Artificial intelligence model training method and communication device

Abstract

The invention provides a training method of an artificial intelligence model and a communication device, and relates to the technical field of communication. In the training method of the artificial intelligence model, the parameters needing to be updated in the first model can be determined according to the obtained identification information of the first scene, and the parameters needing to be updated in the first model are updated instead of all the parameters, so that the overall expenditure of model updating is effectively reduced.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and more specifically, to a training method and a communication device for an artificial intelligence model. Background Art

[0002] Artificial intelligence (AI) has been widely used in the field of communication technology in order to improve the communication performance between terminal devices and network devices.

[0003] In the specific application of AI, the AI ​​model can be applied to certain scenarios. When the AI ​​model switches from one scenario to another, in order to enable the AI ​​model to be applied to the newly switched scenario, the data of the newly switched scenario can be obtained, and the AI ​​model can be updated using the data of the newly switched scenario to obtain an updated AI model, which can be applied to the newly switched scenario. However, the above solution may result in a large overall overhead for model updates. Summary of the invention

[0004] The present application provides a training method and communication device for an artificial intelligence model, in order to reduce the overall overhead of model updating.

[0005] In a first aspect, a training method for an artificial intelligence model is provided, comprising: obtaining identification information of a first scene; determining a first parameter in the first model based on the identification information of the first scene, wherein the first parameter is used to determine a parameter that needs to be updated in the first model.

[0006] The solution described in the first aspect can be applied to a model training network element or a model deployment network element, for example, can be executed by a model training network element or a model deployment network element or a chip or module used for a model training network element or a model deployment network element.

[0007] By determining the parameters that need to be updated in the first model based on the identification information of the scene, it is helpful to update the parameters that need to be updated in the first model without updating all the parameters of the first model. In this way, the overall cost of model updating can be effectively reduced.

[0008] A possible implementation method, which determines the first parameter in the first model based on identification information of the first scene, includes: determining the first parameter based on identification information of the first scene and at least one association relationship, wherein a first association relationship in the at least one association relationship is used to associate the first parameter with the first scene.

[0009] By setting association relationships between multiple different scenes and parameters in the first model, and determining the first parameter associated with the first scene based on the identification information of the first scene and at least one of the above-mentioned association relationships, the parameters that need to be updated in the first model can be determined.

[0010] In a possible implementation manner, at least one association relationship is predefined or configured.

[0011] In this way, the signaling overhead for the association relationship between the interaction scenario and the parameters in the first model can be saved.

[0012] In a possible implementation manner, the method further includes: transmitting the first parameter.

[0013] In this way, other network elements can determine the parameters that need to be updated in the first model.

[0014] In a possible implementation manner, the method further includes: determining, based on identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model; and transmitting a message for instructing to obtain data of the first scene.

[0015] The data of the first scene can be used to update the first model, and the updated first model can be applied to the first scene.

[0016] In this way, it is possible to obtain data of the first scene so that the first model can be updated and trained using the data of the first scene.

[0017] In a possible implementation manner, the method further includes: acquiring first data, the first data being related to the first scene; and using the first data to update parameters that need to be updated in the first model to obtain an updated first model.

[0018] For example, the scenes related to the first scene may be: scenes with the same or similar data features as the first scene, or scenes of the same type as the first scene; or scenes with similar or identical scene features to the first scene. Similar means that they all meet the same conditions, for example, they are all below a certain threshold, or above a certain threshold, or similar means that the difference is below a certain threshold. The specific threshold, feature information, or one or more of the data features may be predefined or configured, and are not limited here.

[0019] Since the first data is related to the first scenario, the first model is updated using the first data, and the obtained updated first model can support the first scenario.

[0020] In a possible implementation method, the first data includes any one of the following: data of the first scene, data of the first scene and data of scenes related to the first scene, or data generated according to data features of the data of the first scene and the second data, the second data is data of scenes currently supported by the first model, and scenes related to the first scene belong to scenes currently supported by the first model.

[0021] Through one or more of the above-listed items, the present application supports completing the update training of the parameters that need to be updated in the first model, so that the updated first model can be applied to the first scenario.

[0022] A possible implementation method of obtaining the first data includes: obtaining the first data from a data collection network element, where the first data is data of a first scene; or, obtaining the first data from a data generation network element, where the first data is data generated according to data of the first scene and data features of second data; or, obtaining the first data from a data storage network element, where the first data is data of the first scene and data of scenes related to the first scene.

[0023] In this way, the present application supports that data for updating training of parameters that need to be updated in the first model can be obtained through multiple approaches.

[0024] In a possible implementation manner, the parameter that needs to be updated in the first model is the first parameter, or the parameter that needs to be updated in the first model is part or all of the parameters in the first model except the first parameter.

[0025] When the parameter that needs to be updated in the first model is the first parameter, the present application supports directly determining the parameter that needs to be updated in the first model; when the parameter that needs to be updated in the first model is part or all of the parameters in the first model except the first parameter, the present application supports indicating the parameters that need to be updated in the first model with lower signaling overhead.

[0026] In a possible implementation manner, the data characteristics of the second data include at least one of the following: signal to noise ratio (SNR), signal to interference plus noise ratio (SINR), reference signal received power (RSRP) or data distribution.

[0027] Among them, the above-mentioned data distribution may include one or more of the following situations: latent variable variational probability distribution of data after variational inference, probability distribution of data after Gaussian mixture model fitting, and potential distribution of data obtained through adversarial process learning).

[0028] In this way, through one or more parameters listed above, the present application supports generating a larger amount of data that can be used for updating the first model.

[0029] In a possible implementation manner, the method further includes: determining, based on the identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model.

[0030] When it is determined based on the identification information of the first scene that the first scene does not belong to the scenes currently supported by the first model, the present application can initiate the data collection process for the first scene, and use the data related to the first scene to update and train the parameters that need to be updated in the first model, so that the updated first model can support the first scene.

[0031] On the second aspect, a training method for an artificial intelligence model is provided, including: obtaining identification information of a first scene; determining scenes related to the first scene based on the identification information of the first scene, the scenes related to the first scene belonging to scenes currently supported by the first model; obtaining quantity information of data of scenes related to the first scene, the data of scenes related to the first scene being used for updating the first model, and the updated first model supports the first scene.

[0032] The solution described in the second aspect can be applied to a model training network element or a model deployment network element, for example, executed by a model training network element or a model deployment network element or a chip or module used for a model training network element or a model deployment network element.

[0033] For example, the scenes related to the first scene may be: scenes with the same or similar data features as the first scene, or scenes of the same type as the first scene; or scenes with similar or identical scene features to the first scene. Similar means that they all meet the same conditions, for example, they are all below a certain threshold, or above a certain threshold, or similar means that the difference is below a certain threshold. The specific threshold, feature information, or one or more of the data features may be predefined or configured, and are not limited here.

[0034] By determining the scenes related to the first scene based on the identification information of the first scene and obtaining the quantity information of the data of the scenes related to the first scene, this is helpful to reduce the amount of data that needs to be obtained from the first scene, thereby reducing the overhead of data collection and further reducing the overall overhead of model updating.

[0035] In a possible implementation, when the second aspect is applied to a model training network element, the method may further include: obtaining data of a scene related to the first scene. In this way, the model training network element can be trained based on the stored data of the scene related to the first scene, and the stored or obtained data can be effectively used, thereby reducing the amount of data to be obtained in the first scene and speeding up the model update.

[0036] In a possible implementation manner, the method further includes: determining, based on identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model; sending first indication information, the first indication information being used to indicate the amount of data to be acquired in the first scene, the quantity information of data of scenes related to the first scene being used to determine the amount of data to be acquired in the first scene, and the data of the first scene being used to update the first model.

[0037] In this way, the amount of data that needs to be acquired from the first scene may be indicated.

[0038] In a possible implementation manner, the above-mentioned determining the scene related to the first scene based on the identification information of the first scene includes: determining based on a predefined rule, or determining based on feature information of the scene.

[0039] In this way, scenes related to the first scene can be determined.

[0040] In a possible implementation manner, the first indication information is also used to instruct to obtain data of the first scene.

[0041] In this way, the data of the first scene can be obtained.

[0042] On the third aspect, a training method for an artificial intelligence model is provided, including: obtaining identification information of a first scene; determining, based on the identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model; transmitting first indication information, the first indication information being used to indicate obtaining data of the first scene, the data of the first scene being used for updating the first model, and the updated first model supporting the first scene.

[0043] The solution described in the third aspect can be applied to a model deployment network element, for example, executed by a model deployment network element or a chip or module used for a model deployment network element.

[0044] By determining whether it is necessary to obtain data for the first scene based on the identification information of the scene, this helps to avoid repeated data collection. When it is determined that it is necessary to obtain data for the first scene, an instruction can be given to obtain the data for the first scene, and the data for the first scene can be used to update and train the first model, so that the updated first model can support the first scene.

[0045] In addition, through the above scheme, the present application can avoid the situation where the first model is updated and trained when there is no need to update and train the first model, thereby effectively reducing the overall cost of model updating.

[0046] In a fourth aspect, a training method for an artificial intelligence model is provided, including: obtaining data of a first scene; generating data according to the data of the first scene and data features of the second data to obtain first data, wherein the second data is data of a scene currently supported by the first model, the first data is used to update the first model, and the updated first model supports the first scene; and sending the first data.

[0047] The solution described in the fourth aspect can be applied to a data generation network element, for example, executed by a data generation network element or a chip or module used for a data generation network element.

[0048] By generating the first data based on the data of the first scene and the data features of the second data, and without the need to obtain a large amount of data in the first scene, the present application can support the updated first model to be used not only for the first scene, but also for updating the previously supported scenes, which can effectively reduce the overhead of data collection, and thus can also effectively reduce the overall overhead of model updates.

[0049] In a possible implementation manner, the data feature of the second data includes at least one of the following: SNR, SINR, RSRP, or data distribution.

[0050] Among them, the above-mentioned data distribution may include one or more of the following situations: latent variable variational probability distribution of data after variational inference, probability distribution of data after Gaussian mixture model fitting, and potential distribution of data obtained through adversarial process learning).

[0051] In this way, through one or more parameters listed above, the present application supports generating a larger amount of data that can be used for updating the first model.

[0052] In a fifth aspect, a training method for an artificial intelligence model is provided, comprising: determining identification information of a first scene; and sending identification information of the first scene.

[0053] The scheme described in the fifth aspect can be applied to a data collection network element, for example, executed by a data collection network element or a chip or module used for a data collection network element. In a possible implementation manner, the above-mentioned determination of the identification information of the first scene may include: obtaining the identification information of the first scene or perceiving the identification information of the first scene, which is not limited to this.

[0054] By reporting the identification information of the first scene, other network elements or devices can determine whether it is necessary to start update training of the first model and whether it is necessary to start the data collection process for the first scene based on the identification information of the first scene, thereby effectively reducing the overall overhead of model updates.

[0055] In a possible implementation manner, the method further includes: receiving indication information, where the indication information is used to instruct to obtain data of the first scene; and obtaining the data of the first scene.

[0056] In this way, data of the first scene can be obtained according to the instruction.

[0057] In a possible implementation manner, the above indication information includes information about the amount of data to be acquired in the first scenario.

[0058] In this way, there is no need to obtain a large amount of data from the first scenario, which is conducive to reducing the overhead of data collection, thereby effectively supporting the reduction of the overall overhead of model updating.

[0059] In a possible implementation manner, the obtaining of the data of the first scene includes: obtaining the data of the first scene according to information on the amount of data to be obtained in the first scene.

[0060] In this way, the data of the first scene can be acquired according to the indicated amount of data to be acquired in the first scene.

[0061] In a sixth aspect, a training method for an artificial intelligence model is provided, comprising: obtaining indication information, the indication information being used to indicate obtaining data of a scene related to a first scene, the indication information comprising identification information of a scene related to the first scene; determining data of the scene related to the first scene based on the identification information of the scene related to the first scene and at least one association relationship, a first association relationship among the at least one association relationship being used to associate the scene related to the first scene with data of the scene related to the first scene; and transmitting data of the scene related to the first scene.

[0062] The solution described in the sixth aspect can be applied to a data storage network element, for example, executed by a data storage network element or a chip or module used for a data storage network element.

[0063] Optionally, the scene related to the first scene may be: a scene having the same or similar data features as the first scene, or a scene of the same type as the first scene; or a scene having similar or identical scene features to the first scene. Similar means that they all satisfy the same condition, for example, they are all below a certain threshold, or above a certain threshold, or similar means that the difference is below a certain threshold. The specific threshold, feature information, or one or more of the data features may be predefined or configured and are not limited here.

[0064] By sending data of scenes related to the first scene, the data of scenes related to the first scene can be used for updating and training the first model, which can effectively utilize the stored data and reduce the amount of data that needs to be obtained from the first scene, and can effectively reduce the overhead of data collection, thereby supporting the reduction of the overall overhead of model updating.

[0065] In a possible implementation manner, the at least one association relationship mentioned above is predefined or configured.

[0066] In this way, the signaling overhead for the association between the interaction scenes and the scene data can be saved.

[0067] In the seventh aspect, a communication device is provided, which includes a module or unit corresponding to the method / operation / step / action described in any one of the first to sixth aspects and any one of the aspects, and the module or unit can be a hardware circuit, or software, or a combination of hardware circuit and software.

[0068] In an eighth aspect, a communication device is provided, comprising a processor, wherein the processor is used to, by executing a computer program or instruction, or by a logic circuit, enable the communication device to execute the method described in the first aspect and any possibility of the first aspect; or enable the communication device to execute the method described in the second aspect and any possibility of the second aspect; or enable the communication device to execute the method described in the third aspect and any possibility of the third aspect; or enable the communication device to execute the method described in the fourth aspect and any possibility of the fourth aspect; or enable the communication device to execute the method described in the fifth aspect and any possibility of the fifth aspect; or enable the communication device to execute the method described in the sixth aspect and any possibility of the sixth aspect.

[0069] In a possible implementation manner, the communication device further includes a memory for storing the computer program or instruction.

[0070] In a possible implementation manner, the communication device further includes a communication interface, which is used to input and / or output signals.

[0071] In a ninth aspect, a communication device is provided, comprising a logic circuit and an input / output interface, the input / output interface being used to input and / or output signals, the logic circuit being used to execute the method described in the first aspect and any possibility of the first aspect; or, the logic circuit being used to execute the method described in the second aspect and any possibility of the second aspect; or, the logic circuit being used to execute the method described in the third aspect and any possibility of the third aspect; or, the logic circuit being used to execute the method described in the fourth aspect and any possibility of the fourth aspect; or, the logic circuit being used to execute the method described in the fifth aspect and any possibility of the fifth aspect; or, the logic circuit being used to execute the method described in the sixth aspect and any possibility of the sixth aspect.

[0072] In the tenth aspect, a computer-readable storage medium is provided, on which a computer program or instruction is stored. When the computer program or the instruction is run on a computer, the method described in the first aspect and any possibility of the first aspect is executed; or, the method described in the second aspect and any possibility of the second aspect is executed; or, the method described in the third aspect and any possibility of the third aspect is executed; or, the method described in the fourth aspect and any possibility of the fourth aspect is executed; or, the method described in the fifth aspect and any possibility of the fifth aspect is executed; or, the method described in the sixth aspect and any possibility of the sixth aspect is executed.

[0073] In the eleventh aspect, a computer program product is provided, comprising instructions, which, when executed on a computer, cause the method described in the first aspect and any possibility of the first aspect to be executed; or, cause the method described in the second aspect and any possibility of the second aspect to be executed; or, cause the method described in the third aspect and any possibility of the third aspect to be executed; or, cause the method described in the fourth aspect and any possibility of the fourth aspect to be executed; or, cause the method described in the fifth aspect and any possibility of the fifth aspect to be executed; or, cause the method described in the sixth aspect and any possibility of the sixth aspect to be executed.

[0074] In a twelfth aspect, a communication system is provided, including: one or more of a model training network element, a data collection network element, a model deployment network element, a data storage network element, or a data generation network element, etc. The network element is used to execute the method described in the corresponding aspect.

[0075] For example, the above-mentioned communication system includes a model training network element and a model deployment network element; or, the above-mentioned communication system includes a model deployment network element and a data collection network element; or, the above-mentioned communication system includes a model training network element, a data collection network element and a data generation network element; or, the above-mentioned communication system includes a model training network element, a model deployment network element, a data collection network element and a data storage network element, etc.; or, the above-mentioned communication system includes a model training network element, a model deployment network element, a data collection network element and a data generation network element, etc.; or, the above-mentioned communication system includes a model training network element, a data collection network element and a data generation network element, etc., and there is no limitation on this.

[0076] The description of the beneficial effects of the seventh to twelfth aspects may correspond to the description of the beneficial effects of the first to sixth aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a schematic diagram of an application framework 100 in a communication system in an embodiment of the present application.

[0078] Figure 2 It is a schematic diagram of an application framework 200 in a communication system in an embodiment of the present application.

[0079] Figure 3 is a schematic diagram of a communication system 300 applicable to an embodiment of the present application.

[0080] Figure 4 4 is a schematic diagram of a communication system 400 applicable to an embodiment of the present application.

[0081] Figure 5 is a schematic diagram of a neuron structure 500 .

[0082] Figure 6 It is a schematic diagram of a training method 600 for an AI model according to an embodiment of the present application.

[0083] Figure 7 It is a schematic diagram of the training method 700 of the AI ​​model according to an embodiment of the present application.

[0084] Figure 8 It is a schematic diagram of the training method 800 of the AI ​​model according to an embodiment of the present application.

[0085] Fig. 9 It is a schematic diagram of the training method 900 of the AI ​​model according to an embodiment of the present application.

[0086] Fig.10 It is a schematic block diagram of a communication device 1000 according to an embodiment of the present application.

[0087] Fig.11 It is a schematic block diagram of a communication device 1100 according to an embodiment of the present application. DETAILED DESCRIPTION

[0088] In order to facilitate understanding of the embodiments of the present application, the following points are first explained.

[0089] 1. In this application, unless otherwise specified, "plurality" means two or more.

[0090] 2. In each embodiment of the present application, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form a new embodiment according to their internal logical relationships.

[0091] 3. The various digital numbers involved in this application are only used for the convenience of description and are not used to limit the scope of protection of this application. The size of the serial numbers involved in this application does not mean the order of execution. The execution order of each process should be determined by its function and internal logic. For example, the terms "first", "second", "third", "fourth" and other various terminology labels (if any) in the specification and claims and drawings of this application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. Among them, the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than what is illustrated or described here.

[0092] At the same time, any embodiment or design described in the present application as "exemplary" or "for example" should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete manner for ease of understanding.

[0093] 4. The terms "comprise", "include", "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product or apparatus.

[0094] 5. In this application, "used to indicate" can be understood as "enable", and "enable" can include direct enablement and indirect enablement. When describing that a certain information is used to enable A, it can include that the information directly enables A or indirectly enables A, and it does not mean that the information must carry A.

[0095] The information enabled by the information is called information to be enabled. In the specific implementation process, there are many ways to enable the information to be enabled, such as but not limited to, the information to be enabled can be directly enabled, such as the information to be enabled itself or the index of the information to be enabled. The information to be enabled can also be indirectly enabled by enabling other information, wherein there is an association relationship between the other information and the information to be enabled. It is also possible to enable only a part of the information to be enabled, while the other parts of the information to be enabled are known or agreed in advance. For example, the enabling of specific information can also be achieved by means of the arrangement order of each piece of information agreed in advance (such as specified by the protocol), thereby reducing the enabling overhead to a certain extent. At the same time, the common parts of each piece of information can also be identified and enabled uniformly to reduce the enabling overhead caused by enabling the same information separately.

[0096] 6. In this application, "pre-configuration" may include pre-definition, such as protocol definition. Among them, "pre-definition" can be implemented by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in a device (for example, including each network element), and this application does not limit its specific implementation method.

[0097] 7. The term "storage" or "saving" as used in this application may refer to saving in one or more memories. The one or more memories may be provided separately or integrated in an encoder or decoder, a processor, or a communication device. The one or more memories may also be partially provided separately and partially integrated in a decoder, a processor, or a communication device. The type of memory may be any form of storage medium, without limitation.

[0098] 8. The “protocol” referred to in this application may refer to a standard protocol in the field of communications, such as the fourth generation (4G) network, the fifth generation (5G) network protocol, the new radio (NR) protocol, the 5.5G network protocol, the sixth generation (6 th The present application does not limit the network protocols used in future communication systems.

[0099] 9. The dotted arrows or boxes in the schematic diagrams of the drawings in the specification of this application represent optional steps or optional modules.

[0100] 10. In this application, unless otherwise specified, “ / ” indicates that the objects associated with each other are in an “or” relationship. For example, A / B can represent A or B. “And / or” in this application is only a description of the association relationship between the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural.

[0101] First, a communication system to which the embodiments of the present application are applicable is described.

[0102] The technical solution provided in the present application can be applied to various communication systems, such as: 5G or NR system, long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD) system, wireless local area network (WLAN) system, satellite communication system, future communication system, such as 6G mobile communication system, or a fusion system of multiple systems, etc. The technical solution provided in the present application can also be applied to device to device (D2D) communication, vehicle to everything (V2X) communication, machine to machine (M2M) communication, machine type communication (MTC), and Internet of things (IoT) communication system or other communication systems.

[0103] A network element in a communication system can send a signal to another network element or receive a signal from another network element. The signal may include information, signaling, or data, etc. The network element may also be replaced by an entity, a network entity, a device, a communication device, a communication module, a node, a communication node, etc. The embodiment of the present application is described by taking a network element as an example. For example, a communication system may include at least one terminal device and at least one network device. The network device may send a downlink signal to the terminal device, and / or the terminal device may send an uplink signal to the network device. It is understandable that the terminal device in the embodiment of the present application may be replaced by a first network element, and the network device may be replaced by a second network element, and the two perform the corresponding communication method in the embodiment of the present application.

[0104] In an embodiment of the present application, the terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device.

[0105] The terminal device may be a device that provides voice / data, such as a handheld device or a vehicle-mounted device with a wireless connection function. At present, some examples of terminals are: mobile phones, tablet computers, laptop computers, PDAs, mobile internet devices (MID), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self driving, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, cellular phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication functions, computing devices or other processing devices connected to wireless modems, wearable devices, terminal devices in 5G networks or future evolved public land mobile communication networks (public land mobile The embodiments of the present application do not limit this.

[0106] As an example but not limitation, in the embodiments of the present application, the terminal device may also be a wearable device. Wearable devices may also be referred to as wearable smart devices, which are a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing, and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also powerful functions achieved through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include full-featured, large-sized, and fully or partially independent of smartphones, such as smart watches or smart glasses, as well as devices that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various types of smart bracelets and smart jewelry for vital sign monitoring.

[0107] In the embodiment of the present application, the device for realizing the function of the terminal device may be a terminal device, or a device capable of supporting the terminal device to realize the function, such as a chip system, which may be installed in the terminal device or used in combination with the terminal device. In the embodiment of the present application, the chip system may be composed of a chip, or may include a chip and other discrete devices. In the embodiment of the present application, only the device for realizing the function of the terminal device is used as an example for explanation, and the scheme of the embodiment of the present application is not limited.

[0108] The network device in the embodiment of the present application may be a device for communicating with a terminal device, and the network device may also be referred to as an access network device or a wireless access network device, such as a base station. The network device in the embodiment of the present application may refer to a wireless access network (RAN) node (or device) that connects a terminal device to a wireless network. The base station can broadly cover various names as follows, or be replaced with the following names, such as: NodeB, evolved NodeB (eNB), next generation NodeB (gNB), relay station, access point, transmission point (TRP), transmission point (TP), master station, auxiliary station, multi-standard wireless (motor slide retainer, MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), radio unit (RU), positioning node, etc. The base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. The base station may also refer to a communication module, modem or chip used to be set in the aforementioned equipment or device. The base station may also be a mobile switching center and a device that performs the base station function in D2D, V2X, and M2M communications, a network side device in a 6G network, and a device that performs the base station function in future communication systems. The base station can support networks with the same or different access technologies. Optionally, the RAN node may also be a server, a wearable device, a vehicle or an on-board device, etc. For example, the access network device in the vehicle to everything (V2X) technology may be a road side unit (RSU). The embodiments of the present application do not limit the specific technology and specific device form adopted by the network equipment.

[0109] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move based on the location of the mobile base station. In other examples, a helicopter or drone can be configured to act as a device that communicates with another base station.

[0110] In some deployments, the network device mentioned in the embodiments of the present application may be a device including a CU, or a DU, or a device including a CU and a DU, or a control plane CU node (central unit control plane (central unit-control plane, CU-CP)) and a user plane CU node (central unit user plane (central unit-user plane, CU-UP)) and a DU node. For example, the network device may include a gNB-CU-CP, a gNB-CU-UP, and a gNB-DU.

[0111] In some deployments, multiple RAN nodes collaborate to assist the terminal in achieving wireless access, and different RAN nodes implement part of the functions of the base station. For example, the RAN node can be a CU, DU, CU-CP, CU-UP, or RU. The CU and DU can be set separately, or can also be included in the same network element, such as a BBU. The RU can be included in a radio frequency device or a radio frequency unit, such as an RRU, AAU, or RRH.

[0112] The RAN node may support one or more types of fronthaul interfaces, and different fronthaul interfaces correspond to DUs and RUs with different functions. If the fronthaul interface between the DU and the RU is a common public radio interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and the RU is another interface, relative to the CPRI, part of the downlink and / or uplink baseband functions, such as, for downlink, one or more of precoding, digital beamforming (BF), or inverse fast Fourier transform (IFFT) / adding cyclic prefix (CP), are moved from the DU to the RU for implementation, and for uplink, one or more of digital beamforming (BF), or fast Fourier transform (FFT) / removing cyclic prefix (CP), are moved from the DU to the RU for implementation. In a possible implementation, the interface may be an enhanced common public radio interface (eCPRI). In the eCPRI architecture, the division between DU and RU is different, corresponding to different types (category, Cat) of eCPRI, such as eCPRI CatA, B, C, D, E, F.

[0113] Taking eCPRI Cat A as an example, for downlink transmission, with layer mapping as the division, the DU is configured to implement layer mapping and one or more functions before it (i.e., one or more functions of coding, rate matching, scrambling, modulation, and layer mapping), while other functions after layer mapping (e.g., resource element (RE) mapping, BF), or one or more functions of IFFT / CP) are moved to the RU for implementation. For uplink transmission, with RE demapping as the division, the DU is configured to implement demapping and one or more functions before it (i.e., decoding, rate dematching, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and one or more functions of RE demapping), while other functions after demapping (e.g., one or more functions of digital BF or fast Fourier transform (FFT) / de-CP) are moved to the RU for implementation. It can be understood that for the functional description of DU and RU corresponding to various types of eCPRI, please refer to the eCPRI protocol, which will not be repeated here.

[0114] In one possible design, the processing unit for implementing the baseband function in the BBU is called a baseband high layer (BBH) unit, and the processing unit for implementing the baseband function in the RRU / AAU / RRH is called a baseband low layer (BBL) unit.

[0115] In different systems, CU (or CU-CP and CU-UP), DU or RU may also have different names, but those skilled in the art can understand their meanings. For example, in the ORAN system, CU may also be called O-CU (open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, CU-UP may also be called O-CU-UP, and RU may also be called O-RU. Any unit in the CU (or CU-CP, CU-UP), DU and RU in this application may be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0116] In the embodiments of the present application, the device for realizing the function of the network device may be a network device; or it may be a device capable of supporting the network device to realize the function, such as a chip system, a hardware circuit, a software module, or a hardware circuit plus a software module. The device may be installed in the network device or used in combination with the network device. In the embodiments of the present application, only the device for realizing the function of the network device is a network device as an example for explanation, and the scheme of the embodiments of the present application is not limited.

[0117] The network device and / or terminal device can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; it can also be deployed on the water surface; it can also be deployed on aircraft, balloons and satellites in the air. The scenarios in which the network device and the terminal device are located are not limited in the embodiments of the present application. In addition, the terminal device and the network device can be hardware devices, or they can be software functions running on dedicated hardware, software functions running on general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform), or entities including dedicated or general-purpose hardware devices and software functions. The present application does not limit the specific forms of the terminal device and the network device.

[0118] In wireless communication networks, such as mobile communication networks, the services supported by the network are becoming more and more diverse, so the demands that need to be met are becoming more and more diverse. For example, the network needs to be able to support ultra-high speed, ultra-low latency, and / or ultra-large connections. This feature makes network planning, network configuration, and / or resource scheduling more and more complicated. In addition, as the network becomes more and more powerful, such as supporting higher and higher spectrum, supporting high-order multiple input multiple output (MIMO) technology, supporting beamforming, and / or supporting new technologies such as beam management, network energy saving has become a hot research topic. These new requirements, new scenarios, and new features have brought unprecedented challenges to network planning, operation and maintenance, and efficient operation. In order to meet this challenge, artificial intelligence technology can be introduced into wireless communication networks to achieve network intelligence.

[0119] In order to support AI technology in wireless networks, AI nodes may also be introduced into the network.

[0120] Optionally, the AI ​​node can be deployed in one or more of the following locations in the communication system: access network equipment, terminal equipment, or core network equipment, etc., or the AI ​​node can also be deployed separately, for example, deployed in a location other than any of the above devices, such as a host or cloud server in an over the top (OTT) system. The AI ​​node can communicate with other devices in the communication system, and the other devices can be, for example, one or more of the following: network equipment, terminal equipment, or network elements of the core network, etc.

[0121] It is understandable that the embodiments of the present application do not limit the number of AI nodes. For example, when there are multiple AI nodes, the multiple AI nodes can be divided based on functions, such as different AI nodes are responsible for different functions.

[0122] It can also be understood that AI nodes can be independent devices, or integrated into the same device to implement different functions, or network elements in hardware devices, or software functions running on dedicated hardware, or virtualized functions instantiated on a platform (e.g., a cloud platform). This application does not limit the specific form of AI nodes. Among them, AI nodes can be AI network elements or AI modules.

[0123] Figure 1 FIG. 1 is a schematic diagram of an application framework 100 in a communication system according to an embodiment of the present application. Figure 1 As shown, network elements in the communication system are connected through interfaces (e.g., NG, Xn) or air interfaces. One or more AI modules are set in one or more devices in these network element nodes, such as core network equipment, access network nodes (RAN nodes), terminals, or operation and maintenance management (OAM) (for clarity, Figure 1 Only one is shown in the figure). The access network node can be a separate RAN node or can include multiple RAN nodes, for example, including CU and DU. CU and / or DU can also be provided with one or more AI modules. Optionally, CU can also be split into CU-CP and CU-UP. One or more AI models are provided in CU-CP and / or CU-UP.

[0124] The AI ​​module is used to implement the corresponding AI function. The AI ​​modules deployed in different network elements may be the same or different. The model of the AI ​​module can implement different functions according to different parameter configurations. The model of the AI ​​module can be configured based on one or more of the following parameters: structural parameters (such as the number of neural network layers, the width of the neural network, the connection relationship between layers, the weight of the neuron, the activation function of the neuron, or at least one of the biases in the activation function), input parameters (such as the type of input parameters and / or the dimension of input parameters), or output parameters (such as the type of output parameters and / or the dimension of output parameters). Among them, the bias in the activation function can also be called the bias of the neural network.

[0125] An AI module can have one or more models. A model can be inferred to obtain an output, which includes one parameter or multiple parameters. The learning process, training process, or inference process of different models can be deployed in different nodes or devices, or can be deployed in the same node or device.

[0126] Figure 2 FIG. 2 is a schematic diagram of an application framework 200 in a communication system according to an embodiment of the present application. Figure 2As shown, the communication system includes a RAN intelligent controller (RAN intelligent controller, RIC). For example, the RIC can be Figure 1 The AI ​​modules 117 and 118 shown in the figure are used to implement AI-related functions. RIC includes near-real time RIC (near-real time RIC, near-RT RIC) and non-real time RIC (non-real time RIC, Non-RT RIC). Among them, non-real time RIC mainly processes non-real time information, such as data that is not sensitive to delay, and the delay of this data can be in the order of seconds. Real-time RIC mainly processes near real-time information, such as data that is relatively sensitive to delay, and the delay of this data is in the order of tens of milliseconds.

[0127] Near real-time RIC is used for model training and reasoning. For example, it is used to train an AI model and use the AI ​​model for reasoning. Near real-time RIC can obtain information on the network side and / or the terminal side from RAN nodes (such as CU, CU-CP, CU-UP, DU and / or RU) and / or terminals. This information can be used as training data or reasoning data. Optionally, near real-time RIC can submit the reasoning results to the RAN node and / or terminal. Optionally, the reasoning results can be exchanged between the CU and DU, and / or between the DU and RU. For example, the near real-time RIC submits the reasoning results to the DU, and the DU sends it to the RU.

[0128] Non-real-time RIC is also used for model training and reasoning. For example, it is used to train an AI model and use the model for reasoning. Non-real-time RIC can obtain information on the network side and / or the terminal side from RAN nodes (such as CU, CU-CP, CU-UP, DU and / or RU) and / or terminals. This information can be used as training data or reasoning data, and the reasoning results can be submitted to the RAN node and / or the terminal. Optionally, the reasoning results can be exchanged between the CU and the DU, and / or between the DU and the RU. For example, the non-real-time RIC submits the reasoning results to the DU, and the DU sends it to the RU.

[0129] The near real-time RIC and the non-real-time RIC may also be separately set as a network element. Optionally, the near real-time RIC and the non-real-time RIC may also be part of other devices, for example, the near real-time RIC is set in the RAN node (for example, in the CU, DU), and the non-real-time RIC is set in the OAM, the cloud server, the core network device, or other network devices.

[0130] Figure 3 3 is a schematic diagram of a communication system 300 applicable to an embodiment of the present application. Figure 3As shown, the communication system 300 may include at least one network device, for example, the network device 110; the communication system 100 may also include at least one terminal device, for example, the terminal device 120 and the terminal device 130. The network device 110 and the terminal device (such as the terminal device 120 and the terminal device 130) may communicate via a wireless link. The communication devices in the communication system, for example, the network device 110 and the terminal device 120, may communicate via a multi-antenna technology.

[0131] Figure 4 4 is a schematic diagram of a communication system 400 applicable to an embodiment of the present application. Compared with the communication system 300, the communication system 400 further includes an AI network element 140. The AI ​​network element 140 is used to perform AI-related operations, such as building a training data set or training an AI model.

[0132] In one possible implementation, the network device 110 may send data related to the training of the AI ​​model to the AI ​​network element 140, which constructs a training data set and trains the AI ​​model. For example, the data related to the training of the AI ​​model may include data reported by the terminal device. The AI ​​network element 140 may send the results of operations related to the AI ​​model to the network device 110, and forward them to the terminal device through the network device 110. For example, the results of operations related to the AI ​​model may include at least one of the following: an AI model that has completed training, an evaluation result or a test result of the model, etc. Exemplarily, a part of the AI ​​model that has completed training may be deployed on the network device 110, and another part may be deployed on the terminal device. Alternatively, the AI ​​model that has completed training may be deployed on the network device 110. Alternatively, the AI ​​model that has completed training may be deployed on the terminal device.

[0133] It should be understood that Figure 4 Only the example of the AI ​​network element 140 being directly connected to the network device 110 is used for illustration. In other scenarios, the AI ​​network element 140 may also be connected to the terminal device. Alternatively, the AI ​​network element 140 may be connected to the network device 110 and the terminal device at the same time. Alternatively, the AI ​​network element 140 may also be connected to the network device 110 through a third-party network element. The embodiment of the present application does not limit the connection relationship between the AI ​​network element and other network elements.

[0134] The AI ​​network element 140 may also be provided as a module in a network device and / or a terminal device, for example, Figure 3 In the network device 110 or terminal device shown.

[0135] It should be noted that Figure 3 and Figure 4This is a simplified schematic diagram for ease of understanding. For example, the communication system may also include other devices, such as wireless relay devices and / or wireless backhaul devices. Figure 3 and Figure 4 In practical applications, the communication system may include multiple network devices (such as network device 110 and network device 150 ( Figure 3 The embodiment of the present application does not limit the number of network devices and terminal devices included in the communication system.

[0136] Next, some technical concepts involved in this application are briefly described.

[0137] Machine learning (ML): ML is an important technical approach to realize AI. ML can be divided into supervised learning, unsupervised learning, and reinforcement learning.

[0138] Supervised learning uses ML algorithms to learn the mapping relationship from sample values ​​to sample labels based on the collected sample values ​​and sample labels, and uses ML models to express the learned mapping relationship. The process of training ML models is the process of learning this mapping relationship. For example, in signal detection, the received signal containing noise is the sample, and the real constellation point corresponding to the signal is the label. ML expects to learn the mapping relationship between samples and labels through training, that is, to make the ML model learn a signal detector. During training, the model parameters are optimized by calculating the error between the model's predicted value and the real label. Once the mapping relationship is learned, the learned mapping can be used to predict the sample label of each new sample. The mapping relationship learned by supervised learning can include linear mapping and nonlinear mapping. According to the type of label, the learning task can be divided into classification task and regression task.

[0139] Unsupervised learning is based only on the collected sample values, using algorithms to discover the inherent patterns of samples. There is a type of algorithm in unsupervised learning that uses the sample itself as a supervisory signal, that is, the model learns the mapping relationship from sample to sample, which is called self-supervised learning. During training, the model parameters are optimized by calculating the error between the model's predicted value and the sample itself. Self-supervised learning can be used in applications such as signal compression and decompression recovery. Common algorithms include autoencoders and adversarial generative networks.

[0140] Reinforcement learning is different from supervised learning. It is a type of algorithm that learns problem-solving strategies by interacting with the environment. Unlike supervised and unsupervised learning, reinforcement learning problems do not have clear "correct" action label data. The algorithm needs to interact with the environment to obtain reward signals from the environment, and then adjust the decision-making actions to obtain a larger reward signal value. For example, in downlink power control, the reinforcement learning model adjusts the downlink transmission power of each user according to the total system throughput fed back by the wireless network, and then expects to obtain a higher system throughput. The goal of reinforcement learning is also to learn the mapping relationship between the state of the environment and the optimal decision action. However, because the label of the "correct action" cannot be obtained in advance, the network cannot be optimized by calculating the error between the action and the "correct action". Reinforcement learning training is achieved through iterative interaction with the environment.

[0141] Deep neural network (DNN) is a specific implementation form of ML. According to the universal approximation theorem, neural networks can theoretically approximate any continuous function, which enables neural networks to have the ability to learn any mapping. Traditional communication systems require rich expert knowledge to design communication modules, while deep learning communication systems based on DNN can automatically discover implicit pattern structures from a large number of data sets, establish mapping relationships between data, and achieve performance that is superior to traditional modeling methods. For a description of DDN, see Figure 5 .

[0142] Figure 5 is a schematic diagram of a neuron structure 500. Figure 5 As shown, assuming that the input of the neuron is x = [x0,…,x n ], and the weight corresponding to the input is w=[w0,…,w n ], the weighted sum bias is b, and the form of nonlinear function can be diversified. An example is the max{0,x} maximum function. Then the effect of a neuron's execution can be

[0143] According to the way the network is constructed, DNN can be divided into feedforward neural network (FNN), convolutional neural network (CNN) and recurrent neural network (RNN).

[0144] CNN is a neural network that is specifically designed to process data with a grid-like structure. For example, time series data (discrete sampling on the time axis) and image data (discrete sampling on two dimensions) can be considered to be data with a grid-like structure. CNN does not use all the input information for calculations at once, but uses a fixed-size window to intercept part of the information for convolution operations, which greatly reduces the amount of calculation of model parameters. In addition, depending on the type of information intercepted by the window (for example, people and objects in a picture are different types of information), each window can use different convolution kernel operations, which enables CNN to better extract the features of the input data.

[0145] RNN is a type of DNN network that uses feedback time series information. Its input includes the new input value at the current moment and its own output value at the previous moment. RNN is suitable for obtaining sequence features that are correlated in time, and is particularly suitable for applications such as speech recognition and channel coding.

[0146] The above-mentioned FNN, CNN, and RNN are common neural network structures, which are all constructed based on neurons. As mentioned above, each neuron performs a weighted sum operation on its input values, and the weighted sum result generates an output through a nonlinear function. We call the weights of the weighted sum operation of neurons in the neural network and the nonlinear function the parameters of the neural network. Take the neuron with max{0,x} as the nonlinear function as an example, The parameters of the operated neuron are weights w = [w0,…,w n ]The weighted sum is biased by b, and the nonlinear function max{0,x}. The parameters of all neurons in a neural network constitute the parameters of the neural network.

[0147] AI model: AI model is an algorithm or computer program that can realize AI functions. AI model represents the mapping relationship between the input and output of the model. The type of AI model can be neural network, linear regression model, decision tree model, support vector machine (SVM), Bayesian network, Q learning model or other ML model.

[0148] Two-end model: The two-end model can also be called a bilateral model, a collaborative model, a dual model, or a two-side model. The two-end model refers to a model composed of multiple sub-models. The multiple sub-models that constitute the model need to match each other. The multiple sub-models can be deployed in different nodes.

[0149] The AI ​​model of the embodiment of the present application can be a dual-end model or a single-end model, which can be deployed on terminal devices and / or network devices.

[0150] Model application or reasoning: Use the trained model to solve practical problems.

[0151] AI function: does not specifically refer to a specific AI model, but refers to a class of AI models with the same function. For example, the AI ​​function can be channel state information (CSI) feedback, CSI prediction, beam prediction, positioning, channel estimation, etc., that is, the AI ​​function can correspond to multiple AI models.

[0152] It is understood that the implementation of the AI ​​model or AI function can be a hardware circuit, or software, or a combination of software and hardware, without limitation. Non-limiting examples of software include: program code, program, subroutine, instruction, instruction set, code, code segment, software module, application, or software application, etc.

[0153] In an embodiment of the present application, indications include direct indications (also called explicit indications) and implicit indications. Specifically, direct indication of information A refers to including the information A; implicit indication of information A refers to indicating information A through the correspondence between information A and information B and direct indication of information B. Specifically, the correspondence between information A and information B may be predefined, pre-stored, pre-burned, or pre-configured. In addition, in an embodiment of the present application, information C is used to determine information D, which includes both information D being determined based only on information C and information D being determined based on information C and other information. In addition, information C is used to determine information D, and it may also be indirectly determined, for example, information D is determined based on information E, and information E is determined based on information C.

[0154] In addition, in each embodiment of the present application, "network element A (which may be device A) sends information A to network element B (which may be device B)" can be understood as the destination end of the information A or the intermediate network element in the transmission path between the destination end and the network element B, which may include directly or indirectly sending information to network element B. "Network element B receives information A from network element A" can be understood as the source end of the information A or the intermediate network element in the transmission path between the source end and the network element A, which may include directly or indirectly receiving information from network element A. The information may be processed as necessary between the source end and the destination end of the information transmission, such as format changes, etc., but the destination end can understand the valid information from the source end. Similar expressions in the present application can be understood similarly and will not be repeated here.

[0155] As described in the background technology section, when the AI ​​model switches from one scene to another, in order to enable the AI ​​model to be applied to the newly switched scene, the data of the newly switched scene can be obtained, and the AI ​​model can be updated using the data of the newly switched scene to obtain an updated AI model, which can be applied to the newly switched scene. However, the above scheme may result in a large overhead for model updates. For example, all parameters of the AI ​​model are updated, which will incur a large update overhead; for another example, a large amount of data in the newly switched scene is obtained, which will incur a large data storage overhead, thereby incurring a large overhead for model updates.

[0156] In view of this, the present application provides an AI model training method and communication device, which can effectively reduce the overall overhead of model updating.

[0157] For ease of understanding and explanation, the following description takes the interaction between network element 1 and network element 2 as an example. Network element 1 and network element 2 may correspond to different entities or to the same entity. For example, network element 1 corresponds to entity 1 (which may be a terminal device, a network device, or a core network device), and network element 2 corresponds to entity 2 (which may be a terminal device, a network device, or a core network device); for another example, network element 1 and network element 2 correspond to entity 1, etc., without limitation.

[0158] In a possible implementation, network element 1 may be a model deployment network element or a model training network element, and network element 2 may include at least one of a scene recognition network element or a data collection network element (it may also be a data provision network element, or other names, which are not limited to this). For the interaction between network element 1 and network element 2, please refer to Figure 6 .

[0159] Figure 6 600 is a schematic diagram of a training method 600 of an AI model according to an embodiment of the present application. Figure 6 As shown, the method 600 includes:

[0160] S601. Network element 1 obtains identification information of a first scene.

[0161] For example, Figure 6 As shown in the figure above, network element 1 is a model training network element, and network element 1 obtains identification information of the first scene from network element 2. In other words, after network element 2 determines or obtains the identification information of the first scene, network element 2 sends the identification information of the first scene to network element 1. At this time, network element 2 can also be called a scene recognition network element or a data collection network element.

[0162] For example, Figure 6As shown in the figure above, network element 1 is a model training network element, and network element 1 obtains the identification information of the first scene from the model deployment network element. In other words, after network element 2 determines or obtains the identification information of the first scene, network element 2 sends the identification information of the first scene to the model deployment network element, and the model deployment network element sends the identification information of the first scene to network element 1.

[0163] For example, Figure 6 As shown in the figure below, network element 1 is a model deployment network element, and network element 1 obtains identification information of the first scenario from network element 2. In other words, after network element 2 determines or obtains the identification information of the first scenario, network element 2 sends the identification information of the first scenario to network element 1.

[0164] In one possible implementation, the identification information of the first scene may include identification information of the first scene. For example, an identification may be configured for each scene, and the identification is associated with the characteristics or type of the scene. Exemplarily, for indoor scenes, identification 1 is configured; for outdoor scenes, identification 2 is configured; for factory scenes, identification 3 is configured; for low power consumption scenes, identification 4 is configured; for high power consumption scenes, identification 5 is configured; for scenes with better channel quality, identification 6 is configured; for scenes with poor channel quality, identification 7 is configured; for low latency scenes, identification 8 is configured; for high latency scenes, identification 9 is configured, and so on. Among them, the identification information of these scenes can be sensed, acquired and determined by sensor devices (for example, network devices, terminal devices, and cameras, etc.). For example, the network device or terminal device can determine the characteristic information of the current scene through the reference signal measurement result, and determine the identification information of the current scene based on the association between the characteristic information of the scene and the identification information of the scene, and report the identification information of the current scene.

[0165] In another possible implementation, the identification information of the first scene may include feature information of the first scene. For example, each scene may have its own unique scene feature.

[0166] Exemplarily, for indoor scenes, it includes feature information 1, and feature information 1 includes one or more of the following features: the delay spread of the reference signal channel measurement result in the scene is lower than a first threshold (the first threshold is preconfigured and predefined), the Doppler frequency deviation of the reference signal channel measurement result in the scene is lower than a second threshold (the second threshold is preconfigured and predefined), the corresponding network device indicates that the current scene is located indoors, the corresponding terminal device determines that the current scene is located indoors based on its own positioning result, and the current scene is perceived as an indoor scene through a sensor device (such as a camera).

[0167] Exemplarily, for outdoor scenes, it includes feature information 2, and feature information 2 includes one or more of the following features: the delay spread of the reference signal channel measurement result in the scene is higher than a third threshold (the third threshold is preconfigured and predefined), the Doppler frequency deviation of the reference signal channel measurement result in the scene is higher than a fourth threshold (the fourth threshold is preconfigured and predefined), the corresponding network device indicates that the current scene is located outdoors, the corresponding terminal device determines that the current scene is located outdoors based on its own positioning results, and the current scene is perceived as an outdoor scene through a sensor device (such as a camera).

[0168] Exemplarily, for a factory scenario, it includes feature information 3, which includes one or more of the following features: the delay spread of the reference signal channel measurement result in the scenario is lower than the fifth threshold (the fifth threshold is preconfigured and predefined), the Doppler frequency deviation of the reference signal channel measurement result in the scenario is lower than the sixth threshold (the sixth threshold is preconfigured and predefined), the corresponding network device indicates that the current scene is located in the factory, the corresponding terminal device determines that the current scene is located in the factory based on its own positioning result, and the current scene is perceived as a factory scene through sensor equipment (such as a camera).

[0169] Exemplarily, for a scenario with better channel quality, it includes feature information 4, and feature information 4 includes one or more of the following features: a signal to interference plus noise ratio (SINR) of a reference signal channel measurement result in the scenario is higher than a seventh threshold (the seventh threshold is preconfigured and predefined), and an interference level of a reference signal channel measurement result in the scenario is lower than an eighth threshold (the eighth threshold is preconfigured and predefined).

[0170] Exemplarily, for a scenario with poor channel quality, it includes feature information 5, and feature information 5 includes one or more of the following features: the SINR of the reference signal channel measurement result in the scenario is lower than the ninth threshold (the ninth threshold is preconfigured, predefined), and the interference level of the reference signal channel measurement result in the scenario is higher than the tenth threshold (the tenth threshold is preconfigured, predefined). The above ten thresholds may be different or partially the same, and this is not limited.

[0171] The characteristic information of these scenes can be sensed, acquired and reported by sensor devices, and the network element 1 determines the identification information of the current scene according to the association between the characteristic information of the scene and the identification information of the scene.

[0172] In the above description, network element 2 can be a network element or entity with the function of sensing and obtaining identification information (such as identification information or feature information) of the current scene, for example, a network device or a terminal device or a sensing device different from the network device or the terminal device, etc. The sensing device can communicate with at least one of the network device or the terminal device.

[0173] S602. Network element 1 determines a first parameter in the first model according to identification information of the first scenario, where the first parameter is used to determine a parameter that needs to be updated in the first model.

[0174] The first model can support (can also be: the first model currently supports) multiple scenarios, for example, the first model can support the second scenario, the third scenario and other scenarios. For the convenience of description, the scenarios supported by the first model can be referred to as historical scenarios, such as the second scenario and the third scenario can be collectively referred to as historical scenarios. Among them, the number of historical scenes supported by the first model is not limited, it can be one or more, and the number of historical scenes supported by the first scenario can be updated or changed (for example, deleted or increased, which is not limited). In addition, the historical scene can also be a scenario for the functional service corresponding to the first model. For example, the scenario to which the first model can be specifically applied can be a historical scene.

[0175] In a possible implementation, the first scenario does not belong to the scenarios currently supported by the first model. In this way, the parameters that need to be updated in the first model can be updated so that the first model obtained after the update can support the first scenario.

[0176] Another possible implementation method is that the first scene belongs to the scene currently supported by the first model, such as the first scene belongs to a scene in the historical scene. The parameters that need to be updated in the first model can also be updated to make the updated first model more powerful.

[0177] Network element 1 may determine the parameters that need to be updated in the first model according to the identification information of the first scenario, and the parameters that need to be updated in the first model may not include all parameters of the first model, that is, include all or part of the parameters of the first model. The first parameters can be used to determine the parameters that need to be updated in the first model.

[0178] In a possible implementation, the parameter to be updated in the first model is the first parameter, as shown in Table 1. When the parameter to be updated in the first model is the first parameter, the present application supports directly determining the parameter to be updated in the first model. The contents described in Table 1 are only understood as examples and are not intended to be final limitations.

[0179] Table 1

[0180]

[0181]

[0182] As shown in Table 1, three scenarios and three parameters (taking the first model including three parameters as an example) are used for description. For example, scenario 1 corresponds to parameter 1, scenario 2 corresponds to parameter 2, and scenario 3 corresponds to parameter 3. Exemplarily, for scenario 1, parameter 1 in the first model is a parameter that needs to be updated; for scenario 2, parameter 2 in the first model is a parameter that needs to be updated; for scenario 3, parameter 3 in the first model is a parameter that needs to be updated. The present application supports establishing an association relationship between a scenario and a parameter in the first model, and storing or configuring or predefining the association relationship between the scenario and the parameter in the first model in network element 1, so that the signaling overhead for the association relationship between the interactive scenario and the parameter in the first model can be saved.

[0183] In summary, by setting association relationships between multiple different scenes and parameters in the first model, and determining the first parameter associated with the first scene based on the identification information of the first scene and at least one of the above-mentioned association relationships, the parameters that need to be updated in the first model can be determined.

[0184] Among them, the above parameters 1, 2 and 3 can be understood as Figure 5 Parameters such as w and b described in .

[0185] In combination with the content shown in Table 1, the identification information of the first scene can be used to indicate scene 1, and the first parameter can be parameter 1.

[0186] In another possible implementation, the parameters that need to be updated in the first model are some or all of the parameters in the first model except the first parameter, as shown in Table 2. When the parameters that need to be updated in the first model are some or all of the parameters in the first model except the first parameter, the present application supports indicating the parameters that need to be updated in the first model with lower signaling overhead. The contents described in Table 2 are understood only as examples and are not intended to be final limitations.

[0187] Table 2

[0188] Scenario parameter Scenario 1 Parameter 2 and Parameter 3 Scenario 2 Parameter 1 and Parameter 3 Scene 3 Parameter 1 and Parameter 2

[0189] As shown in Table 2, three scenarios and three parameters (taking the first model including three parameters as an example) are used for description. For example, scenario 1 corresponds to parameter 2 and parameter 3, scenario 2 corresponds to parameter 1 and parameter 3, and scenario 3 corresponds to parameter 1 and parameter 2. Exemplarily, for scenario 1, parameter 1 in the first model is a parameter that does not need to be updated; for scenario 2, parameter 2 in the first model is a parameter that does not need to be updated; for scenario 3, parameter 3 in the first model is a parameter that does not need to be updated. The present application supports establishing an association relationship between a scenario and a parameter in the first model, and storing or configuring or predefining the association relationship between the scenario and the parameter in the first model in network element 1. In this way, network element 1 can determine the first parameter based on the association relationship between the scenario and the parameter in the first model and the identification information of the first scenario.

[0190] In combination with the content shown in Table 2, the identification information of the first scene can be used to indicate scene 1, and the first parameter can be parameter 1.

[0191] In summary, network element 1 determines the parameters that need to be updated in the first model based on the identification information of the scene, which helps to update the model of the parameters that need to be updated in the first model, and there is no need to update all the parameters of the first model. In this way, the overall overhead of model updating can be effectively reduced.

[0192] Combined with the following Figure 7 and Figure 8 right Figure 6 The method shown is further described.

[0193] For ease of understanding and explanation, the following description is made by taking the interaction between network element 1, network element 2 and network element 3 as an example. Among them, network element 1, network element 2 and network element 3 may correspond to different entities, that is, they are different entities, or they may correspond to the same entity or multiple entities, that is, all or part of the three network elements of network element 1, network element 2 or network element 3 belong to the same entity. For example, network element 1 corresponds to entity 1 (which may be a terminal device, a network device or a core network device), network element 2 corresponds to entity 2 (which may be a terminal device, a network device or a core network device), and network element 3 corresponds to entity 3 (which may be a terminal device, a network device or a core network device); for another example, network element 1 and network element 2 correspond to entity 1, and network element 3 corresponds to entity 2; for another example, network element 1 and network element 3 correspond to entity 1, and network element 2 corresponds to entity 2, etc., without limitation. Among them, network element 3 can be functionally referred to as a data storage / generation network element. When it has a data generation function, it may include a data generator, which is referred to as a data generation network element. When it is used for data storage, it may also be referred to as a data storage network element.

[0194] Figure 7 700 is an interactive flow chart of the training method 700 of the AI ​​model of the embodiment of the present application. Figure 7 As shown, taking network element 1 as a model for deploying network elements as an example, method 700 includes:

[0195] S701. Network element 2 determines identification information of a first scene.

[0196] Among them, the network element 2 can be a network element or entity with the function of sensing and obtaining the identification information (such as identification information or feature information) of the current scene, for example, a network device or a terminal device or a sensing device different from the network device or the terminal device, and the sensing device can communicate with at least one of the network device or the terminal device. At this time, the network element 2 can be called a scene recognition network element.

[0197] The network element 2 determining the identification information of the first scene may include: the network element 2 acquiring the identification information of the first scene or the network element 2 sensing the identification information of the first scene, which is not limited to this.

[0198] In one example, network element 2 performs scene recognition based on a reference signal measurement result or a perception result of one or more of a camera, a radar, or a base station, and determines identification information of the first scene. The specific manner in which network element 2 determines identification information of the first scene is not limited.

[0199] S702. Network element 2 sends identification information of the first scenario to network element 1.

[0200] Correspondingly, network element 1 receives identification information of the first scene.

[0201] By having network element 2 report the identification information of the first scene, other network elements (for example, network element 1) can determine whether it is necessary to start update training of the first model and whether it is necessary to start the data collection process for the first scene based on the identification information of the first scene, thereby supporting the effective reduction of the overall overhead of model updates.

[0202] S703. Network element 1 determines, according to identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model.

[0203] For example, network element 1 may store or be configured with identification information of scenarios supported by the first model.

[0204] After network element 1 obtains the identification information of the first scene, it can determine whether the first scene belongs to the scene currently supported by the first model. For ease of description, the following description takes the example of network element 1 determining that the first scene does not belong to the scene currently supported by the first model, but the first scene may also belong to the possible example of the scene currently supported by the first model. When it is determined that the first scene does not belong to the scene currently supported by the first model based on the identification information of the first scene, the present application can start the data collection process for the first scene, and use the data related to the first scene to update the parameters that need to be updated in the first model for training, so that the updated first model can be used for the first scene.

[0205] S704. Network element 1 sends information 1 to a data collection network element, such as network element 2, where information 1 is used to instruct acquisition of data of a first scenario.

[0206] Correspondingly, a data collection network element, such as network element 2, receives information 1 and obtains data of the first scene based on information 1. The data of the first scene can be used to update the first model. In this way, it is possible to obtain data of the first scene so that the first model can be updated and trained using the data of the first scene. The data collection network element may also be another network element different from the scene recognition network element, such as network element 2. Optionally, the scene recognition network element may determine the identification information of the first scene corresponding to the data based on the data from the data collection network element.

[0207] In addition, network element 1 determines whether it is necessary to obtain data for the first scene based on the identification information of the scene, which helps to avoid repeated data collection. When it is determined that data for the first scene needs to be obtained, network element 1 can instruct to obtain data for the first scene, and can use the data for the first scene to update and train the first model, so that the updated first model can be applied to the first scene.

[0208] In addition, through the above scheme, the present application can also avoid the situation where the first model is updated and trained when there is no need to update and train the first model, thereby effectively reducing the overall cost of model updating.

[0209] S705 . A data collection network element, such as network element 2 , sends data of the first scenario to network element 3 .

[0210] Correspondingly, network element 3 receives data of the first scene. Network element 3 may be a data storage network element, which stores data of the first scene. Network element 3 may be a data generation network element, which may extract data features of the first scene and update the data stored by the data generation network element.

[0211] Optionally, a data collection network element, such as network element 2, may also send data of the first scenario to the model training network element without limitation.

[0212] S706. Network element 1 determines a first parameter in the first model according to the identification information of the first scenario, where the first parameter is used to determine a parameter that needs to be updated in the first model.

[0213] For the description of S706, please refer to S602 and will not be repeated here.

[0214] S707. Network element 1 sends a first parameter to the model training network element.

[0215] Correspondingly, the model training network element receives the first parameter and determines the parameters that need to be updated in the first model according to the first parameter. In this way, other network elements can determine the parameters that need to be updated in the first model.

[0216] S708. The model training network element obtains first data, where the first data is related to the first scenario.

[0217] In a possible implementation manner, the first data may include any one of the following:

[0218] The data of the first scene, the data of the first scene and the data of the scenes related to the first scene, or the data generated according to the data features of the first scene and the second data, the second data is the data of the scenes currently supported by the first model, and the scenes related to the first scene belong to the scenes currently supported by the first model. Compared with the first scene, these scenes can also be called historical scenes. Through one or more of the above-mentioned items, the present application supports the completion of the update training of the parameters that need to be updated in the first model, so that the updated first model can be applied to the first scene. Optionally, the updated first model can also be applied to the aforementioned historical scenes.

[0219] When the first data is data of the first scene, the model training network element obtains the first data from network element 2, and may also obtain the first data from a data storage network element (an example of network element 3).

[0220] When the first data is data of the first scene and data of scenes related to the first scene, the model training network element obtains the first data from the data storage network element. For a description of the model training network element obtaining the first data from the data storage / generation network element (network element 3, such as the data storage network element), see Fig. 9 For descriptions of scenarios related to the first scenario, see Fig. 9 .

[0221] When the first data is data obtained by the data storage / generation network element generating data according to the data of the first scene and the data characteristics of the second data, the model training network element can obtain the first data from the data storage / generation network element (network element 3, such as the data generation network element).

[0222] In summary, the present application supports obtaining data for updating and training parameters that need to be updated in the first model through one or more of a plurality of approaches.

[0223] Further, the data generation network element generates data according to the data of the first scenario and the data features of the second data to obtain the first data, including:

[0224] S1-data generation network element obtains data of the first scene from network element 2;

[0225] S2-The data generation network element generates data according to the data of the first scenario and the data features of the second data to obtain the first data.

[0226] The data generation network element can use generative adversarial networks (GAN) and generate data according to the data of the first scene and the data features of the second data to obtain the first data. Among them, GAN is a typical unsupervised learning method that can automatically extract data features and complete data generation. GAN consists of two parts: one is the generator (generating data through a neural network), the purpose of which is to generate data that is as similar as possible to the original data to deceive the judge; the other is the judge (determining whether the data is real or machine-generated through a neural network), the purpose of which is to find out the "fake data" made by the generator.

[0227] For example, after the data generation network element obtains the data of scene 1 (scene 1 can be the initial scene), the data generation network element extracts the data features of scene 1 and generates new data. When the data generation network element obtains the data of scene 2, the data generation network element extracts the data features of scene 1 and the data features of scene 2 and generates new data; after the data generation network element obtains the data of scene 3, the data generation network element extracts the data features of scene 1, the data features of scene 2, and the data features of scene 3 and generates new data. Therefore, the data that the data generation network element has obtained or saved before obtaining the data of the first scene can be collectively referred to as the second data, and the data generation network element extracts the data features of the second data and the data features of the first scene and generates the first data. In this way, the updated AI model can also support historical scenes before the first scene.

[0228] By generating the first data based on the data features of the first scene and the second data, and without the need to obtain a large amount of data in the first scene, the present application can support the updated first model to be used not only in the first scene, but also in the scenes supported before the model was updated, which can effectively reduce the overhead of data collection, and thus can also effectively reduce the overall overhead of model updating.

[0229] In a possible implementation, the data feature of the second data includes any one of the following:

[0230] Signal to noise ratio (SNR), signal to interference plus noise ratio (SINR), reference signal received power (RSRP) or data distribution.

[0231] The above data distribution may include one or more of the following: latent variable variational probability distribution of data after variational inference, probability distribution of data after Gaussian mixture model fitting, and potential distribution of data learned through adversarial process. Thus, through one or more parameters listed above, the data generation network element can generate a large amount of data that can be used to update the first model.

[0232] S709: The model training network element uses the first data to update the parameters that need to be updated in the first model to obtain an updated first model.

[0233] The parameters that need to be updated in the first model are determined based on the first parameters (S707) received from network element 1, that is, the first parameters are used to determine the parameters that need to be updated in the first model.

[0234] For the description of the model training network element updating the parameters that need to be updated in the first model, please refer to the existing solution and will not be repeated here. Among them, since the first data is related to the first scenario, the model training network element uses the first data to update the first model, and the updated first model obtained can support the first scenario.

[0235] S710. The model training network element sends the updated first model to network element 1.

[0236] Correspondingly, network element 1 receives the updated first model.

[0237] Network element 1 can use the updated first model to infer the first scenario. For a specific description, please refer to the existing solution and will not be repeated here.

[0238] Through the above scheme, the present application supports reducing the overall overhead of model updates, and can also enable the updated AI model to support previous historical scenarios, that is, to maintain the ability to reason about historical scenarios (for example, the ability to reason about historical scenarios can be maintained by updating some parameters or using historical scene data to update the model).

[0239] Figure 7The content shown is described by taking the example of the model deployment network element sending the first parameter to the model training network element. The model training network element can also determine the first parameter by itself. Figure 8 Description.

[0240] Figure 8 800 is an interactive flow chart of the training method 800 of the AI ​​model of the embodiment of the present application. Figure 8 As shown, taking network element 1 as a model training network element as an example, method 800 includes:

[0241] S801. Network element 2 determines identification information of a first scene.

[0242] For the description of S801, please refer to S701 and will not be repeated here.

[0243] S802. Network element 2 sends identification information of the first scenario to the model deployment network element and network element 1.

[0244] Correspondingly, the model deployment network element and network element 1 respectively receive identification information of the first scenario.

[0245] For the description of S802, please refer to S702 and will not be repeated here.

[0246] S803: The model deployment network element determines, based on the identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model.

[0247] For the description of S803, please refer to S703 and will not be repeated here.

[0248] S804. The model deployment network element sends information 1 to the data collection network element, such as network element 2, where information 1 is used to instruct to obtain data of the first scenario.

[0249] For the description of S804, please refer to S704 and will not be repeated here.

[0250] S805 . A data collection network element, such as network element 2 , sends data of the first scenario to network element 3 .

[0251] Correspondingly, network element 3 receives data of the first scene.

[0252] For the description of S805, please refer to S705 and will not be repeated here.

[0253] S806. Network element 1 determines a first parameter in the first model according to the identification information of the first scenario, where the first parameter is used to determine a parameter that needs to be updated in the first model.

[0254] For the description of S806, please refer to S602 and will not be repeated here.

[0255] Among them, the parameter that needs to be updated in the first model is the first parameter.

[0256] S807. Network element 1 obtains first data, where the first data is related to the first scenario.

[0257] For the description of S807, please refer to S708 and will not be repeated here.

[0258] S808. Network element 1 uses the first data to update the parameters that need to be updated in the first model to obtain an updated first model.

[0259] For the description of S808, please refer to S709 and will not be repeated here.

[0260] S809. Network element 1 sends the updated first model to the model deployment network element.

[0261] Correspondingly, the model deployment network element receives the updated first model.

[0262] The model deployment network element can use the updated first model to infer the first scenario. For specific description, please refer to the existing solution and will not be repeated here.

[0263] Through the above solution, the present application supports reducing the overall overhead of model updates, and can also enable the updated AI model to support previous historical scenarios, that is, to maintain the ability to reason about historical scenarios.

[0264] Figure 7 and Figure 8 The content shown is described by taking the example of a model deployment network element sending information 1 to a data collection network element, such as network element 2, to instruct the acquisition of data for the first scene. The model training network element may also send information to the data collection network element, such as network element 2, to instruct the acquisition of data for the first scene. For example, the information indicating the acquisition of data for the first scene is used to indicate the acquisition of data for the first scene and the amount of data to be acquired. At this time, information 1 may not be sent. Alternatively, the model deployment network element sends both information 1 and information indicating the acquisition of data for the first scene to the data collection network element, such as network element 2. For example, the information indicating the acquisition of data for the first scene specifically indicates the amount of data to be acquired. For the process of the model training network element sending information to the data collection network element, such as network element 2, to indicate the acquisition of data for the first scene, see Fig. 9 Description.

[0265] Fig. 9 9 is an interactive flow diagram of the training method 900 of the AI ​​model of the embodiment of the present application. Fig. 9 As shown, taking network element 1 as a model training network element or a model deployment network element as an example, and taking network element 3 as a data storage network element as an example, method 900 includes:

[0266] S901. Network element 1 obtains identification information of a first scene.

[0267] For the description of S901, please refer to S601 and will not be repeated here.

[0268] S902. Network element 1 determines scenes related to the first scene according to identification information of the first scene, where the scenes related to the first scene belong to scenes currently supported by the first model.

[0269] For example, the scene related to the first scene may be: a scene having the same or similar data features as the first scene, or a scene of the same type as the first scene; or a scene having similar or identical scene features to the first scene. The description of the scene features of the scene can refer to the related description at S601 above, which will not be repeated here. The description of the data features can also refer to the description of the data features of the second data above, which will not be repeated here.

[0270] In a possible implementation, the network element 1 may determine the scene related to the first scene according to a predefined rule, or the network element 1 may determine the scene related to the first scene according to the feature information of the scene. In this way, the scene related to the first scene may be determined.

[0271] For example, the predefined rule may be: if the feature information of the scene features of two scenes is consistent or similar, they can be identified as related scenes. Among them, similarity means that the feature information meets the same condition, for example, both are lower than a certain threshold, or higher than a certain threshold, or similarity means that the difference in feature information is lower than a certain threshold. The specific threshold and / or feature information can be predefined or configured, which is not limited here.

[0272] Exemplarily, scenario 1 includes feature information 1, feature information 1 includes the following features: the delay spread of the reference signal channel measurement result in the scenario is lower than a first threshold, and the first scenario includes feature information 2, feature information 2 includes the following features: the delay spread of the reference signal channel measurement result in the scenario is lower than the first threshold. Accordingly, scenario 1 and the first scenario are two related scenarios. For another example, the predefined rule may be: if the data features of the scene features of the two scenarios are consistent or similar, they can be identified as related scenarios. Among them, similarity means that the data features all meet the same conditions, for example, they are all lower than a certain threshold, or, higher than a certain threshold, or, similarity means that the difference in data features is lower than a certain threshold. Specific thresholds and / or data features may be predefined or configured, and are not limited here.

[0273] Exemplarily, scenario 1 includes data feature 1, data feature 1 includes the following feature: the reference signal received power in scenario 1 is lower than threshold 1, and the first scenario includes data feature 2, data feature 2 includes the following feature: the reference signal received power in the first scenario is lower than threshold 1. Accordingly, scenario 1 and the first scenario are two related scenarios.

[0274] For another example, the predefined rule may be: if the scene classifications of two scenes belong to the same classification and the classification rules are predefined or preconfigured, they may be identified as related scenes.

[0275] Exemplarily, scene 1 is an indoor scene, and the first scene is also an indoor scene. Accordingly, scene 1 and the first scene are two related scenes.

[0276] S903. Network element 1 obtains quantity information of scene data related to the first scene.

[0277] Network element 1 may send information 2 to network element 3 (which may be a data storage network element), where information 2 is used to indicate the acquisition of information of a scene related to the first scene, for example, information 2 includes identification information of a scene related to the first scene, and network element 3 may determine at least one of the data of the scene related to the first scene or the quantity information of the data of the scene related to the first scene based on the identification information of the scene related to the first scene. The present application supports configuring the association between the scene and the data of the scene in network element 3, as shown in Table 3. The content described in Table 3 is understood only as an example and is not a final limitation.

[0278] Table 3

[0279] Scenario parameter Scenario 1 Data 1 Scenario 2 Data 2 Scene 3 Data 3

[0280] As shown in Table 3, scene 1 corresponds to data 1, scene 2 corresponds to data 2, and scene 3 corresponds to data 3. The present application supports establishing an association relationship between scenes and data, and stores or configures the association relationship between scenes and data in network element 3. Network element 3 can determine the corresponding data based on the association relationship between scenes and data and the identification information of the scenes. In this way, the signaling overhead for the association relationship between interactive scenes and scene data can be saved.

[0281] Optionally, when network element 1 is a model training network element, network element 3 can send data of scenes related to the first scene, and the data of scenes related to the first scene can be used for updating and training the first model. This can effectively utilize the stored data and reduce the amount of data that needs to be obtained from the first scene, which can effectively reduce the overhead of data collection, thereby supporting the reduction of the overall overhead of model updating.

[0282] In summary, by determining scenes related to the first scene based on the identification information of the first scene and obtaining data of scenes related to the first scene, this is helpful to reduce the amount of data that needs to be obtained from the first scene, thereby reducing the overhead of data collection and further reducing the overall overhead of model updating.

[0283] Optionally, when network element 1 is a model training network element, network element 1 can also obtain data of the first scene from network element 3 or a data collection network element, such as network element 2, and use the data of the first scene and the data of scenes related to the first scene to update the parameters that need to be updated in the first model to obtain an updated first model, and the updated first model can be sent to the model deployment network element.

[0284] In one possible implementation, since obtaining data for the first scene may implicitly include obtaining quantity information of the data for the first scene, in S903, when network element 1 is a model training network element, obtaining quantity information of the data for the first scene may be replaced by obtaining data for the first scene.

[0285] Optionally, the method 900 may further include:

[0286] S904. Network element 1 sends information 3 to a data collection network element, such as network element 2, where information 3 is used to indicate the amount of data to be acquired in the first scenario.

[0287] Specifically, when network element 1 determines that the first scene does not belong to the scenes currently supported by the first model based on the identification information of the first scene, network element 1 sends information 3 to a data collection network element, such as network element 2, to indicate the amount of data to be acquired in the first scene. The amount of data to be acquired in the first scene can be determined by network element 1 based on the amount information of data of scenes related to the first scene. For example, network element 1 determines whether network element 3 has stored data of scenes related to the first scene by interacting with network element 3. For example, network element 1 sends inquiry information carrying identification information of scenes related to the first scene to network element 3, and network element 3 inquires whether data of scenes related to the first scene has been stored based on the inquiry information. When network element 1 determines that data of scenes related to the first scene has been stored, network element 1 determines that it does not need to acquire a larger amount of data in the first scene. In this way, the amount of data that needs to be acquired from the first scene can be indicated.

[0288] The information 3 may also be used to instruct to obtain data of the first scene. In this way, the data of the first scene may be obtained.

[0289] Correspondingly, the data collection network element, such as network element 2, can start the data collection process of the first scenario based on information 3. Through the above embodiments, the present application can support the effective reduction of the overall overhead of model update. Among them, the signaling interaction between different network elements described in the above method embodiments can support the effective reduction of the overall overhead of model update.

[0290] Finally, the device embodiment of the embodiment of the present application is introduced.

[0291] In order to implement the functions in the method provided in this application, the terminal device and the network device may include a hardware structure and / or a software module, and implement the above functions in the form of a hardware structure, a software module, or a hardware structure plus a software module. Whether one of the above functions is executed in the form of a hardware structure, a software module, or a hardware structure plus a software module depends on the specific application and design constraints of the technical solution.

[0292] Fig.10 1 is a schematic block diagram of a communication device 1000 of an embodiment of the present application. The communication device 1000 includes a processing circuit 1010 and a transceiver circuit 1020, and the processing circuit 1010 and the transceiver circuit 1020 can be connected to each other through a bus 1030. The communication device 1000 may include one or more of network element 1, network element 2, or network element 3. In other words, the communication device 1000 may include a data collection network element (such as network element 2), a scene recognition network element (such as network element 2), a model training network element (such as network element 1), a data storage / generation network element (such as network element 3), or a model deployment network element (such as network element 1).

[0293] Optionally, the communication device 1000 may further include a memory 1040. The memory 1040 includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or a portable read-only memory (CD-ROM), and the memory 1040 is used for related instructions and data.

[0294] The processing circuit 1010 may be one or more central processing units (CPUs). When the processing circuit 1010 is a CPU, the CPU may be a single-core CPU or a multi-core CPU. The processing circuit 1010 may be a signal processor, a chip, or other integrated circuits that can implement the method of the present application, or a portion of the circuits used for processing functions in the aforementioned processors, chips, or integrated circuits. In addition, the transceiver circuit 1020 may also be a transceiver, or an input-output interface, which is used for input or output of signals or data, and may also be referred to as an input-output circuit.

[0295] When the communication device 1000 is the network element 1, illustratively, the processing circuit 1010 is used to perform the following operations: obtain identification information of the first scene; determine a first parameter in the first model according to the identification information of the first scene, etc.

[0296] When the communication device 1000 is the network element 2, illustratively, the processing circuit 1010 is used to perform the following operations: determine identification information of the first scene; send the identification information of the first scene, etc.

[0297] When the communication device 1000 is the network element 3, illustratively, the processing circuit 1010 is used to perform the following operations: receiving information 2 or information 3; sending first data to the model training network element, etc.

[0298] The above contents are only exemplary descriptions. When the communication device 1000 is the network element 1, the network element 2 or the network element 3, it will be responsible for executing the methods or steps related to the network element 1, the network element 2 or the network element 3 in the above method embodiments.

[0299] It can be understood that when the communication device 1000 is network element 1, network element 2 or network element 3, the transceiver circuit 1020 is a transceiver. When the communication device 1000 is a chip for network element 1, network element 2 or network element 3, the transceiver circuit 1020 is an input-output circuit. The above description is only an exemplary description. For specific content, please refer to the content shown in the above method embodiment. Fig.10 The implementation of each operation in can also refer to Figures 6 to 9 The corresponding description of the method embodiment shown.

[0300] Fig.11It is a schematic block diagram of a communication device 1100 of an embodiment of the present application. The communication device 1100 may include one or more of network element 1, network element 2 or network element 3, or may be a chip or module including one or more of network element 1, network element 2 or network element 3. In other words, the communication device 1100 may include one or more of a data collection network element (such as network element 2), a scene recognition network element (such as network element 2), a model training network element (such as network element 1), a data storage network element (such as network element 3), or a model deployment network element (such as network element 1), etc., or a chip or module including one or more of a data collection network element (such as network element 2), a scene recognition network element (such as network element 2), a model training network element (such as network element 1), a data storage / generation network element (such as network element 3), or a model deployment network element (such as network element 1), etc., for implementing the method involved in the above embodiment.

[0301] The communication device 1100 includes a transceiver unit 1110 and a processing unit 1120. The transceiver unit 1110 and the processing unit 1120 are described below by way of example.

[0302] The transceiver unit 1110 may include a transmitting unit and a receiving unit. The transmitting unit is used to perform a transmitting action of the communication device, and the receiving unit is used to perform a receiving action of the communication device. For ease of description, the embodiment of the present application combines the transmitting unit and the receiving unit into one transceiver unit. A unified description is given here, and no further description is given later.

[0303] When the communication device 1100 is a network element 1, illustratively, the transceiver unit 1110 is used to obtain identification information of the first scene; the processing unit 1120 is used to execute the content of the steps involving processing and control of the network element 1. For example, the processing unit 1120 is used to determine the first parameter in the first model according to the identification information of the first scene.

[0304] When the communication device 1100 is a network element 2, illustratively, the transceiver unit 1110 is used to receive information 1, and the processing unit 1120 is used to execute the content of the steps involving processing and control of the network element 2. For example, the processing unit 1120 is used to collect data of the first scenario.

[0305] The above contents are only intended as exemplary descriptions.

[0306] Among them, when the communication device 1100 includes one or more of network element 1, network element 2 or network element 3, it will be responsible for executing one or more of the methods or steps related to network element 1, network element 2 or network element 3 in the aforementioned method embodiments.

[0307] Optionally, the communication device 1100 further includes a storage unit 1130, where the storage unit 1130 is used to store a program or code for executing the aforementioned method.

[0308] It should be noted that Fig.11 The transceiver unit in can correspond to Fig.10 The transceiver circuit in Fig.11 The processing units in can correspond to Fig.10 The processing circuit in.

[0309] Fig.10 and Fig.11 The device embodiment shown is used to implement Figures 6 to 9 The content described. Fig.10 and Fig.11 The specific execution steps and methods of the device shown can refer to the contents described in the aforementioned method embodiment.

[0310] The present application also provides a chip, including a processor, for calling and executing instructions stored in a memory from the memory, so that a communication device equipped with the chip executes the methods in the above examples. The memory can be integrated into the chip, or located outside the chip.

[0311] The present application also provides another chip, including: an input interface, an output interface, and a processor, wherein the input interface, the output interface, and the processor are connected via an internal connection path, and the processor is used to execute the code in the memory, and when the code is executed, the processor is used to execute the method in each of the above examples. Optionally, the chip also includes a memory, and the memory is used to store computer programs or codes.

[0312] The present application also provides a processor, which is coupled to a memory and is used to execute the methods and functions involving a network device or a terminal device in any of the above-mentioned embodiments.

[0313] In another embodiment of the present application, a computer program product including instructions is provided. When the computer program product is run on a computer, the method of the above embodiment is implemented.

[0314] The present application also provides a computer program. When the computer program is executed in a computer, the method of the above embodiment is implemented.

[0315] In another embodiment of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a computer, the method described in the above embodiment is implemented.

[0316] It should be understood that in the embodiments of the present application, the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0317] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus RAM (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0318] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instruction can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instruction can be transmitted from a website site, a computer, a server or a data center by wired or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or a data center containing one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0319] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0320] It can be appreciated by a person skilled in the art that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or in combination with computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. It can be clearly understood by a person skilled in the art that for the convenience and simplicity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0321] The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, the functional units in each embodiment of the present application may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit. If the above functions are implemented in the form of software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions for a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory, a random access memory, a disk or an optical disk.

[0322] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

Claims

1. A training method for an artificial intelligence model, characterized in that: include: Obtaining identification information of a first scene; A first parameter in the first model is determined according to the identification information of the first scene, where the first parameter is used to determine a parameter that needs to be updated in the first model.

2. The method according to claim 1, characterized in that: The determining a first parameter in the first model according to the identification information of the first scene includes: The first parameter is determined according to identification information of the first scene and at least one association relationship, wherein a first association relationship among the at least one association relationship is used to associate the first parameter with the first scene.

3. The method according to claim 2, characterized in that The at least one association relationship is predefined or configured.

4. The method according to any one of claims 1 to 3, characterized in that The method further comprises: The first parameter is transmitted.

5. The method according to claim 4, characterized in that The method further comprises: Determining, according to the identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model; A message for instructing to obtain data of the first scenario is transmitted.

6. The method according to any one of claims 1 to 3, characterized in that The method further comprises: Acquire first data, where the first data is related to the first scenario; The first data is used to update the parameters that need to be updated in the first model to obtain the updated first model.

7. The method according to claim 6, characterized in that The first data includes any one of the following: The data of the first scene, the data of the first scene and the data of scenes related to the first scene, or the data generated according to the data features of the data of the first scene and the second data, the second data is the data of the scenes currently supported by the first model, and the scenes related to the first scene belong to the scenes currently supported by the first model.

8. The method according to claim 7, characterized in that The obtaining of the first data includes: Acquire the first data from a data collection network element, where the first data is data of the first scenario; or, Acquire the first data from a data generating network element, where the first data is data generated according to the data of the first scenario and data features of the second data; or, The first data is obtained from a data storage network element, where the first data is data of the first scene and data of scenes related to the first scene.

9. The method according to any one of claims 1 to 8, characterized in that The parameter to be updated in the first model is the first parameter, or, The parameters that need to be updated in the first model are part or all of the parameters in the first model except the first parameter.

10. The method according to any one of claims 7 to 9, characterized in that The data characteristics of the second data include at least one of the following: Signal-to-noise ratio, reference signal received power, or data distribution.

11. The method according to any one of claims 1 to 10, characterized in that The method further comprises: According to the identification information of the first scene, it is determined that the first scene does not belong to the scenes currently supported by the first model.

12. A method for training an artificial intelligence model, characterized in that: include: Obtaining identification information of a first scene; Determining, according to the identification information of the first scene, scenes related to the first scene, where the scenes related to the first scene belong to scenes currently supported by the first model; The quantity information of the scene data related to the first scene is obtained, the scene data related to the first scene is used for updating the first model, and the updated first model supports the first scene.

13. The method according to claim 12, characterized in that The method further comprises: Determining, according to the identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model; Send first indication information, where the first indication information is used to indicate the amount of data to be acquired in the first scene, the quantity information of data of scenes related to the first scene is used to determine the amount of data to be acquired in the first scene, and the data of the first scene is used to update the first model.

14. The method according to claim 12 or 13, characterized in that The determining, according to the identification information of the first scene, a scene related to the first scene includes: Determined based on predefined rules, or determined based on feature information of the scene.

15. The method according to any one of claims 12 to 14, characterized in that The first indication information is also used to instruct to obtain data of the first scene.

16. A method for training an artificial intelligence model, characterized in that: include: Obtaining identification information of a first scene; Determining, according to the identification information of the first scene, that the first scene does not belong to the scenes currently supported by the first model; Transmit first indication information, where the first indication information is used to indicate obtaining data of the first scene, the data of the first scene is used to update the first model, and the updated first model supports the first scene.

17. A method for training an artificial intelligence model, characterized in that: include: Get the data of the first scene; Generate data according to the data of the first scenario and the data features of the second data to obtain first data, where the second data is data of the scenario currently supported by the first model, and the first data is used to update the first model, and the updated first model supports the first scenario; The first data is sent.

18. The method according to claim 17, characterized in that The data characteristics of the second data include at least one of the following: Signal-to-noise ratio, reference signal received power, or data distribution.

19. A method for training an artificial intelligence model, characterized in that: include: Determining identification information of a first scene; Send identification information of the first scenario.

20. The method according to claim 19, characterized in that The method further comprises: receiving instruction information; Based on the indication information, data of the first scene is acquired.

21. The method according to claim 20, characterized in that The indication information includes information about the amount of data to be acquired in the first scene; The acquiring data of the first scenario based on the indication information includes: According to the information about the amount of data to be acquired in the first scene, data of the first scene is acquired.

22. A method for training an artificial intelligence model, characterized in that: include: Acquire indication information, where the indication information is used to instruct acquisition of data of a scene related to the first scene, and the indication information includes identification information of the scene related to the first scene; Determine data of a scene related to the first scene according to identification information of a scene related to the first scene and at least one association relationship, wherein a first association relationship in the at least one association relationship is used to associate the scene related to the first scene with data of a scene related to the first scene; Transmit data of a scene related to the first scene.

23. The method according to claim 22, characterized in that The at least one association relationship is predefined or configured.

24. A communication device, characterized in that: comprising a processor configured to, by executing a computer program or instruction, or, by a logic circuit, enabling the communication device to perform the method according to any one of claims 1 to 11; or, causing the communication device to execute the method according to any one of claims 12 to 15; or, causing the communication device to execute the method according to claim 16; or, The communication device is caused to perform the method according to claim 17 or 18; or, causing the communication device to perform the method according to any one of claims 19 to 21; or, The communication device is enabled to perform the method of claim 22 or 23.

Images