参与者；
分别确定每个所述待选择参与者的数据质量因子，业务因子和稳定性因子；
基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者。
当然，本申请实施例所提供的一种包含计算机可执行指令的存储介质,其计算机可执行指令不限于如上所述的方法操作,还可以执行本申请任意实施例所提供的参与者选择方法中的相关操作。
本实施例提供的联邦学习参与者选择方法、装置、设备和存储介质，包括：获取多个待选择参与者，分别确定每个待选择参与者的数据质量因子，业务因子和稳定性因子，基于多个待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，本实施例中通过引入数据质量因子、业务因子、稳定性因子来挑选参与者，并将结果量化以方便挑选参与者，选出高质量联邦学习参与者，提高联邦学习的效果。
通过以上关于实施方式的描述，所属领域的技术人员可以清楚地了解到，本申请可借助软件及必需的通用硬件来实现，当然也可以通过硬件实现，但很多情况下前者是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中，如计算机的软盘、只读存储器(Read-Only Memory,ROM)、随机存取存储器(Random Access Memory,RAM)、闪存(FLASH)、硬盘或光盘等，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述的方法。
以上所述，仅为本申请的一些实施例而已，并非用于限定本申请的保护范围。
本领域内的技术人员应明白，术语用户终端涵盖任何适合类型的无线用户设备，例如移动电话、便携数据处理装置、便携网络浏览器或车载移动台。
一般来说，本申请的多种实施例可以在硬件或专用电路、软件、逻辑或其任何组合中实现。例如，一些方面可以被实现在硬件中，而其它方面可以被实现在可以被控制器、微处理器或其它计算装置执行的固件或软件中，尽管本申请不限于此。
本申请的实施例可以通过移动装置的数据处理器执行计算机程序指令来实现，例如在处理器实体中，或者通过硬件，或者通过软件和硬件的组合。计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码。
本申请附图中的任何逻辑流程的框图可以表示程序步骤，或者可以表示相互连接的逻辑电路、模块和功能，或者可以表示程序步骤与逻辑电路、模块和功能的组合。计算机程序可以存储在存储器上。存储器可以具有任何适合于本地技术环境的类型并且可以使用任何适合的数据存储技术实现，例如但不限于只读存储器(ROM)、随机访问存储器(RAM)、光存储器装置和系统(数码多功能光碟DVD或CD光盘)等。计算机可读介质可以包括非瞬时性存储介质。数据处理器可以是任何适合于本地技术环境的类型，例如但不限于通用计算机、专用计算机、微处理器、数字信号处理器(DSP)、专用集成电路(ASIC)、可编程逻辑器件(FGPA)以及基于多核处理器架构的处理器。
通过非限制性的示例，上文已提供了对本申请的一些实施例的详细描述。但结合附图和权利要求来考虑，对以上实施例的多种修改和调整对本领域技术人员来说是显而易见的，但不偏离本申请的范围。因此，本申请的恰当范围将根据权利要求确定。

Claims (14)

  1. 一种联邦学习参与者选择方法，包括：
    获取多个待选择参与者；
    分别确定每个所述待选择参与者的数据质量因子，业务因子和稳定性因子；
    基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者。
  2. 根据权利要求1所述的方法，其中，确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：
    基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子。
  3. 根据权利要求2所述的方法，其中，所述联邦学习类型包括如下一个或多个：KPI劣化检测，小区权重优化，光模块故障预测。
  4. 根据权利要求3所述的方法，其中，在所述联邦学习类型是小区权重优化的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：
    基于所述待选择参与者的QoS告警参数确定数据质量因子；
    基于所述待选择参与者的经纬度确定业务因子；
    基于所述待选择参与者的在服时间确定稳定性因子。
  5. 根据权利要求4所述的方法，其中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，包括：
    将数据质量因子，业务因子和稳定性因子满足最优函数取值的待选择参与者确定为被选中的参与者。
  6. 根据权利要求5所述的方法，其中，所述最优函数为：f(data,value,available)＝
    Figure PCTCN2021117779-appb-100001
    Figure PCTCN2021117779-appb-100002
    Figure PCTCN2021117779-appb-100003
    Figure PCTCN2021117779-appb-100004
    其中，n是指待选择参与者的总个数，i是第i个待选择参与者，data是数据质量因子，value是业务因子，available是稳定性因子，a是第一阈值，由各个待选择参与者的value值确定，b是第二阈值，由各个待选择参与者的data值确定，c是第三阈值，由各个待选择参与者的available值确定。
  7. 根据权利要求3所述的方法，其中，在所述联邦学习类型是KPI劣化检测或光模块故障预测的情况下，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者，包括：
    基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数；
    将所述选取参数进行排序，将排在前L个的所述待选择参与者确定为被选中的参与者。
  8. 根据权利要求7所述的方法，其中，在所述联邦学习类型是KPI劣化检测的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：
    基于所述待选择参与者的性能数据完整性参数确定数据质量因子；
    基于所述待选择参与者的QoS告警参数确定业务因子；
    基于所述待选择参与者的在服时间确定稳定性因子。
  9. 根据权利要求8所述的方法，其中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数，包括：
    针对每个待选择的参与者，对数据质量因子，业务因子和稳定性因子分别采用线性标准化方法进行归一化处理；
    将归一化处理后的数据质量因子，业务因子和稳定性因子进行第一加权处理，得到待选择参与者的选取参数。
  10. 根据权利要求7所述的方法，其中，在所述联邦学习类型是光模块故障预测的情况下，基于联邦学习类型确定所述待选择参与者的数据质量因子，业务因子和稳定性因子，包括：
    基于所述待选择参与者的性能数据完整性参数确定数据质量因子；
    基于所述待选择参与者的光模块链路的质量参数确定业务因子；
    基于所述待选择参与者的在服时间确定稳定性因子。
  11. 根据权利要求10所述的方法，其中，基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定每个待选择参与者的选取参数，包括：
    针对每个待选择的参与者，将数据质量因子，业务因子和稳定性因子进行第二加权处理，得到待选择参与者的选取参数。
  12. 一种联邦学习参与者选择装置，包括：
    获取模块，被配置为获取多个待选择参与者；
    因子确定模块，被配置为分别确定每个所述待选择参与者的数据质量因子，业务因子和稳定性因子；
    参与者选择模块，被配置为基于多个所述待选择参与者的数据质量因子，业务因子和稳定性因子确定被选中的参与者。
  13. 一种联邦学习参与者选择设备，包括：
    一个或多个处理器；
    存储器，用于存储一个或多个程序；其中，
    当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如权利要求1-11中任一项所述的方法。
  14. 一种存储介质，存储有计算机程序，其中，所述计算机程序被处理器执行时实现如权利要求1-11中任一项所述的方法。
PCT/CN2021/117779 2020-11-25 2021-09-10 联邦学习参与者选择方法、装置、设备和存储介质 WO2022110975A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
JP2023532158A JP2023550806A (ja) 2020-11-25 2021-09-10 連合学習参加者の選択方法、装置、機器、及び記憶媒体
EP21896479.9A EP4250188A4 (en) 2020-11-25 2021-09-10 METHOD AND DEVICE FOR SELECTING FEDERATED LEARNING PARTICIPANTS AS WELL AS DEVICE AND STORAGE MEDIUM
US18/254,474 US20230394365A1 (en) 2020-11-25 2021-09-10 Federated learning participant selection method and apparatus, and device and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011345470.3 2020-11-25
CN202011345470.3A CN113255924B (zh) 2020-11-25 2020-11-25 联邦学习参与者选择方法、装置、设备和存储介质

Publications (1)

Publication Number Publication Date
WO2022110975A1 true WO2022110975A1 (zh) 2022-06-02

Family

ID=77180628

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/117779 WO2022110975A1 (zh) 2020-11-25 2021-09-10 联邦学习参与者选择方法、装置、设备和存储介质

Country Status (5)

Country Link
US (1) US20230394365A1 (zh)
EP (1) EP4250188A4 (zh)
JP (1) JP2023550806A (zh)
CN (1) CN113255924B (zh)
WO (1) WO2022110975A1 (zh)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113255924B (zh) * 2020-11-25 2023-03-24 中兴通讯股份有限公司 联邦学习参与者选择方法、装置、设备和存储介质
CN116032917A (zh) * 2021-10-26 2023-04-28 展讯通信（上海）有限公司 联邦学习方法及装置、通信方法及装置
WO2023090472A1 (ko) * 2021-11-17 2023-05-25 엘지전자 주식회사 정보를 전송하는 방법 및 장치, 그리고 정보를 수신하는 방법 및 장치
CN116567702A (zh) * 2022-01-26 2023-08-08 展讯通信（上海）有限公司 一种用户设备选择方法、装置、芯片及模组设备
CN117221930A (zh) * 2022-06-02 2023-12-12 中兴通讯股份有限公司 基站在服率确定方法、装置、电子设备和存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150106166A1 (en) * 2012-08-19 2015-04-16 Carrier Iq, Inc Interactive Selection and Setting Display of Components in Quality of Service (QoS) Scores and QoS Ratings and Method of Operation
CN110598870A (zh) * 2019-09-02 2019-12-20 深圳前海微众银行股份有限公司 一种联邦学习方法及装置
CN111325351A (zh) * 2020-02-19 2020-06-23 深圳前海微众银行股份有限公司 一种确定联邦学习参与方的方法及装置
CN113255924A (zh) * 2020-11-25 2021-08-13 中兴通讯股份有限公司 联邦学习参与者选择方法、装置、设备和存储介质

Family Cites Families (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10368359B2 (en) * 2014-06-04 2019-07-30 Guangdong Oppo Mobile Telecommunications Corp., Ltd. Method and user equipment for predicting available throughput for uplink data
CN105101315B (zh) * 2015-07-08 2019-04-30 浙江大学 远程操控移动通信系统及方法
CN108076468B (zh) * 2016-11-08 2020-12-25 上海中兴软件有限责任公司 一种越区覆盖处理方法及装置
CN109167695B (zh) * 2018-10-26 2021-12-28 深圳前海微众银行股份有限公司 基于联邦学习的联盟网络构建方法、设备及可读存储介质
EP3891933A1 (en) * 2018-12-07 2021-10-13 Telefonaktiebolaget LM Ericsson (publ) Predicting network communication performance using federated learning
JP7244292B2 (ja) * 2019-02-19 2023-03-22 株式会社Ｉｈｉエアロスペース 基地局選択装置と基地局選択方法
CN110475258B (zh) * 2019-07-10 2022-09-23 中国联合网络通信集团有限公司 一种基站的可靠性评估方法和系统
CN110443375B (zh) * 2019-08-16 2021-06-11 深圳前海微众银行股份有限公司 一种联邦学习方法及装置
US11423332B2 (en) * 2019-09-27 2022-08-23 Intel Corporation Distributed machine learning in an information centric network
CN111369042B (zh) * 2020-02-27 2021-09-24 山东大学 一种基于加权联邦学习的无线业务流量预测方法
CN111611610B (zh) * 2020-04-12 2023-05-30 西安电子科技大学 联邦学习信息处理方法、系统、存储介质、程序、终端
CN111756848B (zh) * 2020-06-28 2021-05-11 河海大学 移动边缘环境下基于联邦学习和移动感知的QoS优化方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150106166A1 (en) * 2012-08-19 2015-04-16 Carrier Iq, Inc Interactive Selection and Setting Display of Components in Quality of Service (QoS) Scores and QoS Ratings and Method of Operation
CN110598870A (zh) * 2019-09-02 2019-12-20 深圳前海微众银行股份有限公司 一种联邦学习方法及装置
CN111325351A (zh) * 2020-02-19 2020-06-23 深圳前海微众银行股份有限公司 一种确定联邦学习参与方的方法及装置
CN113255924A (zh) * 2020-11-25 2021-08-13 中兴通讯股份有限公司 联邦学习参与者选择方法、装置、设备和存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4250188A4 *

Also Published As

Publication number Publication date
CN113255924B (zh) 2023-03-24
JP2023550806A (ja) 2023-12-05
CN113255924A (zh) 2021-08-13
EP4250188A1 (en) 2023-09-27
US20230394365A1 (en) 2023-12-07
EP4250188A4 (en) 2024-04-24

Similar Documents

Publication Publication Date Title
WO2022110975A1 (zh) 联邦学习参与者选择方法、装置、设备和存储介质
JP2022016589A (ja) コンピューティング・デバイスについてのシステム性能およびイベント・データを収集、追跡、および記憶するためのシステムおよび方法
CN112840600B (zh) 用于改进无线通信网络中的性能的方法、装置和介质
US20150135320A1 (en) Methods and apparatus to identify malicious activity in a network
CN104396188A (zh) 用于对移动网络性能问题进行根本原因分析的系统和方法
US10728773B2 (en) Automated intelligent self-organizing network for optimizing network performance
US20180225592A1 (en) Determining a target device profile including an expected behavior for a target device
CN111935820B (zh) 基于无线网络的定位实现方法及相关设备
US10742482B2 (en) Clustering event records representing anomalous events
US11329893B2 (en) Live network real time intelligent analysis on distributed system
US20230146912A1 (en) Method, Apparatus, and Computing Device for Constructing Prediction Model, and Storage Medium
CN104285159A (zh) 支持存储信息的更新
US20200413459A1 (en) Overshoot analysis based on user equipment metrics
WO2015062109A1 (zh) 一种网络关键性绩效指标的评估方法及装置
US20230004776A1 (en) Moderator for identifying deficient nodes in federated learning
US20210127303A1 (en) Wireless band priority metrics analysis and response
WO2017108106A1 (en) Method and network node for identifiying specific area of wireless communication system
CN109993562B (zh) 一种满意度仿真方法、装置及终端设备
US9609533B2 (en) Method and device for optimizing a handover parameter
US11350290B2 (en) Downlink interference detection and identification of aggressor cells
CN112437469B (zh) 服务质量保障方法、装置和计算机可读存储介质
CN106028399B (zh) 针对功率和性能优化设备中的应用行为
Liu et al. FaLQE: Fluctuation adaptive link quality estimator for wireless sensor networks
FR3045859A1 (fr) Procede et dispositifs associes de formation d'un nuage informatique stockant le resultat de l'execution deportee d'une tache informatique
CN112637904B (zh) 负载均衡方法、装置及计算设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21896479

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 2023532158

Country of ref document: JP

WWE Wipo information: entry into national phase

Ref document number: 2021896479

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2021896479

Country of ref document: EP

Effective date: 20230622

NENP Non-entry into the national phase

Ref country code: DE