Method and apparatus for training artificial intelligence AI models in wireless networks

Abstract

A method and apparatus for training an artificial intelligence AI model in a wireless network are provided. The method includes: sending first setting information to a terminal participating in federated learning, the first setting information being used to set at least one of the following training period, time-frequency resource, and report timing, the same training period, the same time-frequency resource, and the same report timing being set for different terminals participating in federated learning; and receiving a signal obtained by wireless superposition of gradients reported by the terminal, the gradient being completed within the training period and being the gradient of the AI ​​model reported by the terminal at the report timing by using the time-frequency resource. According to the method and apparatus of the present application, problems such as excessively high time-frequency resource utilization and large delay caused by edge nodes participating in federated learning need to use mutually orthogonal uplink time-frequency resources can be eliminated.

Description

[Technical field]

[0001] TECHNICAL FIELD Embodiments of the present application relate to the field of artificial intelligence (AI), and in particular to methods for training AI models in wireless networks. [Background technology]

[0002] In wireless communication networks, e.g., in mobile communication networks, the services supported by the network are becoming more and more diverse, so are the requirements to be satisfied. For example, the network must be capable of supporting very high speed, very low latency, and / or very large capacity connections. Such characteristics make network planning, network configuration, and / or resource scheduling more and more complex. Such new requirements, scenarios, and characteristics bring unprecedented challenges to network planning, operation, and maintenance, and efficient operation. To satisfy these challenges, artificial intelligence techniques may be introduced into wireless communication networks to achieve network intelligence. Federated learning is a popular model training architecture in which distributed data located at edge devices can be invoked to participate in model training without violating privacy. In federated learning, the central node and edge devices need to exchange parameters or gradients of an artificial intelligence (AI) model. How to deploy federated learning in wireless networks is a problem worth studying. Summary of the Invention

[0003] The present application provides a method and apparatus for training an AI model in a wireless network, so as to solve problems such as excessively high time-frequency resource utilization and large delay caused by edge nodes participating in federated learning need to use mutually orthogonal uplink time-frequency resources when federated learning is deployed in a wireless network.

[0004] According to a first aspect, a method for training an artificial intelligence AI model in a wireless network is provided. The method is executed by a second node, and may be executed by a component (processor, chip, etc.) configured in the second node, or may be executed by a software module. The method includes: transmitting first setting information to a terminal participating in federated learning, the first setting information is used to set at least one of the following training period, time-frequency resource, and report timing, the same training period, the same time-frequency resource, and the same report timing are set for different terminals participating in the federated learning; and receiving a signal obtained by wireless superposition of gradients reported by the terminals participating in the federated learning, the gradients being gradients reported by the terminals at the report timing by using the time-frequency resource of an AI model whose training is completed within the training period.

[0005] It should be noted that in the description of this application, the second node may also be referred to as an access network device, the first node may also be referred to as a central node, etc. In the above design, the same training period, the same time-frequency resource, and the same reporting timing are set by the access network device (which may also be referred to as the second node) or the central node (which may also be referred to as the first node) for the terminals participating in the federated learning. For example, the number of terminals participating in the federated learning is n. In this application, the access network device or the central node assigns the same time-frequency resource to n terminals. Compared with the conventional solution in which n orthogonal time-frequency resources are assigned to n terminals participating in the federated learning, the overhead of time-frequency resources can be reduced. Furthermore, when n orthogonal time-frequency resources are assigned to n terminals participating in the federated learning, each terminal reports the gradient of the AI ​​model by using its respective time-frequency resource. The access network device may receive n radio frequency signals and process the n radio frequency signals separately to restore the gradient reported by each terminal. As a result, the delay is large. However, in this application, the access network device allocates one time-frequency resource to n terminals participating in federated learning, and the n terminals report gradients on the time-frequency resource. Therefore, based on the superposition nature of the wireless channel, the n gradients are superposed during wireless transmission. For example, the value of n is 3, the gradient reported by terminal 1 is Y1, the gradient reported by terminal 2 is Y2, and the gradient reported by terminal 3 is Y3. In this case, the above three gradients are transmitted on the same time-frequency resource, and the above three gradients are superposed. It is assumed that the wireless channel satisfies perfect signal superposition (fading, interference, noise, etc. can be ignored). In this case, the superposed signal is Y=Y1+Y2+Y3.When the channel does not satisfy perfect signal superposition, after receiving the above signals in the above time-frequency resources, the access network device may perform signal processing on the received signals to restore the superposed signal Y, and the terminal may perform gradient aggregation by using the superposed signal Y. The process of gradient aggregation may be a process of calculating an arithmetic mean. For example, the superposed signal Y may be divided by 3, and the result is used as the aggregate gradient. According to the solution of the present application, the superposed signal Y may be obtained by processing one radio frequency signal. However, in the existing solution, three radio frequency signals in different slots need to be processed sequentially to restore the corresponding gradients, and then aggregation is further performed. By using the solution of the present application, the utilization of time-frequency resources can be reduced to a certain extent, and the delay of gradient aggregation can be reduced.

[0006] It should be understood that the training period set by the second node (or called access network device) for the terminal may not be the actual training period of the terminal, but is an upper limit of the time required by the terminal for each training round. In other words, the terminal completes the current round of model training within the training period and reports a training completion indication to the base station. Otherwise, the terminal may end the current round of model training and wait until the next training period comes. In this application, the training period is set to ensure that different terminals can simultaneously report the gradient of model training to the access network device. Optionally, the gradient of model training reported by the terminal to the access network device is the gradient in the current round of model training, not the gradient in other rounds of model training, e.g., the gradient in the previous round of model training. For example, the training period may be a global parameter determined by comprehensively considering the computational capabilities of the terminals participating in the federated learning, the complexity of the model, and the like.

[0007] In a possible design, the method further includes receiving a training completion indication from the terminal, the training completion indication being sent by the terminal to the second node if training of the AI ​​model is completed within the training period, and collecting statistics regarding the number of terminals that completed training of the AI ​​model within the training period based on the training completion indication sent by the terminal.

[0008] In a possible design, the method further includes determining an average gradient in the current round of model training based on gradients reported by different terminals participating in the federated learning when the number of terminals that have completed training of the AI ​​model is greater than or equal to a terminal threshold number, or using the average gradient in the previous round of model training as the average gradient in the current round of model training when the number of terminals that have completed training of the AI ​​model is less than the terminal threshold number, updating parameters of the AI ​​model based on the average gradient in the current round of model training, and sending the average gradient in the current round of model training to the terminal.

[0009] According to the above design, the first node may determine a terminal threshold number. When air-based calculation is introduced into federated learning, the number of terminals participating in federated learning affects the accuracy of calculating the average gradient in the current round of model training. The first node may set a terminal threshold number. If the number of terminals reporting gradients is equal to or greater than the terminal threshold number, the average gradient in the current round of model training is calculated and sent to the terminal; otherwise, the average gradient in the previous round of model training is sent to the terminal, or the average gradient in the previous round of model training is used as the average gradient in the current round of model training and sent to the terminal, ensuring that the accuracy of the calculated average gradient in the current round of model training meets the requirement.

[0010] In a possible design, the method further includes transmitting to the first node the number of terminals that completed training of the AI ​​model within the training period and a signal obtained by wireless superposition of the gradients reported by the terminals.

[0011] In a possible design, the first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, modulation scheme, initial AI model, or transmit power.

[0012] In a possible design, the process of determining the transmit power includes measuring a sounding reference signal SRS from the terminal to determine an uplink channel quality of the terminal, and determining the transmit power of the terminal based on the uplink channel quality.

[0013] According to the above design, the second node determines an optimal transmission power by measuring the uplink channel, and sets the transmission power for the terminal to transmit the gradients in the current round of model training, so as to improve the accuracy of the airborne calculation and further enhance the accuracy of the gradient aggregation.

[0014] In a possible design, the first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, a modulation scheme, an initial AI model, a channel state information CSI interval, or a channel inversion parameter.

[0015] In a possible design, the method further includes receiving second configuration information from the first node, where the second configuration information is used to configure at least one of the following: a list of terminals to participate in the federated learning, an initial AI model, a group temporary identifier, a training period, a terminal threshold number, a transport block size, or an uplink requirement.

[0016] In a possible design, the method further includes receiving first terminal information from the terminal and sending second terminal information to the first node, where the first terminal information includes at least one of the following: a communication capability of the terminal, a computational capability of the terminal, or a dataset characteristic of the terminal, and the second terminal information includes at least one of the following: a communication capability of the terminal, a computational capability of the terminal, a dataset characteristic of the terminal, or a terminal temporary identifier, and the terminal temporary identifier is assigned to the terminal by the second node.

[0017] In this design, the communication capabilities of the terminal include, for example, the maximum transmission power that the terminal can support, the antenna configuration of the terminal, etc. The computational capabilities of the terminal include, for example, the performance of a central processing unit (CPU), the performance of a graphics processing unit (GPU), storage space, and power. The dataset characteristics of the terminal include, for example, the size of the dataset, the distribution of the dataset, whether the dataset is complete, and whether the label of the dataset is complete. Optionally, the dataset may be further divided into a training set, a validation set, and a test set based on a percentage. For example, 60% of the dataset is a training set, 20% of the dataset is a validation set, and 20% of the dataset is a test set. It can be understood that the training set is used to train an AI model, the validation set is used to evaluate the trained AI model, and the test set is used to test the trained AI model. The terminal temporary identifier, i.e. the identifier, may be a cell radio network temporary identifier (C-RNTI), another temporary identifier, etc.

[0018] In a possible design, the method further includes sending a model training termination indication to the terminal when the model training termination condition is satisfied, or receiving a model training termination indication from the first node and sending a model training termination indication to the terminal.

[0019] According to a second aspect, a method for training an artificial intelligence AI model in a wireless network is provided. The method corresponds to the first aspect. For advantageous effects, please refer to the description of the first aspect. The method is executed by a terminal, and may be executed by a component (processor, chip, etc.) configured in the terminal, or may be executed by a software module, etc. The method includes: receiving first setting information from a second node, the first setting information is used to set at least one of a next training period, a time-frequency resource, and a reporting timing, and the same training period, the same time-frequency resource, and the same reporting timing are set for different terminals participating in the federated learning; training the AI ​​model within the training period to obtain a gradient of the AI ​​model in a current round of model training; and reporting the gradient of the AI ​​model in the current round of model training to the second node at the reporting timing by using the time-frequency resource.

[0020] In a possible design, the method further includes, when the training period ends, sending a training completion indication to the second node if training of the AI ​​model is complete.

[0021] In a possible design, the method further includes terminating training of the AI ​​model if training of the AI ​​model is not completed within the training period.

[0022] In a possible design, the method further includes receiving an average gradient in a previous round of model training from the second node, and updating gradients of the AI ​​model in the current round of model training based on the average gradients in the previous round of model training, or updating gradients and parameters of the AI ​​model in the current round of model training based on the average gradients in the current round of model training.

[0023] In a possible design, the first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, modulation scheme, initial AI model, or transmit power.

[0024] In a possible design, the first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, a modulation scheme, an initial AI model, a channel state information CSI interval, or a channel inversion parameter.

[0025] In a possible design, when the first configuration information is further used to set channel state information CSI interval and channel inversion parameters, the method further includes: if the same frequency resource is configured for the downlink channel and the uplink channel, determining a CSI of an uplink channel of the terminal based on the measured CSI of the downlink channel, and if the CSI of the uplink channel meets the CSI interval requirements, determining a transmission power based on the channel inversion parameters, and reporting the gradient of the AI ​​model in the current round of model training to the second node includes reporting the gradient of the AI ​​model in the current round of model training to the second node based on the determined transmission power.

[0026] In a possible design, the first configuration information further includes a group temporary identifier, where the group temporary identifier is a group temporary identifier assigned to the terminal by the first node.

[0027] In a possible design, the method further includes receiving a scheduling instruction from the second node, the scheduling instruction including a group temporary identifier, and performing training of the AI ​​model in the current round of model training when the group temporary identifier included in the scheduling instruction is the same as the group temporary identifier assigned to the terminal by the first node, or skipping performance of the training of the AI ​​model in the current round of model training otherwise.

[0028] In a possible design, the method further includes receiving a model training termination indication from the second node and terminating training of the AI ​​model based on the model training termination indication.

[0029] In a possible design, the method further includes transmitting first terminal information to the second node, where the first terminal information includes at least one of the following: communication capabilities of the terminal, computational capabilities of the terminal, or dataset characteristics of the terminal.

[0030] According to a third aspect, a method for training an artificial intelligence AI model in a wireless network is provided. The method corresponds to the first aspect. For advantageous effects, please refer to the description of the first aspect. The method may be executed by a first node, and may be executed by a component (processor, chip, or other component) configured in the first node, or may be executed by a software module, etc. The method includes: determining second setting information, the second setting information being used to set at least one of a list of terminals participating in the next federated learning, an initial AI model, a group temporary identifier, a training period, a terminal threshold number, a size of a transport block, or an uplink requirement; and transmitting the second setting information to the second node.

[0031] In a possible design, the method further includes receiving second terminal information from the second node, the second terminal information including at least one of the following: communication capabilities of the terminal, computational capabilities of the terminal, dataset characteristics of the terminal, or a terminal temporary identifier, the terminal temporary identifier being assigned to the terminal by the second node, and determining a list of terminals to participate in the federated learning based on the terminal information.

[0032] In a possible design, the method further includes receiving from the second node a signal obtained by wireless superposition of gradients reported by terminals participating in the federated learning and the number of terminals that have completed training of the AI ​​model within the training period; if the number of terminals that have completed training of the model within the training period is greater than or equal to a terminal threshold number, determining an average gradient in the current round of model training based on the gradient of the AI ​​model reported by the terminal, or if not, using the average gradient in the previous round of model training as the average gradient in the current round of model training; updating parameters of the AI ​​model based on the average gradient in the current round of model training, and transmitting the average gradient in the current round of model training to the second node, so as to enable the second node to transmit the average gradient in the current round of model training to the terminal.

[0033] In a possible design, the method further includes sending a scheduling instruction to the second node, the scheduling instruction including the group temporary identifier, and the scheduling instruction is used to schedule a terminal corresponding to the group temporary identifier to perform training of the AI ​​model in the current round of model training.

[0034] In a possible design, the method further includes, when the model training termination condition is satisfied, sending a model training termination instruction to the second node to instruct the terminal to terminate training of the AI ​​model in the current round of model training.

[0035] According to a fourth aspect, there is provided an apparatus. The apparatus includes units or modules in one-to-one correspondence with the methods / operations / steps / actions described in the first, second or third aspect. The units or modules may be hardware circuits, or software circuits, or may be implemented by hardware circuits in combination with software.

[0036] According to a fifth aspect, there is provided a communication device. The device includes a processor and a memory. The memory is configured to store a computer program or instruction, and the processor is coupled to the memory. When the processor executes the computer program or instruction, the device can perform the method of the first aspect, the second aspect, or the third aspect.

[0037] According to a sixth aspect, there is provided an apparatus, comprising a processor and an interface circuit, the processor configured to communicate with other devices through the interface circuit and to perform a method as described in any one of the first, second or third aspects.

[0038] According to a seventh aspect, an apparatus is provided, comprising a processor coupled to a memory. The processor is configured to execute a program stored in the memory to perform a method as described in any one of the first, second or third aspects. The memory may be located external or internal to the apparatus. Further, there may be one or more processors.

[0039] According to an eighth aspect, a chip system is provided, comprising a processor or circuitry configured to perform the method described in any one of the first, second or third aspects.

[0040] According to a ninth aspect, there is provided a computer readable storage medium. The computer readable storage medium stores a computer program or instructions. When the computer program or instructions are executed by an apparatus, the apparatus can perform the method in the first aspect, the second aspect, or the third aspect.

[0041] According to a tenth aspect, there is provided a computer program product comprising a computer program or instructions, which when executed by an apparatus, can cause the apparatus to perform the method of the first aspect, the second aspect or the third aspect.

[0042] According to an eleventh aspect, there is provided a system, comprising an apparatus for implementing the method of the first aspect and an apparatus for performing the method of the second aspect. Optionally, the apparatus may further comprise an apparatus for performing the method of the third aspect. [Brief description of the drawings]

[0043] [Figure 1] FIG. 1 is a diagram of a network architecture according to the present application. [Figure 2a] FIG. 2 is a diagram of a neural network according to the present application. [Figure 2b] FIG. 2 is a diagram of a neural network according to the present application. [Figure 2c] FIG. 1 is a diagram of an AI model in accordance with the present application. [Diagram 3] FIG. 2 is a diagram of associative learning training in accordance with the present application. [Figure 4] 1 is a schematic flow chart according to the present application. [Figure 5A] 1 is a schematic flow chart according to the present application. [Figure 5B] 1 is a schematic flow chart according to the present application. [Figure 6] FIG. 2 is a diagram of channel measurements according to the present application. [Figure 7A] 4 is another schematic flow chart according to the present application. [Figure 7B] 4 is another schematic flow chart according to the present application. [Figure 7C] 4 is another schematic flow chart according to the present application. [Figure 8A] 4 is another schematic flow chart according to the present application. [Figure 8B] 4 is another schematic flow chart according to the present application. [Figure 8C] 4 is another schematic flow chart according to the present application. [Figure 8D] 4 is another schematic flow chart according to the present application. [Figure 9] FIG. 1 is a diagram of an apparatus according to the present application. [Figure 10] FIG. 1 is a diagram of an apparatus according to the present application. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0044] FIG. 1 is an architecture diagram of a communication system 1000 to which the present application is applied. As shown in FIG. 1, the communication system includes a radio access network 100 and a core network 200. Optionally, the communication system 1000 may further include the Internet 300. The radio access network 100 may include at least one access network device (e.g., 110a and 110b in FIG. 1), and may further include at least one terminal (e.g., 120a-120j in FIG. 1). The terminal is connected to the access network device in a wireless manner, and the access network device is connected to the core network in a wireless or wired manner. The core network device and the access network device may be different physical devices independent of each other, or the functions of the core network device and the logical functions of the access network device may be integrated into the same physical device, or some of the functions of the core network device and some of the functions of the access network device may be integrated into one physical device. The terminals may be connected to each other in a wired or wireless manner, and the access network devices may be connected to each other in a wired or wireless manner. FIG. 1 is merely a diagram. The communication system may further include other network devices, such as wireless relay devices and wireless backhaul devices, which are not shown in FIG.

[0045] The access network device may be a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next generation NodeB (gNB) in a 5th generation (5G) mobile communication system, an access network device in an open radio access network (O-RAN), a next generation base station in a 6th generation (6G) mobile communication system, a base station in a future mobile communication system, an access node in a wireless fidelity (Wi-Fi) system, etc., or may be a module or unit, such as a central unit (CU), a distributed unit (DU), a central unit control plane (CU-CP) module, or a central unit user plane (CU-UP) module that realizes part of the functions of a base station. The access network device may be a macro base station (e.g., 110a in FIG. 1), or may be a micro base station or an indoor base station (e.g., 110b in FIG. 1), or may be a relay node, a donor node, etc. The specific technology used by the access network device and the specific device form are not limited by this application.

[0046] In the present application, an apparatus configured to implement a function of an access network device may be an access network device, or may be, for example, an apparatus that can support an access network device to implement a function, such as a chip system, a hardware circuit, a software module, or a combination of a hardware circuit and a software module, where the apparatus may be installed in the access network device or adapted for use with the access network device. In the present application, a chip system may include a chip, or may include a chip and other discrete components. For ease of description, the following describes the technical solution provided in the present application by using an example in which the apparatus configured to implement a function of an access network device is an access network device, and the access network device is a base station.

[0047] (1) Protocol layer structure

[0048] The communication between the access network device and the terminal follows a specific protocol layer structure. The protocol layer structure may include a control plane protocol layer structure and a user plane protocol layer structure. For example, the control plane protocol layer structure may include protocol layer functions such as a radio resource control (RRC) layer, a packet data convergence protocol (PDCP) layer, a radio link control (RLC) layer, a media access control (MAC) layer, and a physical layer. For example, the user plane protocol layer structure may include protocol layer functions such as a PDCP layer, an RLC layer, a MAC layer, and a physical layer. In a possible implementation, a service layer adaptation protocol (SDAP) layer may be further included above the PDCP layer.

[0049] Optionally, the protocol layer structure between the access network device and the terminal may further include an artificial intelligence (AI) layer configured to transmit data related to AI functions.

[0050] (2) Central unit (CU) and distributed unit (DU)

[0051] An access device may include a CU and a DU. A plurality of DUs may be controlled by one CU in a centralized manner. For example, an interface between a CU and a DU may be called an F1 interface. A control plane (CP) interface may be F1-C, and a user plane (UP) interface may be F1-U. The specific names of each interface are not limited in this application. The CU and the DU may be defined based on the protocol layers of a wireless network. For example, the functions of the PDCP layer and the protocol layers above the PDCP layer are set in the CU, and the functions of the protocol layers below the PDCP layer (e.g., the RLC layer and the MAC layer) are set in the DU. As another example, the functions of the protocol layers above the PDCP layer are set in the CU, and the functions of the PDCP layer and the protocol layers below the PDCP layer are set in the DU. This is not limited.

[0052] The above division into the processing functions of the CU and the DU based on the protocol layer is only one example, and other divisions may alternatively be used. For example, the CU or the DU may be defined to have more protocol layer functions, and as another example, the CU or the DU may alternatively be defined to have a part of the processing functions of the protocol layers. In a design, a part of the functions of the RLC layer and the functions of the protocol layers above the RLC layer are set in the CU, and the remaining functions of the RLC layer and the functions of the protocol layers below the RLC layer are set in the DU. In other designs, the division into the functions of the CU or the DU may alternatively be performed based on the service type or other system requirements. For example, the division may be performed based on delay. Functions whose processing time needs to satisfy the delay requirement are set in the DU, and functions whose processing time does not need to satisfy the delay requirement are set in the CU. In other designs, the CU may alternatively have one or more functions of the core network. For example, the CU may be deployed on the network side to facilitate centralized management. In other designs, the radio unit (RU) of the DU is deployed remotely. Optionally, the RU may have radio frequency functions.

[0053] Optionally, the DU and the RU may be distinguished by a physical layer (PHY). For example, the DU may implement the upper layer functions of the PHY layer, and the RU may implement the lower layer functions of the PHY layer. When the PHY layer is used for transmission, the functions of the PHY layer may include at least one of the following functions: adding a cyclic redundancy check (CRC) code, channel coding, rate matching, scrambling, modulation, layer mapping, precoding, resource mapping, physical antenna mapping, or radio frequency transmission. When the PHY layer is used for reception, the functions of the PHY layer may include at least one of the following functions: CRC check, channel decoding, de-rate matching, descrambling, demodulation, layer demapping, channel detection, resource demapping, physical antenna demapping, or radio frequency reception. The upper layer functions of the PHY layer may include some of the functions of the PHY layer. For example, some of the functions are closer to the MAC layer. The lower layer functions of the PHY layer may include other parts of the functions of the PHY layer. For example, the parts of the functions are closer to the radio frequency functions. For example, the upper layer functions of the PHY layer may include CRC code addition, channel coding, rate matching, scrambling, modulation, and layer mapping, and the lower layer functions of the PHY layer may include precoding, resource mapping, physical antenna demapping, and radio frequency transmission functions. Alternatively, the upper layer functions of the PHY layer may include CRC code addition, channel coding, rate matching, scrambling, modulation, layer mapping, and precoding, and the lower layer functions of the PHY layer may include resource mapping, physical antenna mapping, and radio frequency transmission functions. For example, the upper layer functions of the PHY layer may include CRC check, channel decoding, de-rate matching, decoding, modulation, and layer demapping, and the lower layer functions of the PHY layer may include channel detection, resource demapping, physical antenna demapping, and radio frequency reception functions.Alternatively, the upper layer functions of the PHY layer may include CRC checking, channel decoding, de-rate matching, decoding, demodulation, layer demapping, and channel detection, and the lower layer functions of the PHY layer may include resource demapping, physical antenna demapping, and radio frequency reception functions.

[0054] For example, the functionality of the CU may be implemented by one entity or may be implemented by different entities. For example, the functionality of the CU may be further divided. Specifically, the control plane and user plane of the CU are separated and implemented by different entities, the different entities being a control plane CU entity (i.e., a CU-CP entity) and a user plane CU entity (i.e., a CU-UP entity). The CU-CP entity and the CU-UP entity may be coupled to the DU to jointly realize the functionality of an access network device.

[0055] Optionally, any one of the DU, CU, CU-CP, CU-UP, and RU may be a software module, a hardware structure, or a combination of a software module and a hardware structure. This is not limited. Different entities may exist in different forms, this is not limited. For example, the DU, CU, CU-CP, and CU-UP are software modules, and the RU is a hardware structure. These modules and the methods performed by these modules also fall within the scope of protection of the present disclosure.

[0056] In a possible implementation, the access network device includes a CU-CP, a CU-UP, a DU, and a RU. For example, the present application is performed by a DU, or a DU and a RU, or a CU-CP, a DU and a RU, or a CU-UP, a DU and a RU. This is not limited. The method performed by the modules also falls within the scope of protection of the present application.

[0057] The terminal may also be referred to as a terminal device, user equipment (UE), mobile station, mobile terminal, etc. The terminal is widely applied to communication in various scenarios, including, but not limited to, at least one of the following scenarios: device-to-device (D2D), vehicle-to-everything (V2X), machine-type communication (MTC), Internet of things (IOT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grid, smart furniture, smart office, smart wearable, smart transportation, smart city, etc. The terminal may be a mobile phone, a tablet computer, a computer with a wireless transceiver function, a wearable device, a vehicle, an unmanned aerial vehicle, a helicopter, an airplane, a ship, a robot, a mechanical arm, a smart home device, etc. The specific technology used by the terminal and the specific device form are not limited by this application.

[0058] In the present application, a device configured to implement the functions of a terminal may be a terminal, or may be, for example, a device that can support a terminal to implement functions, such as a chip system, a hardware circuit, a software module, or a combination of a hardware circuit and a software module, where the device may be installed in a terminal or adapted for use with a terminal. For ease of description, the following describes the technical solution provided in the present application by using an example in which the device configured to implement the functions of a terminal is a terminal.

[0059] The base station and the terminal may be fixed or mobile. The base station and / or the terminal may be deployed on the ground, including indoor or outdoor scenarios, and handheld or vehicle-mounted scenarios, or on water, or in the air, on airplanes, balloons, and satellites. The application scenarios of the base station and the terminal are not limited in this application. The base station and the terminal may be deployed in the same scenario or in different scenarios. For example, the base station and the terminal are both deployed on the ground. Alternatively, the base station is deployed on the ground and the terminal is deployed on the water. The examples are not described one by one.

[0060] The roles of the base station and the terminal may be relative. For example, the helicopter or unmanned aerial vehicle 120i in FIG. 1 may be configured as a mobile base station. In the case of the terminal 120j accessing the wireless access network 100 via 120i, the terminal 120i is a base station, while in the case of the base station 110a, 120i is a terminal, in other words, 110a and 120i communicate with each other based on a wireless air interface protocol. 110a and 120i may alternatively communicate with each other based on an interface protocol between base stations. In this case, 120i is also a base station for 110a. Thus, the base station and the terminal may be collectively referred to as a communication device, and 110a and 110b in FIG. 1 may be referred to as a communication device having a function of a base station, and 120a to 120j in FIG. 1 may be referred to as a communication device having a function of a terminal.

[0061] Communications between base stations and terminals, between base stations and base stations, or between terminals and terminals may be performed using licensed spectrum, or may be performed using unlicensed spectrum, or may be performed using both licensed and unlicensed spectrum. Communications may be performed using spectrum below 6 gigahertz (GHz), or may be performed using spectrum above 6 GHz, or may be performed using both spectrum below 6 GHz and spectrum above 6 GHz. Spectral resources used for wireless communications are not limited in this application.

[0062] In this application, a base station transmits a downlink signal or information to a terminal, the downlink information being carried on a downlink channel, and the terminal transmits an uplink signal or information to the base station, the uplink information being carried on an uplink channel. To communicate with a base station, a terminal may establish a wireless connection to a cell controlled by the base station. The cell to which the terminal establishes a wireless connection is called the serving cell of the terminal. When communicating with a serving cell, the terminal may be interfered with by signals from nearby cells.

[0063] In the present application, an independent network element (e.g., referred to as a central node, AI network element, or AI node) may be introduced into the communication system shown in FIG. 1 to implement AI-related operations. The central node may be directly connected to the access network device in the communication system, or may be indirectly connected to the access network device via a third party network element. The third party network element may be a core network element, such as an authentication management function (AMF) network element or a user plane function (UPF) network element. Alternatively, the AI ​​function, AI module, or AI entity may be configured in another network element in the communication system to implement AI-related operations. For example, the other network element may be an access network device (e.g., gNB), a core network device, or a network management (OAM). In this case, the network element that performs AI-related operations is a network element with built-in AI functions. In the present application, the OAM is configured to operate, manage, and / or maintain the core network device and / or is configured to operate, manage, and / or maintain the access network device.

[0064] In this application, an AI model is a specific way to implement an AI function, and an AI model represents a mapping relationship between the input and output of the model. An AI model may be a neural network or other machine learning model. An AI model may be referred to as a model for short. An AI-related operation may include at least one of the following: data collection, model training, model information release, model inference, inference result release, etc.

[0065] A neural network is used as an example. A neural network is a specific embodiment of machine learning technology. According to the universal approximation theorem, a neural network can theoretically approximate any continuous function, so that the neural network has the ability to learn any mapping. In traditional communication systems, communication modules need to be designed with extensive expertise. However, a deep learning communication system based on a neural network can automatically discover implicit pattern structures from a large amount of data set, establish mapping relationships between data, and obtain better performance than traditional modeling methods.

[0066] The idea of ​​neural networks comes from the neuron structure of the brain tissue. Each neuron performs a weighted sum operation on the neuron's input values ​​and outputs the weighted sum result through an activation function. Figure 2a shows the structure of a neuron. The inputs of a neuron are x = [x0, x1, . . . , x n ], and the weights corresponding to the inputs are w=[w1,w2, ,w n ] and the bias of the weighted sum is assumed to be b. The form of the activation function can be varied. The activation function of a neuron is assumed to be y = f(z) = max(0,z). In this case, the output of the neuron is

number

number

[0067] A neural network usually includes a multi-layer structure, with each layer including one or more neurons. Increasing the depth and / or width of a neural network can improve the representation capability of the neural network, providing a stronger information extraction and abstract modeling capability for complex systems. The depth of a neural network may refer to the number of layers included in the neural network, and the number of neurons included in each layer may be referred to as the layer width. FIG. 2b is a diagram of the layer relationship of a neural network. In an implementation, a neural network includes an input layer and an output layer. After performing neuron processing on the received input, the input layer of the neural network transfers the result to the output layer, and the output layer obtains the output result of the neural network. In another implementation, a neural network includes an input layer, a hidden layer, and an output layer. After performing neuron processing on the received input, the input layer of the neural network transfers the calculation result to the output layer or an adjacent hidden layer, and finally, the output layer obtains the output result of the neural network. A neural network may include one hidden layer or multiple hidden layers connected in series. This is not limited. In the training process of a neural network, a loss function may be defined. The loss function represents the gap or difference between the output value of the neural network and the ideal target value. The specific form of the loss function is not limited in this application. The training process of a neural network is the process of adjusting neural network parameters such as the gradient of the neural network, the number and width of layers, the weights of neurons, parameters in the activation functions of neurons, and / or the like, so that the value of the loss function is less than a threshold or meets the target requirement.

[0068] FIG. 2c is a diagram of an AI application framework. A data source is configured to store training data and inference data. A model training host analyzes or trains the training data provided by the data to obtain an AI model, and deploys the AI ​​model to a model inference host. Optionally, the model training node may further update the AI ​​model deployed in the model inference node. The model inference node may further feed back related information of the deployed model to the model training node, so that the model training node optimizes or updates the deployed AI model, etc.

[0069] The AI ​​model represents a mapping relationship between the input and output of the model. Obtaining an AI model by learning through a model training node is equivalent to obtaining a mapping relationship between the input and output of the model by learning through a model training node using training data. The model inference node uses the AI ​​model to perform inference based on inference data provided by a data source, and obtains an inference result. The method can also be described as follows: the model inference node inputs inference data to the AI ​​model and obtains an output by using the AI ​​model. The output is an inference result. The inference result may indicate a configuration parameter used (operated) by the subject of the action and / or an operation performed by the subject of the action. The inference result may be centrally planned by an actor entity and sent to one or more subjects of the action (e.g., network entities) for action.

[0070] Federated learning (FL) is a popular AI / ML model training framework that can effectively help multiple organizations use data to perform machine learning modeling while meeting the requirements of user privacy protection, data safety, and government regulations. As a distributed machine learning model, federated learning effectively solves the problem of data silos and enables participants to perform collaborative modeling without sharing data, thereby technically breaking data silos and realizing AI collaboration. Federated learning includes a central node and an edge node. A central node (e.g., a server or a base station) can call distributed devices located at edge devices (e.g., smartphones or sensors) to participate in model training without violating privacy.

[0071] Federated learning includes three types: horizontal federated learning, vertical federated learning, and federated transfer learning. This application mainly relates to the process of horizontal federated learning. Figure 3 shows the training process of horizontal federated learning. It can be seen that horizontal federated learning includes one central node and multiple edge nodes. The original data is distributed to the edge nodes, and the central node does not have the original data, and the edge nodes do not allow the original data to be sent to the central node.

[0072] In the training process of federated learning, the central node first sends an initialized AI model (which can be called an initial AI model) to each edge node, and then starts iterative training. Each iterative training process is as follows:

[0073] 1. The edge node trains an initial AI model by using local data and obtains the gradients of the trained AI model.

[0074] 2. Each edge node reports the gradients obtained by the edge node through training to the central node.

[0075] 3. After receiving the gradients reported by the edge nodes, the central node aggregates the gradients and updates the parameters of the AI ​​model based on the aggregated gradients.

[0076] 4. The central node distributes the aggregated gradients to each edge node participating in the training, and the edge nodes update the parameters and gradients of the locally trained AI model based on the aggregated gradients distributed by the central node.

[0077] 5. The central node calculates the loss function of the AI ​​model obtained by updating the parameters, and terminates model training if the loss function meets the condition, or repeats steps 2 to 4 above if the loss function does not meet the condition.

[0078] In the conventional federated learning solution, mutually orthogonal uplink time-frequency resources need to be assigned to edge nodes, and the edge nodes transmit gradients on the mutually orthogonal uplink time-frequency resources. The central node needs to sequentially restore the gradients reported by all terminals participating in the federated learning training to perform gradient aggregation. This causes large overhead of time-frequency resources and large delay, and is not applicable to federated learning scenarios with limited bandwidth and / or high delay requirements.

[0079] The present application provides a method for training an AI model in a wireless network. In the method, the same time-frequency resource and the same report timing can be assigned to terminals participating in federated learning. The terminals participating in federated learning report the gradient of the trained AI model at the same report timing by using the same time-frequency resource, so as to solve the above-mentioned problem of large time-frequency resource overhead and large delay caused by multiple mutually orthogonal time-frequency resources being assigned to the terminals participating in federated learning.

[0080] As shown in Figure 4, a method procedure for training an AI model in a wireless network is provided. The procedure includes at least the following steps:

[0081] Step 401: A base station sends first configuration information to a terminal participating in federated learning, where the first configuration information is used to configure at least one of the following: a training period, a time-frequency resource, or a report timing. Correspondingly, the terminal receives the first configuration information from the base station.

[0082] Step 402: The terminal trains the AI ​​model within a training period to obtain the gradient of the AI ​​model in the current round of model training.

[0083] Step 403: The terminal reports the gradient of the AI ​​model in the current round of model training to the base station at the reporting timing by using the time-frequency resource. Correspondingly, the base station receives a signal obtained by wireless superposition of the gradient reported by the terminal.

[0084] In the present application, the same training period, the same time-frequency resource, and the same report timing are set by the base station or the central node for the terminals participating in the federated learning. For example, the number of terminals participating in the federated learning is n. In the present application, the base station or the central node assigns the same time-frequency resource to the n terminals. Compared with the conventional solution in which n orthogonal time-frequency resources are assigned to the n terminals participating in the federated learning, the overhead of the time-frequency resource can be reduced. Furthermore, when n orthogonal time-frequency resources are assigned to the n terminals participating in the federated learning, each terminal reports the gradient of the AI ​​model by using the respective time-frequency resource. The base station receives n radio frequency signals and may process the n radio frequency signals separately to restore the gradient reported by each terminal. As a result, the delay is large. However, in the present application, the base station assigns one time-frequency resource to the n terminals participating in the federated learning, and the n terminals report the gradient in the time-frequency resource. Therefore, based on the superposition nature of the wireless channel, the n gradients are superposed during wireless transmission. For example, the value of n is 3, the gradient reported by terminal 1 is Y1, the gradient reported by terminal 2 is Y2, and the gradient reported by terminal 3 is Y3. In this case, the above three gradients are transmitted in the same time-frequency resource, and the above three gradients are superimposed. It is assumed that the wireless channel satisfies perfect signal superimposition (fading, interference, noise, etc. can be ignored). In this case, the superimposed signal is Y=Y1+Y2+Y3. When the channel does not satisfy perfect signal superimposition, after receiving the above signals in the above time-frequency resource, the base station may perform signal processing on the received signals to restore the superimposed signal Y, and the terminal may perform gradient aggregation by using the superimposed signal Y. The process of gradient aggregation may be a process of calculating an arithmetic mean. For example, the superimposed signal Y may be divided by 3, and the result is used as the aggregated gradient. According to the solution of the present application, the superimposed signal Y may be obtained by processing one radio frequency signal.However, in the existing solution, three radio frequency signals in different slots need to be processed sequentially to restore the corresponding gradients, and then aggregation is further performed. By using the solution of the present application, the utilization of time-frequency resources can be reduced to a certain extent, and the delay of gradient aggregation can be reduced.

[0085] The training period set by the base station (or other devices in the access network device) for the terminal may not be the actual training period of the terminal, and is usually an upper limit of the time required by the terminal for each training round. In other words, the terminal completes the current round of model training within the training period and reports a training completion indication to the base station. Otherwise, the terminal may end the current round of model training and wait until the next training period comes. In this application, the base station sets the training period to ensure that different terminals can simultaneously report the gradient of model training to the base station. Optionally, the gradient of model training reported by the terminal to the base station is the gradient in the current round of model training, not the gradient in other rounds of model training, e.g., the gradient in the previous round of model training. For example, the training period may be a global parameter determined by comprehensively considering the computational capabilities of the terminals participating in the federated learning, the complexity of the model, etc., and then set by the base station for each terminal.

[0086] In the present application, the number of terminals participating in the federated learning may be n, and each terminal may be regarded as an edge node; The central node may be a base station, or may be an OAM, or may be placed independently as a module in the core network, etc. This is not limited. In the following description, an example in which the central node is independent of the base station, in other words, the central node and the base station are two devices, is used for explanation.

[0087] 5A and 5B, a method procedure for training an AI model in a wireless network is provided. The procedure includes at least the following steps:

[0088] Step 501: each of the n terminals reports terminal information to the base station, and the base station centrally collects the terminal information and reports the collected information to the central node. In this application, the terminal information includes at least one of the following:

[0089] 1. The communications capabilities of the terminal, including, for example, the maximum transmit power that can be supported by the terminal and the antenna configuration of the terminal.

[0090] 2. The computing capabilities of the terminal, including, for example, the performance of the central processing unit (CPU), the performance of the graphics processing unit (GPU), memory space, and power.

[0091] 3. Dataset characteristics of the terminal, such as the size of the dataset, the distribution of the dataset, whether the dataset is complete, and whether the label of the dataset is complete. Optionally, the dataset may be further divided into a training set, a validation set, and a test set based on a percentage. For example, 60% of the dataset is a training set, 20% of the dataset is a validation set, and 20% of the dataset is a test set. It can be understood that the training set is used to train an AI model, the validation set is used to evaluate the trained AI model, and the test set is used to test the trained AI model.

[0092] Optionally, the base station assigns a terminal temporary identifier to the terminal. The identifier may be a cell radio network temporary identifier (C-RNTI), another temporary identifier, etc. Optionally, the other temporary identifier may be differentiated using a coding scheme, for example, using a sequence number as shown in Table 1 or Table 2. [Table 1] [Table 2]

[0093] Optionally, before step 501, the procedure may further include: the base station sending an instruction to the n terminals to report terminal information. The n terminals separately report their respective terminal information in the above step 501 based on the instruction.

[0094] Step 502: The central node sends second configuration information to the base station. Optionally, the second configuration information is used to set at least one of the following: a list of terminals participating in federated learning, an initial AI model, a training period, a terminal threshold number, a size of a transport block, an uplink requirement, etc. The uplink requirement may include a rate, a bit error rate, a delay, etc. during the uplink transmission of the terminal. The transport block is a data block including a MAC protocol data unit (PDU), and the data block is transmitted in a transmission time interval (TTI).

[0095] It should be noted that in this application, when the base station receives the terminal information reported by the terminal, the terminal information may be referred to as first terminal information. The base station assigns a temporary identifier to the terminal, and adds the temporary identifier to the terminal information to form second terminal information. In addition to the terminal temporary identifier, the second terminal information may further include at least one of the following: communication capability of the terminal, computation capability of the terminal, or dataset characteristics of the terminal. The base station reports the second terminal information to a central node.

[0096] Specifically, the central node may determine a list of terminals participating in federated learning based on the second terminal information reported by the base station. For example, the second terminal information includes the terminal's communication capability, computation capability, dataset feature, temporary identifier, etc. For example, the central node comprehensively considers the terminal's communication capability, computation capability, dataset feature, etc., and generates a list of terminals participating in federated learning. For example, during the comprehensive consideration, priorities may be set for the terminal's communication capability, computation capability, dataset feature, etc. The priority of the terminal's dataset feature is higher than the priority of the terminal's communication capability, and the priority of the terminal's communication capability is higher than the priority of the terminal's computation capability. Furthermore, corresponding thresholds are set for the terminal's communication capability, computation capability, and dataset feature, respectively, and a terminal that does not meet the threshold condition is not considered to be placed in the list of terminals participating in federated learning. For example, the central node may use a terminal whose communication capability is equal to or greater than the communication capability threshold, whose computation capability is equal to or greater than the computation capability threshold, and whose dataset feature meets the dataset feature requirement, as a terminal participating in federated learning.

[0097] In this application, the central node may set a training period for the terminals participating in the federated learning. The training period should be set to ensure that all terminals participating in the federated learning can complete local model training within the training period, taking into account the computational complexity of the AI ​​model to be trained and the computational capabilities of the terminals. However, the training period should not be set too long, so as not to affect the overall efficiency of the training of the AI ​​model.

[0098] In this application, the central node may determine the terminal threshold number. When air-based computation is introduced into federated learning, the number of terminals participating in federated learning affects the accuracy of calculating the average gradient in the current round of model training. Therefore, the central node may set the terminal threshold number. When the number of terminals reporting gradients is equal to or greater than the terminal threshold number, the average gradient in the current round of model training is calculated and sent to the terminal; otherwise, the average gradient in the previous round of model training may be sent to the terminal, or the average gradient in the previous round of model training may be used as the average gradient in the current round of model training and sent to the terminal.

[0099] Step 503: The base station sends the first setting information to the terminal.

[0100] In the present application, the first configuration information is used to configure at least one of the following: training period, reporting timing, time-frequency resources, dedicated bearer radio bearer (RB) resources, modulation scheme, initial AI model, etc.

[0101] For example, the base station may determine time-frequency resources, dedicated bearer RB resources, modulation schemes, etc. based on uplink requirements. In this application, the same time-frequency resources, the same reporting timing, and the same training period are assigned by the base station to the n terminals participating in the joint learning.

[0102] For example, the base station needs to allocate a dedicated bearer RB resource to the terminal. The dedicated bearer RB resource may be a signal radio bearer (SRB) resource, a data radio bearer (DRB) resource, etc., and is used to transmit the gradient of the AI ​​model. The dedicated bearer RB resource may only be used for transmitting the gradient and cannot be used for transmitting other data. The modulation scheme set by the base station to the terminal may be phase shift keying (PSK), quadrature amplitude modulation (QAM), other modulation schemes, etc. The specific order of PSK or QAM used for modulation may be determined based on uplink requirements, uplink channel quality, communication capabilities of the base station, communication capabilities of the terminal, etc.

[0103] Step 504: The terminal trains the AI ​​model within a training period.

[0104] In the present application, when receiving the first setting information, the terminal may train the AI ​​model. It can be understood that in the first round of training process, the terminal specifically trains an initial AI model, and the initial AI model is configured for the terminal by the central node. In the subsequent training process, the terminal specifically trains the AI ​​model obtained by the previous training. The training period of the present application can be represented by T. For a terminal participating in federated learning, if the training of the AI ​​model is completed within the training period T, a training completion indication is reported to the base station. If the training of the AI ​​model is not completed within the training period T, the model training is terminated.

[0105] Step 505: According to the training completion indication reported by the terminal, the base station collects statistics on the number of terminals that have completed the training of the AI ​​model within the training period T, and measures the uplink channel quality of the terminals that have completed the current round of model training. Optionally, for a terminal that does not complete model training in the current round of model training, the base station no longer measures the uplink channel quality of the corresponding terminal. Correspondingly, the terminal that does not complete model training no longer reports the gradient of the AI ​​model in the current round of model training to the base station.

[0106] For example, the base station may set a counter. When each round of training begins, the base station counts the terminals that report a training completion indication, and resets the counter to 0 when each round of training ends. When the training period T ends, if the counter count is greater than or equal to the terminal threshold number, the base station measures the uplink channel quality of the terminal, otherwise, the base station triggers the terminal to perform the next training round.

[0107] Optionally, for a terminal, the process of the base station measuring the uplink channel quality of the terminal includes: the base station receiving a sounding reference signal (SRS) from the terminal. The base station measures the SRS to determine the uplink channel quality of the terminal.

[0108] Step 506: The base station determines the transmission power of the corresponding terminal according to the uplink channel quality of the terminal, and the base station sends third configuration information to the terminal, where the third configuration information is used to configure the transmission power, report timing, etc. of the terminal. In this application, the third configuration information and the first configuration information may be collectively referred to as one configuration information.

[0109] For example, the base station may determine the transmission power of the terminal by comprehensively considering conditions such as the error requirements of the airborne calculation, the uplink channel quality of the terminal, the maximum transmission power supported by the terminal, and the total power. Since the airborne calculation has requirements on node synchronization, excessively large synchronization will affect the accuracy of the airborne calculation. Therefore, the base station may set the same report timing for n terminals participating in federated learning. Furthermore, since the channel is constantly changing, there is a validity period for the channel quality measurement. Therefore, the terminal needs to simultaneously report the gradient of the AI ​​model in the current round of model training at the report timing, and the report timing should be within the validity period of the channel quality measurement.

[0110] Optionally, the effective period of the channel quality measurement result may be time-domain, taking into account the time overhead for determining the transmission power of the terminal by the base station, which may result in a mismatch between the power allocation and the uplink channel quality, etc. In the design, the base station may predict the uplink channel quality at the current reporting time point based on the terminal's past uplink channel quality, and optimize the optimal transmission power in advance. A specific prediction method is shown in FIG. 6. The base station collects the past uplink channel condition quality of each terminal between time t1 and time t2, starts to predict the uplink channel quality of each terminal at time t3 based on the terminal's past uplink channel quality at time t2, and delivers the optimal power distribution solution to the terminal at time t3. After receiving the power distribution solution, the terminal immediately reports the gradient in the current round of model training.

[0111] Step 507: The terminal reports the gradient in the current round of model training, the base station calculates the average gradient in the current round of model training, the base station updates the parameters of the AI ​​model based on the average gradient in the current round of model training, and distributes the average gradient in the current round of model training to the n terminals participating in the federated learning.

[0112] For example, if n terminals participating in federated learning complete training within the training period T at the reporting timing, The terminal may report the gradient in the current round of model training to the base station. When receiving the gradient in the current round of model training reported by n terminals, the base station may calculate an average gradient in the current round of model training based on the gradient reported by the terminal. For example, since n terminals report the gradient in the current round of model training in the same time-frequency resource, the signal received by the base station is a signal obtained by over-the-air superposition of n gradients. For example, if the gradient obtained by superposition of n gradients is Y, the base station may determine that the average gradient in the current round of model training is Y / n. Optionally, in the present application, a terminal that does not complete model training within the training period T no longer reports the gradient in the current round of model training to the base station.

[0113] In this application, in order to reduce the reporting time error of the terminal as much as possible, each terminal should start reporting the gradient in the current round of model training as soon as possible when the reporting timing arrives. The base station needs to process the received signal obtained by air calculation to restore the average gradient in the current round of model training. For example, if the count of the counter in the base station is equal to or greater than the terminal threshold number, the base station distributes the locally calculated average gradient in the current round of model training to each terminal. Otherwise, the average gradient in the previous round of model training is used as the average gradient in the current round of model training and distributed to each terminal.

[0114] Optionally, if the terminals participating in the federated learning are changed, the base station may send updated model parameters to the terminals.

[0115] It can be understood that steps 504 to 507 are a cyclic process. After the average gradient in the current round of model training is delivered in step 507, the terminal may update the parameters and gradients of the AI ​​model based on the average gradient in the current round of model training, and report a training completion indication to the base station if the model training is completed within the training period T. It should be noted that in machine learning, in order to reduce the loss function, the parameters of the AI ​​model need to fall along the negative direction of the gradient, that is, the gradient falls. In the present application, the parameters of the AI ​​model may first be updated in the current round of model training based on the average gradient in the previous round of model training, and then the gradient of the AI ​​model is updated based on the updated parameters, that is, the gradient is updated.

[0116] Step 508: The base station determines a model training end condition, and sends a model training end instruction to the terminal.

[0117] For example, the base station may determine a model training end condition and send a model training end instruction to each terminal. The model training end condition may be at least one of the following: model parameters converge, a maximum number of times of model training is reached, a maximum time of model training is reached, etc.

[0118] In this application, air-based computation is introduced to reduce communication delay, time-frequency resource overhead, and signaling overhead in the federated learning process. The set training period and set report timing can reduce the time synchronization error of reporting gradients by terminals. Before the gradients are reported, the base station first measures the uplink channel quality of all participating terminals and optimizes the transmission power of each terminal to improve the performance of air-based computation and further improve the training effect of federated learning.

[0119] As shown in Figure 7A to Figure 7C, the present application provides a procedure of a method for training an AI model in a wireless network. The main difference between this procedure and the procedure shown in Figure 5A and Figure 5B is that in this procedure, the terminal determines the transmission power for reporting the gradient in the current round of model training by itself, and the transmission power is no longer set by the base station. The procedure includes at least the following steps:

[0120] Step 701: n terminals report terminal information to a base station, and the base station centrally reports the terminal information to a central node.

[0121] Specifically, after receiving the terminal information of the terminal, the terminal may assign a temporary identifier to the terminal, add the temporary identifier to the terminal information, and report the terminal information to the central node. See the description of Figures 5A and 5B.

[0122] Step 702: The central node sends second setting information to the base station, where the second setting information is used to set at least one of the following: a list of terminals participating in federated learning, an initial AI model, a training period, a terminal threshold number, a channel state information (CSI) interval, a channel inversion parameter, a transport block size, or an uplink requirement. For the list of terminals participating in federated learning, the initial AI model, the training period, the terminal threshold number, the transport block size, the uplink requirement, etc., please refer to the description of Figures 5A and 5B. This procedure focuses on the CSI interval and the channel inversion parameter.

[0123] In this application, CSI is channel state information, which includes signal-to-noise ratio, Doppler frequency shift, multipath delay spread, etc. The CSI interval includes a signal-to-noise ratio interval, a maximum Doppler frequency shift interval, and a maximum delay spread interval. The signal-to-noise ratio interval is [γ min ,γ max ], and γ min and γmax are the lower and upper limits of the signal-to-noise ratio, respectively. The maximum Doppler frequency shift interval is [f min ,f max ] and f min and f max are the lower and upper bounds of the maximum Doppler frequency shift, respectively. The maximum delay spread interval is [τ min , τ max ], and τ min and τ max are the lower and upper bounds, respectively, of the maximum delay spread.

[0124] In this application, the gradient in the current round of model training can be reported to the base station only when the terminal's downlink CSI fills the CSI interval. Otherwise, the gradient is not reported. For example, the signal-to-noise ratio, maximum Doppler frequency shift, and maximum delay spread measured by the terminal are γ1, f1, and τ1, respectively. γ min ≦γ1≦γ max , f min ≦f1≦f max , and τ min ≦τ1≦τ max If , the terminal may report the gradient in the current round of model training to the base station at the reporting timing. Otherwise, the terminal does not report the gradient in the current round of model training.

[0125] In this application, the channel inversion parameter α is a parameter used for power control. The maximum power and channel gain of the kth terminal are P k and h k and P1|h1| 2 ≦ ≦ P k |h k | 2 ≦ ≦ P K |h K | 2 In this case, it is assumed that α=P1|h1| 2 and the transmit power used by the kth terminal to report the gradient is p k =α / |h k | 2 It should be.

[0126] Step 703: The base station determines time-frequency resources, dedicated bearer RB resources, scheduling schemes, etc. based on uplink requirements. The base station sends first configuration information to the terminal, where the first configuration information is used to configure at least one of the following: training period, report timing, time-frequency resources, initial AI model, dedicated bearer RB resources, modulation schemes, etc.

[0127] In this application, the dedicated bearer RB resource includes a dedicated SRB resource, a dedicated DRB resource, and / or the like. The base station may determine the time-frequency resource, the SRB / DRB resource, the modulation scheme, etc. based on the uplink requirement. To meet the requirement of the air computation, all the terminals need to use the same time-frequency resource. Optionally, the base station may assign the same time-frequency resource to the n terminals to improve the performance of the air computation by using time diversity or frequency diversity. It should be noted that, when the base station configures the same time-frequency resource to the n terminals, the n terminals report the gradient in the current round of the model training to the base station at a specific report timing by using the time-frequency resource. For example, when the base station configures three time-frequency resources to the n terminals, the n terminals report the gradient in the current round of the model training at the report timing by simultaneously using the first time-frequency resource, the second time-frequency resource, and the third time-frequency resource among the three time-frequency resources. That is, at the reporting timing, when n terminals report the gradients in the current round of model training, the same time-frequency resource is specifically used. To avoid interference from other data, the base station may allocate independent SRB / DRB resources to the terminals to transmit the gradients in the current round of model training. The base station may set a modulation scheme, such as PSK or QAM, for the terminals. The specific order of PSK or QAM used for modulation may be determined based on at least one of the following: uplink requirements, uplink channel quality, communication capability of the base station, communication capability of the terminal, etc.

[0128] Step 704: The terminal locally trains the model.

[0129] In this application, when the terminal receives the first setting information, the terminal may start to perform model training. If the terminal completes the model training within the training period T, the terminal reports a training completion indication to the base station. If the terminal does not complete the model training within the training period T, the terminal terminates the model training.

[0130] Step 705: The terminal measures the CSI of the downlink channel and determines the transmit power of the terminal.

[0131] In this application, the same frequency resource may be configured for the uplink channel and the downlink channel of the terminal. Based on the channel reciprocity, the terminal may measure the CSI of the downlink channel to obtain the CSI of the uplink channel. The terminal may determine whether the obtained CSI of the uplink channel is within the CSI interval. If the obtained CSI of the uplink channel is within the CSI interval, the transmission power is determined based on the channel inversion parameter. From the above description, it can be seen that the channel inversion parameter is α=P1|h1| 2 and the transmit power used by the kth terminal to report the gradient is p k =α / |h k | 2 In this application, for the k-th terminal, the parameter h k can be obtained using the CSI, and the transmission power p k may be determined with reference to a channel inversion parameter α, where k is a positive integer greater than or equal to 1 and less than or equal to n.

[0132] Step 706: The terminal reports the gradient in the current round of model training, and the base station calculates the average gradient in the current round of model training, updates the parameters of the AI ​​model according to the average gradient in the current round of model training, and sends the average gradient in the current round of model training to each terminal participating in federated learning. For the specific process of step 706, please refer to step 507 above.

[0133] Step 707: The base station determines a model termination condition and sends a model training termination instruction to each terminal.

[0134] It should be noted that in the design, the terminal may use a set training period and a set reporting timing to ensure that the terminal reports the gradients in local training at the same time. For example, after receiving the average gradients in the previous round of model training distributed by the base station, the terminal starts to perform the current round of model training, and reports the local training gradients in the current round of model training T seconds after receiving the average gradients in the previous round of model training. T seconds is the training period, and T seconds after receiving the average gradients in the previous round of model training is the reporting timing.

[0135] In this application, air-based computation is introduced to reduce communication delay and bandwidth overhead in the federated learning process, and the set training period and the set report timing can reduce the time error of reporting the gradient by the terminal. The base station transmits parameters such as the training period, the power adjustment solution, and the report timing to each terminal participating in the federated learning in a pre-configured manner, and the terminal periodically trains the model and reports the gradient. Before reporting the gradient, the terminal optimizes the transmission power of the terminal through the measurement of the downlink channel quality by using the channel reciprocity, thereby improving the performance of the air-based computation and further improving the training effect of the federated learning. Furthermore, the terminal actively reports the training gradient when the training period ends, so that the signaling overhead of the base station for scheduling the terminal can be reduced.

[0136] As shown in Figure 8A to Figure 8D, a flowchart of a method for training an AI model in a wireless network is provided. The main difference between the flowchart and the procedure shown in Figure 5A and Figure 5B is that the terminals participating in federated learning are grouped, and in this case, a central node schedules the terminals in a certain group to centrally report the gradient in the current round of model training. The flowchart includes at least the following steps:

[0137] Step 801: A terminal reports terminal information to a base station, and the base station centrally reports the terminal information to a central node.

[0138] For a specific explanation, please refer to the explanations of Figures 5A and 5B.

[0139] Step 802: The central node sends second configuration information to the base station.

[0140] In the present application, the central node may determine a list of terminals participating in federated learning, an initial AI model, a training period, a terminal threshold number, a transport block size, an uplink requirement, a group temporary identifier, etc. The second configuration information is used to configure at least one of the following: a list of terminals participating in federated learning, an initial AI model, a training period, a terminal threshold number, a transport block size, an uplink requirement, a group temporary identifier, etc.

[0141] In this application, the central node may comprehensively consider the communication capabilities, computing capabilities, data set features, etc. of the terminals to participate in the federated learning within the service range of each base station, group the terminals that participate in the federated learning and are within the service range of the base station into one group, and assign a temporary identifier to the group. The identifier may be called a group temporary identifier. For example, during the comprehensive consideration, the central node may set priorities for the communication capabilities, computing capabilities, data set features, etc. of the terminals, where the priority of the data set features is higher than the priority of the communication capabilities of the terminals, and the priority of the communication capabilities of the terminals is higher than the priority of the computing capabilities of the terminals. Furthermore, corresponding thresholds are set for the communication capabilities, computing capabilities, and data set features of the terminals, respectively, and the terminals that do not meet the threshold conditions are not considered to be placed in the list of terminals to participate in the federated learning. In this application, the terminals that participate in the federated learning and are within the service range of the base station may be grouped into one group. For example, the procedures in Figures 8A to 8D use an example in which m terminals in the coverage area of ​​base station 1 are grouped into one group, and n terminals in the coverage area of ​​base station N are grouped into another group. Both m and n are positive integers, and the values ​​of m and n may be the same or different. For each group, the terminal may assign a temporary identifier, called a group temporary identifier, to the group. In a possible implementation, two groups are used as an example. Please refer to Table 3 below for a list of terminals participating in federated learning in each group. [Table 3]

[0142] Step 803: The base station sends first setting information to the terminal, and the first setting information is used to set at least one of the following: an initial AI model, a group temporary identifier, a training period, time-frequency resources, dedicated bearer RB resources, a modulation scheme, etc.

[0143] In this application, the base station may determine the time-frequency resources, the dedicated bearer RB resources, the modulation scheme, etc. based on the uplink requirements in the first configuration information. Alternatively, the time-frequency resources may be configured by a central node and distributed to each terminal through the base station. The central node may assign the same time-frequency resources to all terminals in the same group.

[0144] Step 804: The central node sends a scheduling instruction to the base station, and the base station forwards the scheduling instruction to the terminal, where the scheduling instruction includes the group temporary identifier, and the scheduling instruction is used to schedule the terminal corresponding to the group temporary identifier to perform training of the AI ​​model in the current round of model training.

[0145] In this application, an example is used in which n terminals in the coverage area of ​​one base station are grouped into one group. When the central node specifically schedules a specific group to perform model training and report the gradient in the current round of model training, the central node sends a scheduling instruction to the base station corresponding to the group. Upon receiving the scheduling instruction, the base station may broadcast the scheduling instruction in the coverage area of ​​the base station. Upon receiving the scheduling instruction, the terminal may compare the group temporary identifier carried in the scheduling instruction with the group temporary identifier assigned to the terminal by the central node. If the two are the same, the terminal performs training of the AI ​​model in the current round of model training, or otherwise, the terminal does not perform training of the AI ​​model in the current round of model training.

[0146] Similar to the procedure shown in Figures 5A and 5B, if the terminal has completed the training of the AI ​​model when the training period ends, the terminal reports a training completion indication to the base station, and if the terminal has not completed the training of the AI ​​model when the training period ends, the terminal also ends the training of the AI ​​model. The base station counts the number of terminals that have completed model training within the training period based on the training completion indication reported by the terminal, and reports the number of terminals to the central node.

[0147] Step 805: The base station sends third setting information to the terminals, and the third setting information is used to set the transmission power and the report timing of each terminal. Optionally, the base station sets the same report timing for the terminals. Alternatively, the report timing may be set by a central node and transferred to each terminal by the base station, in other words, the second setting information may be further used to set the report timing. In this application, the third setting information and the first setting information may be referred to as one setting information.

[0148] For example, the process of determining the transmission power by the base station includes the base station measuring the SRS from the terminal to determine the uplink channel quality of the terminal. The terminal determines the transmission power of the terminal based on the uplink channel quality. For details, please refer to the description of Figures 5A and 5B. The details will not be described again here.

[0149] Step 806: The terminal reports the gradient in the current round of model training to the central node, and the central node calculates the average gradient in the current round of model training, updates the parameters of the AI ​​model based on the average gradient in the current round of model training, and distributes the average gradient in the current round of model training to the terminals in the current scheduling group via the base station.

[0150] Specifically, the central node may compare the value relationship between the number of terminals that have completed the training of the AI ​​model within the training period and the terminal threshold number. If the number of terminals that have completed the training of the AI ​​model within the training period is greater than the terminal threshold number, the central node calculates the average gradient in the current round of model training based on the gradient in the current round of model training reported by the terminal; otherwise, the central node uses the average gradient in the previous round of model training as the average gradient in the current round of model training. The central node updates the parameters of the AI ​​model based on the average gradient in the current round of model training, and transmits the average gradient in the current round of model training to the terminals in the scheduling group through the base station.

[0151] Step 807: The central node determines a model training termination condition, and sends a model training termination instruction to the corresponding terminal.

[0152] For the model training end condition, please refer to the procedure description of Figures 5A and 5B or 7A to 7C. The difference is that in the procedure description of Figures 8A to 8D, the central node determines whether the model training end condition is satisfied, and in the procedure shown in Figures 5A and 5B or 7A to 7C, the base station determines whether the model training end condition is satisfied. It can be understood that the central node may first send a model training end instruction to the base station, and the base station forwards the model training end instruction to the terminal.

[0153] In the procedure shown in Figures 8A to 8D, the central node may schedule the terminals in all groups to participate in the federated learning training, or may schedule only the terminals in some groups to participate in the federated learning training. In the above description, when receiving a scheduling instruction, the terminal may perform the federated learning training, or otherwise, the terminal does not perform the federated learning training. Optionally, the terminal may further set monitoring parameters of the terminal, where the parameters include a monitoring timing and a monitoring period. When the monitoring timing arrives, the scheduling instruction is monitored based on the set monitoring period. When the monitoring period is reached, the terminal stays in a sleep mode to save the power of the terminal.

[0154] In the above solution, the terminals are grouped, and only a part of the terminals may be scheduled to perform federated learning training. Compared with the solution in which all terminals need to participate in federated learning, the power consumption of the terminals may be reduced.

[0155] It can be understood that, to implement the functions in the above methods, the base station, the terminal, and the central node include corresponding hardware structures and / or software modules for performing the functions. Those skilled in the art should easily realize that the present application can be implemented by hardware or a combination of hardware and computer software with reference to the units and method steps in the examples described in this application. Whether the functions are performed by hardware or hardware driven by computer software depends on the specific application scenario and design constraints of the technical solution.

[0156] 9 and 10 are diagrams of possible communication device structures according to the present application. These communication devices may be configured to implement the functions of the terminal, base station, or central node in the above method, and thus can also realize the advantageous effects of the above method. In the present application, when implementing the functions of the terminal, the communication device may be one of the terminals 120a-120j shown in FIG. 1, and when implementing the functions of the base station, the communication device may be the base station 110a or 110b shown in FIG. 1, or may be a module (e.g., a chip) used in the terminal or base station.

[0157] As shown in Fig. 9, the communication device 900 includes a processing unit 910 and a transceiver unit 920. The communication device 900 is configured to implement functions of a terminal, a base station, or a central node in the methods shown in Fig. 4, Fig. 5A and Fig. 5B, Fig. 7A-7C, or Fig. 8A-8D.

[0158] When the communication device 900 is configured to implement the functions of the base station in the method shown in Figure 4, Figure 5A and Figure 5B, Figure 7A-Figure 7C, or Figure 8A-Figure 8D, the transceiver unit 920 is configured to transmit first setting information to the terminal participating in the federated learning, and the first setting information is used to set at least one of the following training period, time-frequency resource, and report timing, where the same training period, the same time-frequency resource, and the same report timing are set for different terminals participating in the federated learning, and receive a signal obtained by wireless superposition of gradients reported by the terminals participating in the federated learning, where the gradient is the gradient reported by the terminal at the report timing by using the time-frequency resource of the AI ​​model whose training is completed within the training period. The processing unit 910 is configured to generate the first setting information and process the gradient reported by the terminal.

[0159] When the communication device 900 is configured to implement the functions of a terminal in the method shown in Figures 4, 5A and 5B, 7A to 7C, or 8A to 8D, the transceiver unit 920 is configured to transmit first setting information from the second node, the first setting information is used to set at least one of the next training period, time-frequency resources, and reporting timing, the same training period, the same time-frequency resources, and the same reporting timing are set for different terminals participating in the federated learning, the processing unit 910 is configured to train the AI ​​model within the training period to obtain a gradient of the AI ​​model in the current round of model training, and the transceiver unit 920 is further configured to report the gradient of the AI ​​model in the current round of model training to the second node at the reporting timing by using the time-frequency resources.

[0160] When the communication device 900 is configured to implement the functionality of a central node in the method shown in Figures 4, 5A and 5B, 7A to 7C, or 8A to 8D, the processing unit 910 is configured to determine second setting information, which is used to set at least one of a list of terminals to participate in the next federated learning, an initial AI model, a group temporary identifier, a training period, a number of terminal thresholds, a transport block size, or an uplink requirement, and the transceiver unit 920 is configured to transmit the second setting information to the second node.

[0161] For more detailed description of the processing unit 910 and the transceiver unit 920, please refer to the relevant descriptions in the methods shown in Figures 4, 5A and 5B, 7A-7C, or 8A-8D. The details will not be described again here.

[0162] As shown in Fig. 10, the communication device 1000 includes a processor 1010 and an interface circuit 1020. The processor 1010 and the interface circuit 1020 are coupled to each other. It can be understood that the interface circuit 1020 may be a transceiver or an input / output interface. Optionally, the communication device 1000 may further include a memory 1030 configured to store instructions to be executed by the processor 1010, or to store input data required by the processor 1010 to execute the instructions, or to store data generated after the processor 1010 executes the instructions.

[0163] When the communications device 1000 is configured to implement the above method, the processor 1010 is configured to implement the functionality of the processing unit 910 and the interface circuit 1020 is configured to implement the functionality of the transceiver unit 920.

[0164] If the communication device is a chip used in a terminal, the chip of the terminal implements the functions of the terminal in the above method. The chip in the terminal receives information from other modules (such as a radio frequency module or an antenna) in the terminal, and the information is transmitted to the terminal by the base station, or the chip in the terminal transmits information to other modules (such as a radio frequency module or an antenna) in the terminal, and the information is transmitted to the base station by the terminal.

[0165] If the communication device is a module used in a base station, the module in the base station implements the function of the base station in the above method. The module in the base station receives information from other modules (e.g., radio frequency modules or antennas) in the base station, and the information is transmitted to the base station by the terminal, or the module in the base station transmits information to other modules (e.g., radio frequency modules or antennas) in the base station, and the information is transmitted to the terminal by the base station. The module in the base station here may be a baseband chip in the base station, or may be a DU or other module. The DU here may be a DU in an open radio access network (O-RAN) architecture.

[0166] If the above device is a module used in a central node, the central node implements the function of the central node in the above method. The central node receives information from other modules (such as radio frequency modules or antennas) in the central node, and the information is transmitted to the central node by the base station, or the modules in the central node transmit information to other modules (radio frequency modules or antennas) in the central node, and the information is transmitted to the base station by the central node. The modules in the central node may be baseband chips or other modules in the central node.

[0167] The processor of the present application may be a central processing unit (CPU) or other general-purpose processor, digital signal processor (DSP), or application-specific integrated circuit (ASIC), field programmable gate array (FPGA), or other programmable logic device, transistor logic device, hardware component, or any combination thereof. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0168] The memory of the present application may be random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disk, removable hard disk, CD-ROM, or any other form of storage medium well known in the art.

[0169] For example, a storage medium may be coupled to the processor such that the processor can read information from, and write information to, the storage medium. The storage medium may alternatively be components of the processor. The processor and the storage medium may reside in an ASIC. Further, the ASIC may reside in a base station or terminal. Indeed, the processor and the storage medium may reside as discrete components in a base station or terminal.

[0170] Part or all of the method of the present application may be implemented by software, hardware, firmware, or any combination thereof. When software is used to implement the method, part or all of the method may be implemented in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer programs or instructions are loaded into a computer and executed, the procedures or functions according to the present application are generated completely or partially. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, a core network device, an OAM, or other programmable device. The computer program or instructions may be stored in a computer-readable storage medium or may be transmitted from a computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless manner. The computer-readable storage medium may be any useful medium or data storage device that can be accessed by a computer, such as a server or a data center incorporating one or more useful media. The useful medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape, or an optical medium, such as a digital video disk, or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include two types of storage media, i.e., a volatile storage medium and a non-volatile storage medium.

[0171] In this application, unless otherwise stated or there is no logical contradiction, the terms and / or descriptions in different embodiments may be consistent and cross-referenced, and the technical features in different embodiments may be combined based on their internal logical relationships to form new embodiments.

[0172] In this application, "at least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship between related objects, and indicates that three relationships may exist. For example, A and / or B can indicate that A exists alone, that both A and B exist, and that B exists alone, and A and B may be singular or plural. In the text description of this application, the " / " character generally indicates an "or" relationship between related objects. In formulas in this application, the " / " character indicates a "division" relationship between related objects. "Comprising at least one of A, B, or C" may indicate including A, including B, including C, including A and B, including A and C, including B and C, or including A, B, and C.

[0173] It can be understood that various numbers used in this application are only distinguished for ease of description, and are not intended to limit the scope of this application. The sequential numbers of the above processes do not imply an execution order, and the execution order of the processes should be determined based on the functions and internal logic of the processes.

[0174] [CROSS REFERENCE TO RELATED APPLICATIONS] This application claims priority to Chinese Patent Application No. 202111505116.7, entitled "METHOD FOR TRAINING ARTIFICIAL INTELLIGENCE AI MODEL IN WIRELESS NETWORK AND APPARATUS," filed with the State Intellectual Property Office of the People's Republic of China on December 10, 2021, and the prior Chinese patent application is incorporated herein by reference in its entirety.

Claims

1. 1. A method for training an artificial intelligence (AI) model in a wireless network, comprising: Sending first setting information to terminals participating in the federated learning, the first setting information being used to set at least one of a next training period, a time-frequency resource, or a report timing, and the same training period, the same time-frequency resource, or the same report timing is set for different terminals participating in the federated learning; receiving a signal obtained by wireless superposition of gradients reported by the terminals participating in federated learning, the gradients being gradients reported by the terminals at the reporting timing by using the time-frequency resources of an AI model whose training is completed within the training period; The method according to claim 1,

2. receiving a training completion indication from the terminal, the training completion indication being sent by the terminal to a second node if training of the AI ​​model is completed within the training period; Collecting statistics regarding the number of terminals that have completed training of the AI ​​model within the training period based on the training completion indication sent by the terminal; The method of claim 1 further comprising:

3. When the number of terminals that have completed training of the AI ​​model is equal to or greater than a terminal threshold number, determine an average gradient in the current round of model training based on the gradients reported by the terminals participating in the federated learning, or when the number of terminals that have completed training of the AI ​​model is less than a terminal threshold number, use the average gradient in the previous round of model training as the average gradient in the current round of model training; updating parameters of the AI ​​model based on the average gradient in the current round of model training, and sending the average gradient in the current round of model training to the terminal; The method of claim 2 further comprising:

4. and transmitting to a first node the number of terminals that have completed training of the AI ​​model within the training period and the signal obtained by the wireless superposition of the gradients reported by the terminals. The method of claim 2.

5. The first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, modulation scheme, initial AI model, or transmission power.

5. The method according to any one of claims 1 to 4.

6. The process of determining a transmission power comprises: measuring a sounding reference signal (SRS) from the terminal to determine an uplink channel quality of the terminal; determining the transmit power of the terminal based on the uplink channel quality; having The method according to claim 5.

7. The first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, modulation scheme, initial AI model, channel state information (CSI) interval, or channel inversion parameters.

5. The method according to any one of claims 1 to 4.

8. receiving second configuration information from the first node; The second setting information is used to set at least one of the following: a list of terminals participating in federated learning, the initial AI model, a group temporary identifier, the training period, the terminal threshold number, a transport block size, or an uplink requirement.

8. The method according to any one of claims 1 to 7.

9. receiving first terminal information from the terminal and transmitting second terminal information to the first node; The first terminal information includes at least one of the following: a communication capability of the terminal, a computation capability of the terminal, or a dataset characteristic of the terminal; and the second terminal information includes at least one of the following: the communication capability of the terminal, the computation capability of the terminal, the dataset characteristic of the terminal, or a terminal temporary identifier, and the terminal temporary identifier is assigned to the terminal by the second node.

9. The method according to any one of claims 1 to 8.

10. sending a model training end instruction to the terminal when a model training end condition is satisfied; or receiving a model training end instruction from the first node and transmitting the model training end instruction to the terminal; The method of any one of claims 1 to 9, further comprising:

11. 1. A method for training an artificial intelligence (AI) model in a wireless network, comprising: receiving first configuration information from a second node, the first configuration information being used to configure at least one of a next training period, a time-frequency resource, or a report timing, and the same training period, the same time-frequency resource, or the same report timing is configured for different terminals participating in the federated learning; training an AI model within the training period to obtain a gradient of the AI ​​model in a current round of model training; reporting the gradients of the AI ​​model in the current round of the model training to the second node at the reporting timing by using the time-frequency resources; The method according to claim 1,

12. When the training period ends, if training of the AI ​​model is complete, sending a training completion indication to the second node. The method of claim 11 further comprising:

13. and terminating training of the AI ​​model if training of the AI ​​model is not completed within the training period.

13. The method according to claim 11 or 12.

14. receiving an average gradient in a previous round of model training from the second node, and updating the gradient of the AI ​​model in the current round of model training based on the average gradient in the previous round of model training; or updating the gradients and parameters of the AI ​​model in the current round of model training based on an average gradient in the current round of model training.

14. The method according to any one of claims 11 to 13.

15. The first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, modulation scheme, initial AI model, or transmission power.

15. The method according to any one of claims 11 to 14.

16. The first configuration information is further used to configure at least one of the following: dedicated bearer RB resources, modulation scheme, initial AI model, channel state information (CSI) interval, or channel inversion parameters.

15. The method according to any one of claims 11 to 14.

17. When the first setting information is further used to set the channel state information (CSI) interval and the channel inversion parameter, the method further comprises: if the same frequency resources are configured for a downlink channel and an uplink channel, determining a CSI of the uplink channel of the terminal based on the measured CSI of the downlink channel; If the CSI of the uplink channel meets the CSI interval requirement, determining the transmit power based on the channel inversion parameter; Reporting the gradients of the AI ​​model in the current round of model training to the second node includes: reporting the gradient of the AI ​​model in a current round of the model training to the second node based on the determined transmit power.

17. The method of claim 16.

18. The first setting information further includes a group temporary identifier, the group temporary identifier being a group temporary identifier assigned to the terminal by a first node; 18. The method according to any one of claims 11 to 17.

19. receiving a scheduling indication from the second node, the scheduling indication including a group temporary identifier; When the group temporary identifier included in the scheduling instruction is the same as the group temporary identifier assigned to the terminal by the first node, executing the training of the AI ​​model in the current round of model training, or when the group temporary identifier included in the scheduling instruction is not the same as the group temporary identifier assigned to the terminal by the first node, skipping the execution of the training of the AI ​​model in the current round of model training.

20. The method of claim 18, further comprising:

20. receiving a model training end indication from the second node; Ending the training of the AI ​​model based on the model training end instruction.

20. The method of any one of claims 11 to 19, further comprising:

21. transmitting first terminal information to the second node; The first terminal information includes at least one of the following: communication capabilities of the terminal, computational capabilities of the terminal, or dataset characteristics of the terminal; 21. The method according to any one of claims 11 to 20.

22. 1. A method for training an artificial intelligence (AI) model in a wireless network, comprising: determining second setting information, the second setting information being used to set at least one of a list of terminals participating in the federated learning, an initial AI model, a group temporary identifier, a training period, a terminal threshold number, a transport block size, or an uplink requirement; transmitting the second configuration information to a second node; The method according to claim 1,

23. receiving second terminal information from the second node, the second terminal information including at least one of the following: a communication capability of the terminal, a computation capability of the terminal, a dataset characteristic of the terminal, or a terminal temporary identifier, the terminal temporary identifier being assigned to the terminal by the second node; determining the list of terminals that will participate in federated learning based on the terminal information; 23. The method of claim 22, further comprising:

24. Receive from the second node a signal obtained by wireless superposition of gradients reported by the terminals participating in the federated learning and a number of terminals that have completed training of the AI ​​model within the training period; When the number of the terminals that have completed the training of the model within the training period is equal to or greater than the terminal threshold number, determine an average gradient in the current round of model training based on the gradient of the AI ​​model reported by the terminal; when the number of the terminals that have completed the training of the model within the training period is less than the terminal threshold number, use the average gradient in the previous round of model training as the average gradient in the current round of model training; updating parameters of the AI ​​model based on the average gradient in the current round of the model training, and transmitting the average gradient in the current round of the model training to the second node, so that the second node can transmit the average gradient in the current round of the model training to the terminal; 24. The method of claim 22 or 23, further comprising:

25. and sending a scheduling instruction to the second node, the scheduling instruction including a group temporary identifier, the scheduling instruction being used to schedule a terminal corresponding to the group temporary identifier to perform training of the AI ​​model in the current round of the model training.

25. A method according to any one of claims 22 to 24.

26. and if a model training termination condition is satisfied, sending a model training termination instruction to the second node to instruct the terminal to terminate training of the AI ​​model in the current round of model training.

26. A method according to any one of claims 22 to 25.

27. 1. An apparatus for training an artificial intelligence (AI) model in a wireless network, comprising: Apparatus comprising a unit adapted to carry out the method according to any one of claims 1 to 10.

28. 1. An apparatus for training an artificial intelligence (AI) model in a wireless network, comprising: A processor and a memory, The processor is configured to perform a method according to any one of claims 1 to 10. Device.

29. 1. An apparatus for training an artificial intelligence (AI) model in a wireless network, comprising: Apparatus comprising a unit adapted to carry out the method according to any one of claims 11 to 21.

30. 1. An apparatus for training an artificial intelligence (AI) model in a wireless network, comprising: A processor and a memory, The processor is configured to perform a method according to any one of claims 11 to 21. Device.

31. 1. An apparatus for training an artificial intelligence (AI) model in a wireless network, comprising: Apparatus comprising a unit adapted to carry out the method according to any one of claims 22 to 26.

32. 1. An apparatus for training an artificial intelligence (AI) model in a wireless network, comprising: A processor and a memory, The processor is configured to perform a method according to any one of claims 22 to 26. Device.

33. An apparatus according to claim 27 or 28; A system comprising an apparatus according to claim 29 or 30.

34. 33. The apparatus according to claim 31 or 32, 34. The system of claim 33.

35. A computer-readable storage medium storing instructions, comprising: The instructions, when executed on a computer, are capable of causing the computer to carry out the method according to any one of claims 1 to 10, the method according to any one of claims 11 to 21, or the method according to any one of claims 22 to 26. A computer-readable storage medium.

36. 1. A computer program product having instructions, comprising: The instructions, when executed on a computer, are capable of causing the computer to carry out the method according to any one of claims 1 to 10, the method according to any one of claims 11 to 21, or the method according to any one of claims 22 to 26. Computer program products.

37. A chip system comprising a processor or circuitry configured to perform the method of any one of claims 1 to 10, the method of any one of claims 11 to 21, or the method of any one of claims 22 to 26.

Images