Model training method, node, communication device, communication system and storage medium

Abstract

Provided in the embodiments of the present disclosure are a model training method, a node, a communication device, a communication system and a storage medium. The method is executed by a first node. The method comprises: receiving first information, wherein the first information is used for a third node to request a training service of an artificial intelligence (AI) model; responding to the training service requested by the first information; and sending second information, wherein the second information comprises a response result provided to the third node. In the technical solution provided in the embodiments of the present disclosure, a first node centrally executes a training service of an AI model, and makes the training service of the AI model available to an external third node, such that the third node does not need to execute the training service of the AI model by itself, and directly uses a network resource of a first node to obtain the training service of the AI model.

Description

Model training methods, nodes, communication equipment, communication systems, and storage media Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to model training methods, nodes, communication devices, communication systems and storage media. Background Technology

[0002] Artificial intelligence (AI) will become one of the core technologies of future communications, particularly in 6G mobile communications (6G). th In Generation 6 (G) networks, AI will become an indispensable part, playing a crucial role.

[0003] Summary of the Invention

[0004] This disclosure provides a model training method, a node, a communication device, a communication system, and a storage medium.

[0005] According to a first aspect of the present disclosure, a model training method is provided, wherein the method is executed by a first node, the method comprising: receiving first information, the first information being used by a third node to request training services for an artificial intelligence (AI) model; responding to the training service requested by the first information; and sending second information, the second information including a response result provided to the third node.

[0006] According to a second aspect of the present disclosure, a model training method is provided, wherein the method is executed by a fourth node, the method comprising: receiving third information sent by a first node based on first information; the first information being used by the third node to request training services for an AI model; and sending a response result of the training service to the first node.

[0007] According to a third aspect of the present disclosure, a model training method is provided, wherein the method is executed by a second node, the method comprising: receiving first information sent by a third node, the first information being used by the third node to request training services for an AI model; sending the first information to a first node responding to the training services; receiving second information sent by the first node, the second information including a response result provided to the third node; and sending the second information to the third node.

[0008] According to a fourth aspect of the present disclosure, a model training method is provided, wherein the method is executed by a third node, the method comprising: sending first information, the first information being used by the third node to request training services for an AI model; and receiving second information, the second information including a response result provided to the third node.

[0009] According to a fifth aspect of the present disclosure, a model training method is provided, wherein the method is executed by a communication system, the method comprising: a first node receiving first information sent by a second node or a third node, the first information being used by the third node to request training services for an AI model; the first node responding to the training services requested by the first information; and the first node sending second information to the second node or the third node, the second information including a response result provided to the third node.

[0010] According to a sixth aspect of the present disclosure, a first node is provided, wherein the first node includes: a first receiving module configured to receive first information sent by a second node, the first information being used by a third node to request training services for an artificial intelligence (AI) model; a processing module configured to respond to the training services requested by the first information; and a first sending module configured to send second information to the second node, the second information including a response result provided to the third node.

[0011] According to a seventh aspect of the present disclosure, a second node is provided, wherein the second node includes: a second receiving module configured to receive first information sent by a third node, the first information being used by the third node to request training services for an AI model; a second sending module configured to send the first information to a first node responding to the training service; the second receiving module is further configured to receive second information sent by the first node, the second information including a response result provided to the third node; and the second sending module is further configured to send the second information to the third node.

[0012] According to an eighth aspect of the present disclosure, a third node is provided, wherein the third node includes: a third sending module configured to send first information to a second node, the first information being used by the third node to request training services for an AI model; and a third receiving module configured to receive second information sent by the second node, the second information including a response result provided to the third node.

[0013] According to a ninth aspect of the present disclosure, a fourth node is provided, wherein the fourth node includes: a fourth receiving module configured to receive third information sent by a first node based on first information; the first information is used by the third node to request training services for an AI model; and a fourth sending module configured to send a response result of the training service to the first node.

[0014] According to a tenth aspect of the present disclosure, a communication system is provided, wherein the communication system includes a first node, a second node, a third node, and a fourth node, the first node being configured to implement the model training method provided in the first aspect, the fourth node being configured to implement the model training method provided in the second aspect, the second node being configured to implement the model training method provided in the third aspect, and the third node being configured to implement the model training method provided in the fourth aspect.

[0015] According to an eleventh aspect of the present disclosure, a communication device is provided, wherein the communication device includes:

[0016] One or more processors;

[0017] The processor is used to invoke instructions to cause the communication device to execute the model training method provided by the first aspect, the second aspect, the third aspect, or the fourth aspect.

[0018] According to a twelfth aspect of the present disclosure, a storage medium is provided, wherein the storage medium stores instructions that, when executed on a communication device, cause the communication device to perform a model training method provided in the first aspect, the second aspect, the third aspect, or the fourth aspect.

[0019] According to a thirteenth aspect of the present disclosure, a program product is provided, wherein when the program product is executed by a communication device, the communication device performs the model training method provided by the first aspect, the second aspect, the third aspect, or the fourth aspect.

[0020] According to a fourteenth aspect of the present disclosure, a computer program is provided that, when run on a computer, causes the computer to perform the model training method provided in the first, second, third, or fourth aspect.

[0021] The technical solution provided in this embodiment involves a first node centrally executing the AI ​​model training service, and a second node opening the AI ​​model training service to an external third node, so that the third node does not need to execute the AI ​​model training service itself, but can directly obtain the AI ​​model training service by utilizing the network resources of the first node.

[0022] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the embodiments of this disclosure. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the embodiments of the invention.

[0024] Figure 1A is a schematic diagram of the architecture of a communication system according to an exemplary embodiment;

[0025] Figure 1B is a schematic diagram illustrating a usage scenario of a novel mobile communication system according to an exemplary embodiment;

[0026] Figure 1C is a schematic diagram illustrating the network capabilities of a novel mobile communication system according to an exemplary embodiment;

[0027] Figure 1D is a schematic diagram illustrating an ML model training process according to an exemplary embodiment;

[0028] Figure 2A is an interactive schematic diagram of a model training method according to an exemplary embodiment;

[0029] Figure 2B is an interactive schematic diagram of a model training method according to an exemplary embodiment;

[0030] Figure 3A is a flowchart illustrating a model training method according to an exemplary embodiment;

[0031] Figure 3B is a schematic flowchart of a model training method according to an exemplary embodiment;

[0032] Figure 4 is a flowchart illustrating a model training method according to an exemplary embodiment.

[0033] Figure 5 is a flowchart illustrating a model training method according to an exemplary embodiment.

[0034] Figure 6 is a flowchart illustrating a model training method according to an exemplary embodiment;

[0035] Figure 7 is an interactive schematic diagram of a model training method according to an exemplary embodiment;

[0036] Figure 8A is a schematic diagram of a service-based AI-as-a-service architecture in a 6G system according to an exemplary embodiment;

[0037] Figure 8B is a schematic diagram illustrating a process for providing AI model training as a service in a 6G system according to an exemplary embodiment.

[0038] Figure 9A is a schematic diagram of the structure of a first node according to an exemplary embodiment;

[0039] Figure 9B is a schematic diagram of the structure of a second node according to an exemplary embodiment;

[0040] Figure 9C is a schematic diagram of the structure of a third node according to an exemplary embodiment;

[0041] Figure 9D is a schematic diagram of the structure of a fourth node according to an exemplary embodiment;

[0042] Figure 10A is a schematic diagram of the structure of a communication device according to an exemplary embodiment;

[0043] Figure 10B is a schematic diagram of the structure of a chip according to an exemplary embodiment. Detailed Implementation

[0044] This disclosure provides a model training method, a node, a communication device, a communication system, and a storage medium.

[0045] In a first aspect, embodiments of this disclosure provide a model training method, wherein the method is executed by a first node, and the method includes: receiving first information, the first information being used by a third node to request training services for an artificial intelligence (AI) model; responding to the training service requested by the first information; and sending second information, the second information including a response result provided to the third node.

[0046] In the above embodiment, upon receiving the first information, the first node responds to the third node's request for AI model training services and provides the third node with the training service response results. Thus, the first node centrally executes the AI ​​model training services and exposes these services to external third nodes, allowing the third nodes to directly utilize the first node's network resources to obtain the AI ​​model training services without needing to execute them themselves.

[0047] In conjunction with some embodiments of the first aspect, in some embodiments, receiving the first information includes at least one of the following:

[0048] If the third node is a trusted node, the first information sent by the third node is received;

[0049] If the third node is not a trusted node, the first information sent by the second node is received.

[0050] In the above embodiments, if the third node requesting the training service is a trusted node, the second node does not need to participate, and the third node can directly send the first information to the first node; if the third node requesting the training service is not a trusted node, the first node receives the first information forwarded by the second node; thus, security risks are reduced in the process of opening the AI ​​model training service provided by the first node.

[0051] In conjunction with some embodiments of the first aspect, in some embodiments, the training service responding to the first information request includes: selecting one or more fourth nodes to respond to the training service based on at least one of the request information for the training service and the status information of candidate nodes, wherein the fourth nodes are used to provide the response result of the training service; sending third information to the fourth nodes based on the first information; and receiving the response result sent by the fourth nodes based on the third information.

[0052] In the above embodiments, the first node selects a suitable fourth node to respond to the training service requested by the third node based on the training service request information from the third node and / or the status information of the alternative nodes. In this way, the first node can rationally allocate the AI ​​model's training service and effectively utilize the network resources of the fourth node, enabling the AI ​​model's training service to be responded to more efficiently.

[0053] In conjunction with some embodiments of the first aspect, in some embodiments, sending the second information includes at least one of the following:

[0054] Based on the response result provided by the fourth node, send the second information to the second node;

[0055] Based on the response provided by the fourth node, a second message is sent to the trusted third node.

[0056] In the above embodiments, the first node centrally manages the allocation and delivery of AI model training services, and the fourth node responds to the allocated training services and directly sends the response results to the trusted third node, or sends the response results to the second node, which then forwards them to the third node; thereby providing AI services to the third node more effectively.

[0057] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes: determining the demand information requesting the training service based on the first information, the demand information including: fourth information for indicating service parameters for providing the training service, the service parameters including at least one of the following: the type of the AI ​​model and the training method of the AI ​​model.

[0058] In the above embodiment, the first node determines the service parameters for providing the training service requested by the third node based on the first information; thus, in the subsequent selection process of the fourth node, the first node can select a suitable fourth node based on the service parameters of the training service, i.e., the type of AI model and / or the training method of the AI ​​model related to the training service, to ensure that the selected fourth node has the corresponding capabilities to respond to the training service.

[0059] In conjunction with some embodiments of the first aspect, in some embodiments, sending the third information to the fourth node based on the first information further includes:

[0060] Based on the determined training method of the AI ​​model, assign training tasks of the AI ​​model to one or more selected fourth nodes;

[0061] Send third information to one or more fourth nodes, the third information including the assigned training task of the AI ​​model.

[0062] In the above embodiments, the first node selects one or more fourth nodes that are compatible with the determined training method of the AI ​​model, and assigns the AI ​​model training task to the one or more compatible fourth nodes. The assigned AI model training task is distributed to the fourth nodes through third information, so that the fourth nodes can execute the training task assigned by the first node. In this way, it is ensured that the AI ​​model training service requested by the third node can be responded to efficiently.

[0063] In conjunction with some embodiments of the first aspect, in some embodiments, the method further includes: receiving the status information sent by the fourth node, the status information including at least one of the following: capability information, used to indicate the processing capability of the fourth node; load information, used to indicate the load intensity of the fourth node.

[0064] In the above embodiment, the first node receives status information sent by the fourth node to know the processing capacity and load intensity of the fourth node, so as to assist the first node in selecting a suitable fourth node to respond to the training service requested by the third node, and ensure that the training service can be responded to more efficiently.

[0065] In conjunction with some embodiments of the first aspect, in some embodiments, the operation of the training service includes at least one of the following: training operation of the AI ​​model; collection operation of training data; and delivery operation of training results.

[0066] In the above embodiments, when a third node requests training services for an AI model, the first node provides a response result of the training service to the third node by performing at least one of the following operations: training the AI ​​model, collecting training data, and delivering training results.

[0067] In conjunction with some embodiments of the first aspect, in some embodiments, the first information includes at least one of the following: an AI model to be trained; training data of the AI ​​model to be trained; metadata of the AI ​​model to be trained, the metadata being used to indicate model attributes of the AI ​​model to be trained; and fifth information for requesting the first node to collect the training data.

[0068] In the above embodiments, the first information sent by the third node may include the AI ​​model to be trained (model parameters), training data, metadata and / or fifth information, so that the first node can train the AI ​​model based on the training data in the first information or the training data collected by the first node, so as to obtain a trained AI model and complete the training service requested by the second node.

[0069] In conjunction with some embodiments of the first aspect, in some embodiments, the metadata includes at least one of the following: the model version of the AI ​​model; the model identifier of the AI ​​model; the model size of the AI ​​model; the vendor identifier of the AI ​​model; and the training time information of the AI ​​model.

[0070] In the above embodiments, the metadata of the AI ​​model may include at least one of the following: the model version of the AI ​​model, the model identifier of the AI ​​model, the model size of the AI ​​model, the vendor identifier of the AI ​​model, and the training time information of the AI ​​model, so that the first node can more accurately determine the AI ​​model requested for training by the third node based on the metadata of the AI ​​model to be trained.

[0071] In conjunction with some embodiments of the first aspect, in some embodiments, sending third information to the fourth node based on the first information includes: collecting training data of the AI ​​model when the first information does not contain the training data or when the first information includes the fifth information; and sending the third information to the fourth node, wherein the third information includes the collected training data.

[0072] In the above embodiments, if the first information does not contain training data, or if the first information includes the fifth information, i.e., the third node requests the first node to collect training data, in order to ensure the training of the AI ​​model, the first node collects training data on its own; and sends the training data collected by the first node to the fourth node through the third information, so that the fourth node can perform the training operation of the AI ​​model based on the training data trained by the first node, so as to obtain the trained AI model.

[0073] In conjunction with some embodiments of the first aspect, in some embodiments, the second information includes at least one of the following: model parameters of the trained AI model; metadata of the trained AI model.

[0074] In the above embodiment, the first node responds to the AI ​​model training service and sends the model parameters and / or metadata of the trained AI model to the second node via second information, so that the second node can send the second information to the third node. In this way, the third node can utilize the network resources of the first node to perform the AI ​​model training service.

[0075] In conjunction with some embodiments of the first aspect, in some embodiments, the fourth node responding to the training service includes at least one of the following: a first network function, wherein the first network function is a core network function having a first capability and a second capability; a second network function, wherein the second network function is a core network function having the first capability; the first capability is the training capability of an AI model; and the second capability is different from the first capability.

[0076] In the above embodiments, two network functions with different capability ranges are used to respond to the training service of the AI ​​model, so that the first node can select the network function with a more suitable capability range to respond to the processing parameters required by the training service requested by the third node.

[0077] Secondly, embodiments of this disclosure provide a model training method, wherein the method is executed by a fourth node, the method comprising: receiving third information sent by a first node based on first information; the first information being used by the third node to request training services for an AI model; and sending the response result of the training service to the first node.

[0078] In the above embodiments, the fourth node receives the third information sent by the first node and responds to the training service of the AI ​​model allocated by the first node, providing the response result of the training service to the first node; thus, the first node centrally manages the allocation and delivery of the training service of the AI ​​model, and the fourth node responds to the allocated training service; thereby providing AI services to the third node in a better manner.

[0079] Thirdly, embodiments of this disclosure provide a model training method, wherein the method is executed by a second node, the method comprising: receiving first information sent by a third node, the first information being used by the third node to request training services for an AI model; sending the first information to a first node responding to the training service; receiving second information sent by the first node, the second information including a response result provided to the third node; and sending the second information to the third node.

[0080] In the above embodiment, upon receiving the first information sent by the third node, the second node forwards the first information to the first node, requesting the first node to provide AI model training services for the third node. Furthermore, upon receiving the second information sent by the first node, the second node sends the second information to the third node, conveying the response result of the AI ​​model training service provided by the first node. Thus, the second node enables the first node to expose its AI model training service to an external third node, allowing the third node to directly utilize the first node's network resources to obtain the AI ​​model training service without having to execute it itself.

[0081] Fourthly, embodiments of this disclosure provide a model training method, wherein the method is executed by a third node, and the method includes: sending first information, the first information being used by the third node to request training services for an AI model; and receiving second information, the second information including a response result provided to the third node.

[0082] In the above embodiment, the third node requests the training service of the AI ​​model from the first node by sending the first message; and obtains the response result of the training service by receiving the second message; in this way, the first node centrally executes the AI ​​service and opens the result of the AI ​​service to the external third node; so that the third node does not need to execute the AI ​​service itself, reducing the occupation of the third node's network resources.

[0083] Fifthly, embodiments of this disclosure provide a model training method, which is executed by a communication system. The method includes: a first node receiving first information sent by a second node or a third node, the first information being used by the third node to request training services for an AI model; the first node responding to the training services requested by the first information; and the first node sending second information to the second node or the third node, the second information including a response result provided to the third node.

[0084] In a sixth aspect, embodiments of this disclosure provide a first node, wherein the first node includes: a first receiving module configured to receive first information, the first information being used by a third node to request training services for an artificial intelligence (AI) model; a processing module configured to respond to the training services requested by the first information; and a first sending module configured to send second information, the second information including a response result provided to the third node.

[0085] In a seventh aspect, embodiments of this disclosure provide a second node, wherein the second node includes: a second receiving module configured to receive first information sent by a third node, the first information being used by the third node to request training services for an AI model; a second sending module configured to send the first information to a first node responding to the training service; the second receiving module is further configured to receive second information sent by the first node, the second information including a response result provided to the third node; and the second sending module is further configured to send the second information to the third node.

[0086] Eighthly, embodiments of this disclosure provide a third node, wherein the third node includes: a third sending module configured to send first information, the first information being used by the third node to request training services for an AI model; and a third receiving module configured to receive second information, the second information including a response result provided to the third node.

[0087] In a ninth aspect, embodiments of this disclosure provide a fourth node, wherein the fourth node includes: a fourth receiving module configured to receive third information sent by a first node based on first information; the first information is used by the third node to request training services for an AI model; and a fourth sending module configured to send the response result of the training service to the first node.

[0088] In a tenth aspect, embodiments of this disclosure provide a communication system, wherein the communication system includes a first node, a second node, a third node, and a fourth node, the first node being configured to implement the model training method described in the optional implementation of the first aspect, the fourth node being configured to implement the model training method described in the optional implementation of the second aspect, the second node being configured to implement the model training method described in the optional implementation of the third aspect, and the third node being configured to implement the model training method described in the optional implementation of the fourth aspect.

[0089] Eleventhly, embodiments of this disclosure provide a communication device, the communication device comprising:

[0090] One or more processors;

[0091] The processor is used to invoke instructions to cause the communication device to execute the model training method described in the optional implementation of the first, second, third, or fourth aspect.

[0092] In a twelfth aspect, embodiments of this disclosure provide a storage medium storing instructions that, when executed on a communication device, cause the communication device to perform the model training method described in the optional implementations of the first, second, third, or fourth aspects.

[0093] In a thirteenth aspect, embodiments of this disclosure provide a program product that, when executed by a communication device, causes the communication device to perform the model training method described in the optional implementations of the first, second, third, or fourth aspects.

[0094] In a fourteenth aspect, embodiments of this disclosure provide a computer program that, when run on a computer, causes the computer to perform the model training method described in optional implementations of the first, second, third, or fourth aspects.

[0095] Understandably, the aforementioned first node, second node, third node, fourth node, communication device, communication system, storage medium, program product, and computer program are all used to execute the methods provided in the embodiments of this disclosure. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0096] This disclosure provides a model training method, a node, a communication device, a communication system, and a storage medium. In some embodiments, the terms "model training method" and "information processing method" and "information transmission method" can be used interchangeably; the terms "information indication device" and "information processing device" and "information transmission device" can be used interchangeably; and the terms "communication system" and "information processing system" can be used interchangeably.

[0097] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0098] In each of the disclosed embodiments, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0099] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0100] In this embodiment of the disclosure, unless otherwise stated, elements expressed in the singular form, such as "a," "an," "the," "the," "the," "the," "the," "the," "this," etc., can mean "one and only one," or "one or more," "at least one," etc. For example, when using articles such as "a," "an," "the," etc. in translation, the noun following the article can be understood as either a singular expression or a plural expression.

[0101] In the embodiments disclosed herein, "multiple" refers to two or more.

[0102] In some embodiments, the terms “at least one of”, “one or more”, “a plurality of”, “multiple”, etc., may be used interchangeably.

[0103] In some embodiments, the notation "at least one of A and B", "A and / or B", "A in one case, B in another", "A in one case, B in another", etc., may include the following technical solutions depending on the situation: in some embodiments, A (A is executed regardless of B); in some embodiments, B (B is executed regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). The same applies when there are more branches such as A, B, C, etc.

[0104] In some embodiments, the notation "A or B" may include the following technical solutions, depending on the situation: in some embodiments, A (execution of A regardless of B); in some embodiments, B (execution of B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed). The same applies when there are more branches such as A, B, C, etc.

[0105] The prefixes "first," "second," etc., used in the embodiments of this disclosure are merely for distinguishing different descriptive objects and do not impose restrictions on the position, order, priority, quantity, or content of the descriptive objects. The description of the descriptive objects is found in the claims or the context of the embodiments, and the use of prefixes should not constitute unnecessary restrictions. For example, if the descriptive object is a "field," the ordinal numbers preceding "field" in "first field" and "second field" do not restrict the position or order of the "fields." "First" and "second" do not restrict whether the "fields" they modify are in the same message, nor do they restrict the order of "first field" and "second field." Similarly, if the descriptive object is a "level," the ordinal numbers preceding "level" in "first level" and "second level" do not restrict the priority between "levels." Furthermore, the number of descriptive objects is not limited by ordinal numbers and can be one or more. For example, in "first device," the number of "devices" can be one or more. Furthermore, the objects modified by different prefixes can be the same or different. For example, if the object being described is "device", then "first device" and "second device" can be the same device or different devices, and their types can be the same or different. Similarly, if the object being described is "information", then "first information" and "second information" can be the same information or different information, and their content can be the same or different.

[0106] In some embodiments, “including A,” “containing A,” “for indicating A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0107] In some embodiments, terms such as “…”, “determine…”, “in the case of…”, “when…”, “when…”, “if…”, etc. can be used interchangeably.

[0108] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0109] In some embodiments, devices, etc., can be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. Terms such as “device”, “equipment”, “circuit”, “network element”, “node”, “function”, “unit”, “section”, “system”, “network”, “chip”, “chip system”, “entity”, and “subject” can be used interchangeably.

[0110] In some embodiments, "network" can be interpreted as devices included in a network (e.g., access network devices, core network devices, etc.).

[0111] In some embodiments, the terms "access network device (AN device)," "radio access network device (RAN device)," "base station (BS)," "radio base station," "fixed station," "node," "access point," "transmission point (TP)," "reception point (RP)," "transmission / reception point (TRP)," "panel," "antenna panel," "antenna array," "cell," "macro cell," "small cell," "femto cell," "pico cell," "sector," "cell group," "serving cell," "carrier," "component carrier," and "bandwidth part (BWP)" can be used interchangeably.

[0112] In some embodiments, the terms "terminal", "terminal device", "user equipment (UE)", "user terminal", "mobile station (MS)", "mobile terminal (MT)", "subscriber station", "mobile unit", "subscriber unit", "wireless unit", "remote unit", "mobile device", "wireless device", "wireless communication device", "remote device", "mobile subscriber station", "access terminal", "mobile terminal", "wireless terminal", "remote terminal", "handset", "user agent", "mobile client", and "client" can be used interchangeably.

[0113] In some embodiments, access network devices, core network devices, or network devices can be replaced by terminals. For example, embodiments of this disclosure can also be applied to structures where communication between access network devices, core network devices, or network devices and terminals is replaced by communication between multiple terminals (e.g., device-to-device (D2D), vehicle-to-everything (V2X), etc.). In this case, the structure can also be configured such that the terminal has all or part of the functions of the access network device. Furthermore, terms such as "uplink" and "downlink" can be replaced with terms corresponding to communication between terminals (e.g., "sidelink"). For example, uplink channel, downlink channel, etc., can be replaced with sidelink channel, and uplink link, downlink, etc., can be replaced with sidelink link.

[0114] In some embodiments, the terminal may be replaced by an access network device, a core network device, or a network device. In this case, the access network device, core network device, or network device may also be configured to have all or some of the functions of the terminal.

[0115] In some embodiments, the acquisition of data, information, etc., may comply with the laws and regulations of the country where the location is situated.

[0116] In some embodiments, data, information, etc., may be obtained with the user's consent.

[0117] Furthermore, each element, each row, or each column in the table of this disclosure can be implemented as an independent embodiment, and any combination of any element, any row, or any column can also be implemented as an independent embodiment.

[0118] Figure 1A is a schematic diagram of the architecture of a communication system according to an exemplary embodiment.

[0119] As shown in Figure 1A, the communication system 100 includes: a first node 101, a second node 102, a third node 103, and a fourth node 104.

[0120] In some embodiments, the first node may be a core network device.

[0121] In some embodiments, the second node may be a core network device.

[0122] In some embodiments, the second node may be a Network Exposure Function (NEF).

[0123] In some embodiments, the third node may be an application function or an application service (AS).

[0124] In some embodiments, the fourth node may be a core network device.

[0125] In some embodiments, the terminal includes, but is not limited to, at least one of the following: mobile phone, wearable device, Internet of Things device, car with communication function, smart car, tablet computer, computer with wireless transceiver function, virtual reality (VR) terminal device, augmented reality (AR) terminal device, wireless terminal device in industrial control, wireless terminal device in self-driving, wireless terminal device in remote medical surgery, wireless terminal device in smart grid, wireless terminal device in transportation safety, wireless terminal device in smart city, and wireless terminal device in smart home.

[0126] In some embodiments, a core network device may be a single device including one or more network functions, or it may be multiple devices or a group of devices, each including network functions. Network functions may be virtual or physical. The core network may include, for example, at least one of an Evolved Packet Core (EPC), a 5G Core Network (5GCN), or a Next Generation Core (NGC).

[0127] It is worth noting that the first node, the second node, and the fourth node can be different network functions.

[0128] It is understood that the communication system described in this disclosure is for the purpose of more clearly illustrating the technical solutions of this disclosure, and does not constitute a limitation on the technical solutions provided in this disclosure. As those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions provided in this disclosure are also applicable to similar technical problems.

[0129] The following embodiments of this disclosure can be applied to the communication system 100 shown in FIG1A, or to some of the main bodies, but are not limited thereto. The main bodies shown in FIG1A are illustrative. The communication system may include all or some of the main bodies in FIG1A, or it may include other main bodies outside of FIG1A. The number and form of each main body are arbitrary. The connection relationship between the main bodies is illustrative. The main bodies may not be connected or may be connected. The connection can be in any way, it can be a direct connection or an indirect connection, it can be a wired connection or a wireless connection.

[0130] The embodiments disclosed herein can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 5G new radio (NR), Future Radio Access (FRA), New-Radio Access Technology (RAT), New Radio (NR), New radio access (NX), Future generation radio access (FX), Global System for Mobile communications (GSM), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20, Ultra-Wideband (UWB), Bluetooth (a registered trademark), Public Land Mobile Network (PLMN) networks, Device-to-Device (D2D) systems, Machine-to-Machine (M2M) systems, Internet of Things (IoT) systems, Vehicle-to-Everything (V2X) systems, systems utilizing other communication methods, and next-generation systems built upon them, etc. Furthermore, multiple systems can be combined (e.g., a combination of LTE or LTE-A with 5G).

[0131] As shown in Figures 1B and 1C, Figure 1B is a schematic diagram illustrating a usage scenario of a novel mobile communication system according to an exemplary embodiment. Figure 1C is a schematic diagram illustrating the network capabilities of a novel mobile communication system according to an exemplary embodiment.

[0132] Compared to the three existing use cases of mobile communication systems—Enhance Mobile Broadband (eMBB), Ultra-Reliable Low-Latency Communication (URLLC), and Massive Machine Type Communication (mMTC)—the aforementioned new mobile communication systems are expected to support expanded and new use cases.

[0133] In some embodiments, the novel mobile communication system described above supports use cases including: immersive communication, AI and communication, ultra-reliable low-latency communication, ubiquitous connectivity, massive communication, and integrated sensing and communication.

[0134] Specifically, for AI and communication scenarios, distributed computing and artificial intelligence applications will be supported. Typical use cases include: autonomous driving assistance; autonomous collaboration between devices to achieve medical assistance applications; decentralizing computationally intensive operations across devices and networks; and creating digital twins for event prediction.

[0135] This use case will require support for high regional traffic capacity and user experience rates, as well as low latency and high reliability, depending on the specific use case. Beyond communications, this use case is expected to include a range of new capabilities related to the integration of artificial intelligence and computing functions into IMT-2030, including data acquisition, preparation, and processing from diverse sources; distributed AI model training, model sharing, and distributed inference across IMT systems; and computing resource coordination and interlocking.

[0136] At the same time, these new mobile communication systems offer enhanced and new capabilities. Furthermore, each capability may have different relevance and adaptability in different use cases.

[0137] For example, AI-related capabilities and interoperability capabilities. AI-related capabilities refer to the ability to provide certain functions in new mobile communication systems to support AI applications. These functions include distributed data processing, distributed learning, AI computation, AI model execution, and AI model inference. Interoperability capabilities refer to radio interfaces based on member inclusion and transparency to enable functionality between different entities within the system.

[0138] In some embodiments, the Network Data Analytics Function (NWDAF) provides the Analytics Data Repository Function (ADRF) service;

[0139] ADRF service operations can be used to store data or analytics in ADRF, retrieve data or analytics from ADRF, or delete data or analytics from ADRF.

[0140] ADRF operations can also be used to store machine learning (ML) models in an ADRF, retrieve ML models from an ADRF, or delete ML models from an ADRF.

[0141] As shown in Table 1, Table 1 lists the network function services provided by ADRF.

[0142] Table 1

[0143] NWDAF provides analytics for other network functions (NFs) and operation administration and maintenance (OAM) systems in the 5G core network (5GC). NWDAF may include the following logical functions:

[0144] (1) Analytics Logical Function (AnLF): A logical function in NWDAF that performs inference, derives analytics information (i.e., derives statistical data and / or forecasts based on analytics consumer requests) and exposes analytics services, namely Nnwdaf_AnalyticsSubscription or Nnwdaf_AnalyticsInfo.

[0145] (2) Model Training Logical Function (MTLF): A logical function in NWDAF that is used to train ML models and expose new training services (e.g., providing a fully trained ML model).

[0146] Note: NWDAF may include MTLF and / or AnLF.

[0147] As shown in Figure 1D, which is a schematic diagram of an ML model training process according to an exemplary embodiment, a consumer of an NWDAF service (i.e., an NWDAF containing MTLF) uses the process shown in Figure 1D to subscribe to another NWDAF (i.e., an NWDAF containing MTLF) to obtain an ML model trained based on ML model files or ML model information (such as that provided by the NWDAF service consumer). The NWDAF containing MTLF can use this service to enable or update the ML model. The NWDAF service consumer also uses this service to request the NWDAF containing MTLF to prepare for ML model training or modify the training of an existing ML model.

[0148] In NWDAF, model training, model inference, or storage are performed only within NWDAF. It does not provide training, inference, or storage services to AF, or support 6G NF with AI capabilities.

[0149] Currently, NWDAF including MTLF has been extended to train AI or ML models for AI localization only for the Location Management Function (LMF), but NWDAF does not support providing AI training or inference services for Application Function (AF).

[0150] It has become a consensus to reuse the capabilities of 5G to expose potential 6G network services. However, AI as a Service is not yet well defined in 5G or 6G systems. For example, how to reuse the Network Exposure Function (NEF) interface to expose 6G AI as a Service.

[0151] Figure 2A is an interactive schematic diagram illustrating a model training method according to an exemplary embodiment. As shown in Figure 2A, this disclosure relates to a model training method for a communication system 100, the method comprising:

[0152] Step S2101: The third node sends the first information to the second node.

[0153] In some embodiments, the second node receives the first information sent by the third node.

[0154] In some embodiments, the first information is used by the third node to request AI services.

[0155] In some embodiments, the first information is used by the third node to request training services for the AI ​​model.

[0156] In some embodiments, the first information may be a training request for an AI model.

[0157] It should be noted that the third node can be a consumer of AI services.

[0158] In some embodiments, the third node is deployed outside the trust domain of the first node.

[0159] It is understandable that the third node is not a trusted node. It is worth noting that, in this embodiment, because the third node is not a trusted node, in order to reduce the security risks during the process of the third node requesting the AI ​​model training service provided by the first node, the second node needs to participate in the process.

[0160] In some embodiments, the second node may develop the AI ​​services provided by the first node to the third node.

[0161] Here, the second node can be a network open function.

[0162] In some embodiments, the third node may include at least one of the following: AF; Application Service (AS); Core Network Function; Access Network Function.

[0163] It should be noted that the third node can be a third-party AF or AS; or, the third node can be any network function in the network. The AI ​​model training service provided by the first node can be made available to any network function in the network or to a third-party AF or AS.

[0164] It is worth noting that while the third node is a consumer of AI services, it can also be a provider of other business services.

[0165] In some embodiments, the first information may include at least one of the following: the AI ​​model to be trained; the training data of the AI ​​model to be trained; the metadata of the AI ​​model to be trained; and the fifth information, used to request the first node to collect training data.

[0166] Here, the first information includes the AI ​​model to be trained, which may be the model parameters of the AI ​​model to be trained carried in the first information, so that the first node can determine the AI ​​model to be trained based on the model parameters in the first information.

[0167] Model parameters include at least one of the following: model features, model structure, and weights. Model features refer to the data attributes or dimensions input into the AI ​​model. Model structure indicates the optimization algorithm, loss function, and network structure used by the AI ​​model. Weights are parameters indicating the connections of each neuron or layer in the model.

[0168] It's important to note that AI model parameters can be divided into two categories: trainable parameters (e.g., weights) and non-trainable parameters (e.g., model structure). Trainable parameters are learned through training. Non-trainable parameters are fixed and do not update during model training.

[0169] In some embodiments, the model parameters of the AI ​​model to be trained may include at least one of the following: non-training parameters; initial parameter values ​​for trainable parameters.

[0170] In some embodiments, training data is used for training, validating, and testing the AI ​​model to be trained.

[0171] It should be noted that training data is one of the most important parts of training an AI model; training an AI model refers to using the training data to make the AI ​​model fit the patterns in the training data through certain methods, and determining the model parameters (i.e., training parameters) of the AI ​​model.

[0172] In some embodiments, the metadata of the AI ​​model to be trained is used to indicate the model attributes of the AI ​​model to be trained.

[0173] It's important to note that metadata is data that describes various attributes and information about an AI model. Metadata helps in understanding and managing AI models.

[0174] In some embodiments, metadata may include at least one of the following: the model version of the AI ​​model; the model identifier of the AI ​​model; the model size of the AI ​​model; the vendor identifier of the AI ​​model; and the training time information of the AI ​​model.

[0175] In some embodiments, the fifth information is used to trigger the first node to collect training data for the AI ​​model to be trained.

[0176] In some embodiments, the fifth information is used to instruct the training of the AI ​​model to be trained using network data.

[0177] It should be noted that, in cases where the first information includes the fifth information, the first node is responsible for collecting the training data required to train the AI ​​model.

[0178] In some embodiments, the first information further includes a sixth information, which is used to indicate the third node.

[0179] Here, the sixth information can be identification information indicating the third node, or it can be other information that can be used to indicate the identity of the third node.

[0180] In some embodiments, the sixth information is also used to determine whether the third node is authorized to obtain training services.

[0181] It should be noted that after receiving the first message from the third node, the second node needs to perform authorization verification on the third node based on the sixth message within the first message to determine whether the third node has the authority to obtain training services.

[0182] Step S2102: The second node sends the first information to the first node.

[0183] In some embodiments, the first node receives first information sent by the second node.

[0184] In some embodiments, the second node sending first information to the first node includes: sending first information to the first node when it is determined that the third node is authorized to obtain training services.

[0185] Understandably, the second node verifies the authorization of the third node. If it determines that the third node has the authority to access the training service, the second node sends a first message to the first node, requesting the first node to provide the training service to the third node. If it determines that the third node does not have the authority to access the training service, the second node may choose not to send the first message.

[0186] In some embodiments, the first node may be an AI-related network function.

[0187] In some embodiments, the first node can be used to provide AI model-related services.

[0188] In some embodiments, the first node can be used to provide training services for AI models.

[0189] It should be noted that the first node is the provider of AI services; the first node can be a network function that can be used to provide training services for AI models.

[0190] In some embodiments, the first node includes: network functions that participate in the training service of the AI ​​model.

[0191] In some embodiments, the operation of the training service includes at least one of the following: training the AI ​​model; collecting training data; and delivering the training results.

[0192] It should be noted that the first node provides the training service response to the third node by performing at least one of the following operations: training the AI ​​model, collecting training data, and delivering training results.

[0193] Step S2103: The first node responds to the training service of the first information request.

[0194] In some embodiments, the method further includes:

[0195] Based on the first piece of information, determine the required information for requesting training services.

[0196] In some embodiments, the demand information is used to indicate the second node's demand for the response results of the requested training service.

[0197] For example, the demand information determined based on the first information may include training accuracy requirements, training step requirements, and / or training stopping condition requirements.

[0198] In some embodiments, demand information can be used to assist the first node in responding to the first information request's AI service.

[0199] In some embodiments, the demand information includes: fourth information, which indicates the service parameters for providing training services.

[0200] The fourth information indicates that the service parameters for providing training services can be execution parameters related to the training services.

[0201] In some embodiments, the service parameters include at least one of the following: the type of AI model; the training method of the AI ​​model.

[0202] It's important to note that AI models can encompass several different types, each suitable for different tasks and problems. The network structures of different types of AI models may differ. For example, AI models can include feedforward neural networks (FNNs), recurrent neural networks (RNNs), and generative adversarial networks (GANs). FNNs consist of input layers, hidden layers, and output layers, and are suitable for classification and regression problems. RNNs are neural networks with recurrent connections, suitable for sequential data, such as natural language processing. GANs consist of a generator and a discriminator, used to generate realistic data samples.

[0203] In some embodiments, the type of AI model may be determined by the business scenario to which the AI ​​model is applicable.

[0204] For example, in extended reality (XR) business scenarios, the second node needs to process the audio and video data streams in the XR business using an AI model; the second node can send a first message to the first node to request training of an AI model for the audio and video data streams.

[0205] For example, in a location-based service scenario, a second node (e.g., LMF) can send the first message to the first node to request the training of an AI model for location services.

[0206] It should be noted that the network structure of the AI ​​model may differ depending on the business scenario.

[0207] The type of AI model is used in the first node to determine how the AI ​​model is trained.

[0208] AI model training methods can include supervised learning, unsupervised learning, and federated learning. It should be noted that different training methods are suitable for different types of AI models, different problems, and / or different application scenarios.

[0209] In some embodiments, the training service in which the first node responds to the first information request includes:

[0210] Based on the demand information, execute the requested training service.

[0211] It should be noted that the first node can determine the operations related to the training service based on the demand information, and execute the operations related to the training service to obtain the response results of the training service.

[0212] Understandably, when the first information requests training services for an AI model, the first node determines the training method for the AI ​​model based on the request information, and performs the training operation of the AI ​​model according to the training method to obtain a trained AI model; then, the trained AI model is sent to the second node.

[0213] It is worth noting that, in this embodiment of the disclosure, the training service of the AI ​​model responds to the first information request by the first node itself, and provides the response result of the training service.

[0214] Step S2104: Select the fourth node from the first node.

[0215] In some embodiments, the fourth node is used to provide the response results of the training service.

[0216] It should be noted that the fourth node can be the executor of the training service.

[0217] It is worth noting that the response results of the training service are provided by the fourth node, and the first node selects the appropriate fourth node based on the first information.

[0218] In this embodiment of the disclosure, the communication system may include multiple fourth nodes, wherein different fourth nodes may have different capabilities to respond to training services. Therefore, when selecting a fourth node, the first node needs to select a fourth node with suitable capabilities based on the training service requested by the first information.

[0219] In some embodiments, the fourth node of the response training service includes at least one of the following:

[0220] The first network function is a core network function with first and second capabilities.

[0221] The second network function is a core network function that has the first capability.

[0222] The first capability is the ability to train AI models; the second capability is different from the first capability.

[0223] In some embodiments, the second capability may be a capability required to process any service of the core network. For example, the second capability may be a core network capability such as session management capability, access management capability, or mobility management capability.

[0224] It should be noted that the first network function can be an existing network function in the communication system. This embodiment of the disclosure provides AI model training services by reusing the existing first network function of the communication system. It is understood that the first network function, in addition to performing its original function, can also be used to train the AI ​​model. For example, the first network function can be an LMF with training capabilities. Another example is a Network Repository Function (NRF) with training capabilities.

[0225] The second network function can be a newly defined network function in a communication system, specifically designed for handling AI model training services within the core network. It's understood that in a communication system, the second network function can only be used for AI model training services and cannot perform other core network functions.

[0226] It is worth noting that in the embodiments of this disclosure, both the first network function and the second network function can respond to the training service of the AI ​​model; however, the processing capabilities of the first network function and the second network function are different.

[0227] In some embodiments, the method further includes: receiving status information sent by a fourth node, the status information including at least one of the following: capability information, used to indicate the processing capability of the fourth node; load information, used to indicate the load intensity of the fourth node.

[0228] It should be noted that, in this embodiment of the disclosure, the fourth node in the communication system can periodically send status information to the first node to assist the first node in selecting a fourth node to respond to the training service. Alternatively, upon receiving the first information, the first node can publish task information related to the training service; after receiving the task information, the fourth node in the communication system can send status information to the first node.

[0229] In some embodiments, the first node selecting the fourth node includes: selecting one or more fourth nodes that respond to the training service based on at least one of the request information for the training service and the status information of the candidate nodes.

[0230] It should be noted that the first node can select the fourth node to respond to the training service based on the request information. For example, the first node selects the fourth node to respond to the training service based on the training method of the AI ​​model indicated by the fourth information. If the fourth information indicates that the AI ​​model is trained using federated learning, the first node can select multiple first network functions as the fourth node to respond to the AI ​​service.

[0231] Alternatively, the first node can select a fourth node to respond to the training service based on the status information of the candidate nodes. For example, the first node can select a fourth node whose capability information matches the requested training service based on the capability information of the candidate nodes.

[0232] Alternatively, the first node can select a fourth node to respond to the training service based on the demand information and the status information of the candidate nodes. For example, the first node can select a fourth node to respond to the training service based on the inference method of the AI ​​model indicated by the fourth information and the load information of the candidate nodes.

[0233] Step S2105: The first node sends the third information to the fourth node.

[0234] In some embodiments, the fourth node receives third information sent by the first node.

[0235] In some embodiments, the third information is used to indicate the response result of the fourth node providing the training service.

[0236] In some embodiments, the first node sends third information to the fourth node based on the first information.

[0237] It should be noted that after the fourth node of the response training service is determined, the first node can determine the third information based on the first information and send the third information to the fourth node.

[0238] In some embodiments, sending third information to the fourth node based on the first information further includes:

[0239] Based on the determined training method of the AI ​​model, assign training tasks of the AI ​​model to one or more selected fourth nodes;

[0240] Send third information to one or more fourth nodes. The third information includes the training task of the assigned AI model.

[0241] It should be noted that the number of fourth nodes required to train an AI model varies depending on the training method. Therefore, the first node selects the corresponding number of fourth nodes based on the determined training method of the AI ​​model and assigns the AI ​​model training tasks to the fourth nodes; the assigned AI model training tasks are distributed to the fourth nodes by sending third information.

[0242] In some embodiments, the third information includes the model parameters and training data of the AI ​​model to be trained.

[0243] Understandably, the first node sends the model parameters and training data of the AI ​​model to be trained to the fourth node, so that the fourth node can use the training data to train the model parameters of the AI ​​model to be trained, thereby obtaining a trained AI model.

[0244] Here, the training data in the third information can be the training data carried in the first information.

[0245] In some embodiments, sending third information to the fourth node based on the first information includes:

[0246] If the first information does not contain training data or if the first information includes the fifth information, then the training data for the AI ​​model is collected.

[0247] Send a third message to the fourth node, which includes the collected training data.

[0248] It should be noted that, in the case where the first information does not contain training data, or in the case where the first information includes the fifth information, in order to ensure that the training of the AI ​​model can proceed normally, the first node needs to collect the training data required to train the AI ​​model.

[0249] Here, the training data in the third information can be the training data collected by the first node.

[0250] In some embodiments, training data for the AI ​​model is collected, including:

[0251] The first node determines the data source node based on the first piece of information;

[0252] The first node sends a seventh message to the data source node, which instructs the data source node to provide training data.

[0253] It should be noted that during the process of the first node collecting training data for the AI ​​model to be trained, the first node can first determine the data type of the training data of the AI ​​model to be trained based on the first information; and determine the data source node based on the data type of the training data, so that the first node can collect training data from the determined data source node.

[0254] In some embodiments, the method further includes:

[0255] The fourth node performs training service operations based on the third information.

[0256] Understandably, after receiving the third information, the fourth node can perform the training operation of the AI ​​model based on the training data and the model data of the AI ​​model to be trained in the third information, so as to obtain the model parameters of the trained AI model.

[0257] Step S2106: The fourth node sends the response result of the training service to the first node.

[0258] In some embodiments, the first node receives the response result of the training service sent by the fourth node.

[0259] In some embodiments, the response result of the training service includes: model parameters of the AI ​​model that has completed training.

[0260] It should be noted that after receiving the third information, the fourth node trains the AI ​​model based on the third information to obtain the model parameters of the trained AI model; and sends the model parameters of the trained AI model as the training result to the first node, so that the first node can deliver the training result to the third node through the second node.

[0261] In some embodiments, the training results may also include metadata of the trained AI model.

[0262] Understandably, after the fourth node completes the training of the AI ​​model, it can send the metadata of the trained AI model to the first node, so that the first node can manage the trained AI model based on the metadata of the trained AI model.

[0263] Step S2107: The first node sends the second information to the second node.

[0264] In some embodiments, the second node receives second information sent by the first node.

[0265] In some embodiments, the second information includes the response results of the training service.

[0266] In some embodiments, the second information may be a training service response.

[0267] In some embodiments, the first node sending second information to the second node includes: the first node sending second information to the second node based on the response result provided by the fourth node.

[0268] In some embodiments, the second information includes at least one of the following: model parameters of the trained AI model; metadata of the trained AI model.

[0269] It should be noted that the first node sends the response results of the training service, namely the model parameters and / or metadata of the AI ​​model that has completed training, to the second node through the second information.

[0270] In some embodiments, the second information further includes: training data collected by the first node.

[0271] It should be noted that, since the training data used to train the AI ​​model may differ, the model parameters of the trained AI model may also differ. Therefore, if the first information does not include training data or includes the fifth information, since the training data for training the AI ​​model is collected by the first node, the first node can also send the collected training data to the second node through the second information when performing the delivery operation of the training results.

[0272] Step S2108: The second node sends the second information to the third node.

[0273] In some embodiments, the third node receives the second information sent by the second node.

[0274] Understandably, the second node sends a second message to the third node to relay the response results of the training service provided by the first node. This allows the response results of the training service provided by the first node to be made available to the external third node.

[0275] In some embodiments, the term "information" may be used interchangeably with terms such as "message," "signal," "signaling," "report," "configuration," "indication," "instruction," "command," "channel," "parameter," "field," and "data."

[0276] In some embodiments, the term "send" may be used interchangeably with terms such as "transmit," "report," or "transmit."

[0277] The model training method disclosed herein may include at least one of steps S2101 to S2108. For example, step S2101 may be implemented as a standalone embodiment, and steps S2101 to S2103, as well as steps S2107 and S2108, may be implemented as standalone embodiments, but are not limited thereto.

[0278] In some embodiments, steps S2102, S2103, S2104, S2105, and S2106 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that if the second node determines that the third node does not have permission to obtain training services, the second node may reject the third node's AI model training request.

[0279] In some embodiments, steps S2104, S2105, and S2106 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that the first node itself can respond to the AI ​​model training service, and it is not necessary to select a fourth node to respond to the AI ​​model training service.

[0280] Figure 2B is a schematic diagram of an interactive model training method according to an exemplary embodiment. As shown in Figure 2B, this disclosure relates to a model training method for a communication system 100, the method comprising:

[0281] Step S2201: The third node sends the first information to the first node.

[0282] In some embodiments, the first node receives first information sent by the third node.

[0283] In some embodiments, the third node is deployed within the trust domain of the first node.

[0284] Understandably, the third node is a trusted node; the second node does not need to participate in the process of the third node requesting training services for the AI ​​model.

[0285] In some embodiments, the first information is used by the third node to request AI services.

[0286] In some embodiments, the first information is used by the third node to request training services for the AI ​​model.

[0287] In some embodiments, the first information may be a training request for an AI model.

[0288] It should be noted that the third node can be a consumer of AI services.

[0289] In some embodiments, the third node may include at least one of the following: AF; AS; core network function; access network function.

[0290] It should be noted that the third node can be a third-party AF or AS; or, the third node can be any network function in the network. The AI ​​model training service provided by the first node can be made available to any network function in the network or to a third-party AF or AS.

[0291] It is worth noting that while the third node is a consumer of AI services, it can also be a provider of other business services.

[0292] In some embodiments, the first information may include at least one of the following: the AI ​​model to be trained; the training data of the AI ​​model to be trained; the metadata of the AI ​​model to be trained; and the fifth information, used to request the first node to collect training data.

[0293] Here, the first information includes the AI ​​model to be trained, which may be the model parameters of the AI ​​model to be trained carried in the first information, so that the first node can determine the AI ​​model to be trained based on the model parameters in the first information.

[0294] Model parameters include at least one of the following: model features, model structure, and weights. Model features refer to the data attributes or dimensions input into the AI ​​model. Model structure indicates the optimization algorithm, loss function, and network structure used by the AI ​​model. Weights are parameters indicating the connections of each neuron or layer in the model.

[0295] It's important to note that AI model parameters can be divided into two categories: trainable parameters (e.g., weights) and non-trainable parameters (e.g., model structure). Trainable parameters are learned through training. Non-trainable parameters are fixed and do not update during model training.

[0296] In some embodiments, the model parameters of the AI ​​model to be trained may include at least one of the following: non-training parameters; initial parameter values ​​for trainable parameters.

[0297] In some embodiments, training data is used for training, validating, and testing the AI ​​model to be trained.

[0298] It should be noted that training data is one of the most important parts of training an AI model; training an AI model refers to using the training data to make the AI ​​model fit the patterns in the training data through certain methods, and determining the model parameters (i.e., training parameters) of the AI ​​model.

[0299] In some embodiments, the metadata of the AI ​​model to be trained is used to indicate the model attributes of the AI ​​model to be trained.

[0300] It's important to note that metadata is data that describes various attributes and information about an AI model. Metadata helps in understanding and managing AI models.

[0301] In some embodiments, metadata may include at least one of the following: the model version of the AI ​​model; the model identifier of the AI ​​model; the model size of the AI ​​model; the vendor identifier of the AI ​​model; and the training time information of the AI ​​model.

[0302] In some embodiments, the fifth information is used to trigger the first node to collect training data for the AI ​​model to be trained.

[0303] In some embodiments, the fifth information is used to instruct the training of the AI ​​model to be trained using network data.

[0304] It should be noted that, in cases where the first information includes the fifth information, the first node is responsible for collecting the training data required to train the AI ​​model.

[0305] In some embodiments, the first information further includes a sixth information, which is used to indicate the third node.

[0306] Here, the sixth information can be identification information indicating the third node, or it can be other information that can be used to indicate the identity of the third node.

[0307] In some embodiments, the sixth information is also used to determine whether the third node is authorized to obtain training services.

[0308] It should be noted that after receiving the first message from the third node, the first node needs to perform authorization verification on the third node based on the sixth message within the first message to determine whether the third node has the authority to obtain training services.

[0309] Step S2202: The first node responds to the training service of the first information request.

[0310] In some embodiments, optional implementations of step S2202 can be found in optional implementations of step S2103 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0311] Step S2203: Select the fourth node for the first node.

[0312] In some embodiments, optional implementations of step S2203 can be found in optional implementations of step S2104 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0313] Step S2204: The first node sends the third information to the fourth node.

[0314] In some embodiments, optional implementations of step S2204 can be found in optional implementations of step S2105 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0315] Step S2205: The fourth node sends the response result of the training service to the first node.

[0316] In some embodiments, optional implementations of step S2205 can be found in optional implementations of step S2106 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0317] Step S2206: The first node sends the second information to the third node.

[0318] In some embodiments, the third node receives the second information sent by the first node.

[0319] In some embodiments, the second information includes the response results of the training service. This allows the response results of the training service provided by the first node to be made available to an external third node.

[0320] In some embodiments, the second information may be a training service response.

[0321] In some embodiments, the first node sending second information to the third node includes: the first node sending second information to the third node based on the response result provided by the fourth node.

[0322] In some embodiments, the second information includes at least one of the following: model parameters of the trained AI model; metadata of the trained AI model.

[0323] It should be noted that the first node sends the response results of the training service, namely the model parameters and / or metadata of the AI ​​model that has completed training, to the third node through the second information.

[0324] In some embodiments, the second information further includes: training data collected by the first node.

[0325] It should be noted that the model parameters of the trained AI model may be different depending on the training data used to train the AI ​​model. Therefore, if the first information does not include training data or includes the fifth information, since the training data for training the AI ​​model is collected by the first node, the first node can also send the collected training data to the third node through the second information when performing the delivery operation of the training results.

[0326] The model training method disclosed herein may include at least one of steps S2201 to S2206. For example, step S2201 may be implemented as a standalone embodiment, and steps S2201, S2202, and S2206 may be implemented as standalone embodiments, but are not limited thereto.

[0327] In some embodiments, steps S2202, S2203, S2204, S2205, and S2206 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that if the first node determines that the third node does not have permission to obtain training services, the first node may reject the third node's AI model training request.

[0328] In some embodiments, steps S2203, S2204, and S2205 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that the first node itself can respond to the AI ​​model training service, and it is not necessary to select a fourth node to respond to the AI ​​model training service.

[0329] Figure 3A is a flowchart illustrating a model training method according to an exemplary embodiment. As shown in Figure 3A, this embodiment relates to a model training method, executed by a first node, the method including:

[0330] Step S3101: Receive the first information.

[0331] In some embodiments, optional implementations of step S3101 can refer to optional implementations of step S2102 in FIG2A and optional implementations of step S2201 in FIG2B, as well as other related parts in the embodiments involved in FIG2A and other related parts in the embodiments involved in FIG2B, which will not be repeated here.

[0332] Step S3102: Respond to the training service in response to the first information request.

[0333] In some embodiments, optional implementations of step S3102 can be found in optional implementations of step S2103 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0334] Step S3103: Select the fourth node.

[0335] In some embodiments, optional implementations of step S3103 can be found in optional implementations of step S2104 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0336] Step S3104: Send the third message.

[0337] In some embodiments, optional implementations of step S3104 can be found in optional implementations of step S2105 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0338] Step S3105: Receive the response result from the training service.

[0339] In some embodiments, optional implementations of step S3105 can be found in optional implementations of step S2106 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0340] Step S3106: Send the second message.

[0341] In some embodiments, optional implementations of step S3106 can refer to optional implementations of step S2107 in FIG2A and optional implementations of step S2206 in FIG2B, as well as other related parts in the embodiments involved in FIG2A and other related parts in the embodiments involved in FIG2B, which will not be repeated here.

[0342] The model training method disclosed herein may include at least one of steps S3101 to S3106. For example, steps S3101, S3102, and S3106 may be implemented as independent embodiments, but are not limited thereto.

[0343] In some embodiments, steps S3103, S3104, S3105, and S3106 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that the first node itself can respond to the AI ​​model training service, and it is not necessary to select a fourth node to respond to the AI ​​model training service.

[0344] In some embodiments, steps S3202, S3203, S3204, and S3205 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that if the first node determines that the third node does not have permission to obtain training services, the first node may reject the third node's AI model training request.

[0345] In some embodiments, steps S3203 and S3204 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that the first node itself can respond to the AI ​​model training service, and it is not necessary to select a fourth node to respond to the AI ​​model training service.

[0346] Figure 3B is a flowchart illustrating a model training method according to an exemplary embodiment. As shown in Figure 3C, this disclosure relates to a model training method, executed by a first node, the method including:

[0347] Step S3201: Receive the first information.

[0348] In some embodiments, the first information is used by the third node to request training services for the AI ​​model.

[0349] The optional implementation of step S3201 can be found in the optional implementation of step S2102 in Figure 2A and the optional implementation of step S2201 in Figure 2B, as well as other related parts in the embodiments involved in Figure 2A and the embodiments involved in Figure 2B, which will not be repeated here.

[0350] Step S3202: Respond to the training service in response to the first information request.

[0351] The optional implementation of step S3202 can be found in the optional implementation of step S2103 in Figure 2A and the optional implementation of step S2202 in Figure 2B, as well as other related parts in the embodiments involved in Figure 2A and the embodiments involved in Figure 2B, which will not be repeated here.

[0352] Step S3203: Send the second message.

[0353] In some embodiments, the second information includes the response results provided to the third node.

[0354] The optional implementation of step S3203 can be found in the optional implementation of step S2107 in Figure 2A and the optional implementation of step S2206 in Figure 2B, as well as other related parts in the embodiments involved in Figure 2A and the embodiments involved in Figure 2B, which will not be repeated here.

[0355] Figure 4 is a flowchart illustrating a model training method according to an exemplary embodiment. As shown in Figure 4, this embodiment relates to a model training method, executed by a second node, the method including:

[0356] Step S4101: Receive the first information.

[0357] In some embodiments, optional implementations of step S4101 can be found in optional implementations of step S2101 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0358] Step S4102: Send the first message.

[0359] In some embodiments, optional implementations of step S4102 can be found in optional implementations of step S2102 in FIG2A and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0360] Step S4103: Receive the second information.

[0361] In some embodiments, optional implementations of step S4103 can be found in optional implementations of step S2107 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0362] Step S4104: Send the second message.

[0363] In some embodiments, optional implementations of step S4104 can be found in optional implementations of step S2108 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0364] The model training method disclosed in this embodiment may include at least one of steps S4101 to S4104. For example, step S4101 may be implemented as a standalone embodiment, but is not limited thereto.

[0365] In some embodiments, steps S4102, S4103, and S4104 are optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that if the second node determines that the third node does not have permission to obtain training services, the second node may reject the third node's AI model training request without sending the first information to the first node.

[0366] Figure 5 is a flowchart illustrating a model training method according to an exemplary embodiment. As shown in Figure 5, this embodiment relates to a model training method executed by a third node, the method including:

[0367] Step S5101: Send the first message.

[0368] In some embodiments, optional implementations of step S5101 may refer to optional implementations of step S2101 in FIG2A and optional implementations of step S2201 in FIG2B, as well as other related parts in the embodiments involved in FIG2A and other related parts in the embodiments involved in FIG2B, which will not be repeated here.

[0369] Step S5102: Receive the second information.

[0370] In some embodiments, optional implementations of step S5102 may refer to optional implementations of step S2108 in FIG2A and optional implementations of step S2206 in FIG2B, as well as other related parts in the embodiments involved in FIG2A and other related parts in the embodiments involved in FIG2B, which will not be repeated here.

[0371] The model training method disclosed in this embodiment may include at least one of steps S5101 to S5102. For example, step S5101 may be implemented as a standalone embodiment, but is not limited thereto.

[0372] In some embodiments, step S5102 is optional, and one or more of these steps may be omitted or substituted in different embodiments. It is understood that if the second node determines that the third node does not have permission to obtain training services, the second node may reject the third node's AI model training request.

[0373] Figure 6 is a flowchart illustrating a model training method according to an exemplary embodiment. As shown in Figure 6, this embodiment relates to a model training method, executed by a fourth node, the method including:

[0374] Step S6101: Receive third information.

[0375] In some embodiments, optional implementations of step S6101 can be found in optional implementations of step S2105 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0376] Step S6102: Send the response result of the training service.

[0377] In some embodiments, optional implementations of step S6102 can be found in optional implementations of step S2106 in FIG2A, and other related parts in the embodiments involved in FIG2A, which will not be repeated here.

[0378] Figure 7 is an interactive schematic diagram of a model training method according to an exemplary embodiment. As shown in Figure 7, this disclosure relates to a model training method for a communication system 100, the method including one of the following steps:

[0379] Step S7101: The first node receives the first information sent by the second or third node.

[0380] In some embodiments, the first information is used by the third node to request training services for the AI ​​model;

[0381] Step S7102: The first node responds to the training service of the first information request.

[0382] Step S7103: The first node sends the second information to the second or third node.

[0383] In some embodiments, the second information includes the response result provided to the third node.

[0384] In some embodiments, the above methods may include the methods of the above-described communication system side, first node side, second node side, third node side, fourth node side, etc., which will not be described in detail here.

[0385] This disclosure proposes a unified framework for providing AI as a service in a 6G system, including three AI-related network functions.

[0386] (1) AI Model Storage Service Function (AiSSF) provides AI model storage services, including AI model management, such as uploading, downloading or deleting AI models.

[0387] (2) AI model inference service function (AiISF) provides AI model inference services, including AI inference node selection and AI model selection.

[0388] (3) AI Model Training Service Function (AiTSF) provides AI model training services, including training method selection and training node selection.

[0389] This architecture also defines a new type of 6G NF.

[0390] (1) 6G NF with reasoning capability supports AI model reasoning, such as LMF with reasoning capability, for calculating the location of the terminal.

[0391] (2) 6G NF with training capability, supporting AI model training, such as LMF with training capability, used to train AI models.

[0392] As shown in Figure 8A, Figure 8A is a schematic diagram of a service-based AI-as-a-Service architecture in a 6G system according to an exemplary embodiment. AiSSF is an AI-related network function responsible for storing AI models. Here, AiSSF corresponds to the network function participating in the storage service of AI models in this disclosure.

[0393] AiSSF provides the following services:

[0394] (1) Uploading AI Models; The training NF of the AI ​​model can upload the trained AI model to AiSSF. Optionally, the AF or UE can upload the AI ​​model to AiSSF after passing the necessary authentication or authorization.

[0395] (2) AI model download operation; the AI ​​model's inference NF (e.g., AnLF) or 6G NF with inference capability can download the AI ​​model for inference. Optionally, the UE or AI can download the AI ​​model from AiSSF after passing the necessary authentication or authorization.

[0396] (3) Deletion of AI model; Authorized AF, UE or 6G NF may request AiSSF to perform the deletion of AI model.

[0397] AiISF is an AI-related network function responsible for inference services for AI models. Here, AiISF corresponds to the network function involved in the inference service of AI models in this disclosure.

[0398] AiISF can provide the following services:

[0399] (1) Inference operations of AI models; AiISF performs inference on AI models through input data and provides inference results. Inference operations include: selecting inference methods and inference nodes, delivering inference results, etc.

[0400] (2) Delivery of inference results; performed by AiISF, delivering inference results to one or a group of consumers based on requests, service requirements or policies.

[0401] AiTSF is an AI-related network function responsible for training AI models. Here, AiTSF corresponds to the network function in this disclosure that participates in the training service of AI models.

[0402] AiTSF can provide the following services:

[0403] (1) Training operations for AI models; executed by AiTSF, providing training operations for specific AI models, including selecting training methods and training nodes.

[0404] (2) Training data collection operation; performed by AiTSF, collecting data from other data source nodes (e.g., AF, NF or UE); the collection operation may include the selection of data source nodes.

[0405] (3) Management and operation of trained AI models; AiTSF processes trained AI models, including uploading them to AiSSF and delivering them to consumers who need them.

[0406] The 6G NF1 with inference capability is a 6G network function with AI model inference capability, such as an LMF with the ability to calculate the terminal location using an inference AI positioning model. Here, the 6G NF1 with inference capability corresponds to the third network function in this disclosure.

[0407] The 6G NF2 with training capabilities is a 6G network function capable of training AI models. For example, in federated learning training, this NF can serve as a training node for training AI models. Here, the 6G NF2 with training capabilities corresponds to the first network function in this disclosure.

[0408] 6G AI Training NF is a dedicated network function for training AI models, offering high performance. Here, 6G AI Training NF corresponds to the second network function in this disclosure.

[0409] The 6G Exposure Function is a network function that exposes the services provided by the 6G system.

[0410] In some embodiments, AiSST is connected to the communication bus via a first interface Naix1; AiISF is connected to the communication bus via a second interface Naix2; AiTSF is connected to the communication bus via a third interface Naix3; the 6G AI training NF is connected to the communication bus via a fourth interface Naix4; the 6G NF2 with training capability is connected to the communication bus via a fifth interface Naix5; and the 6G NF1 with inference capability is connected to the communication bus via a sixth interface Naix6.

[0411] In some embodiments, the 6G system further includes: 6G UDR, or 6G ADRF. It should be noted that 6G UDR or 6G ADRF are 6G network functions with AI model storage capabilities.

[0412] The 6G UDR or 6G ADRF connects to the communication bus via the seventh interface, Naix7.

[0413] In some embodiments, the 6G system further includes a 6G open function, which is connected to the communication bus via an eighth interface Nxx.

[0414] It should be noted that for AI-related network functions or 6G NFs, AI-related services are provided or consumed through newly defined interfaces such as Naix1, Naix2, Naix3, Naix4, Naix5, Naix6, and Nxx. For example, the Naix1 interface is used to transmit control plane signaling or user plane data.

[0415] This disclosure proposes a framework for providing AI as a service using a 6G system, which provides AI model training as a service to licensed AFs via a 6G NEF.

[0416] As shown in Figure 8B, Figure 8B is a schematic diagram illustrating a process for providing AI model training as a service in a 6G system according to an exemplary embodiment.

[0417] In step S8101, the third node sends a training request for the AI ​​model to the second node.

[0418] Here, the training request for the AI ​​model is the first piece of information disclosed. The training request for the AI ​​model may include the AI ​​model, training data, and model metadata.

[0419] In some embodiments, model metadata includes model identifier, vendor identifier, size, version, etc.

[0420] In some embodiments, the third node may be AF or AS. The second node may be 6G NEF.

[0421] Step S8102: The second node performs authorization verification on the third node.

[0422] Understandably, 6G NEF authorizes and verifies training requests for AI models from AF / AS.

[0423] It should be noted that the 6G NEF and 6G UDM perform authorization verification on the AF / AS to determine whether to accept or reject training requests from the AF / AS.

[0424] In step S8103, if the authorization verification is successful, the second node sends a training request for the AI ​​model to the first node.

[0425] Here, the first node can be AiTSF. Understandably, the 6G NEF sends AI model training requests to the AiTSF, which is responsible for providing AI model training services.

[0426] Step S8104: The first node selects the fourth node that provides training services.

[0427] Here, the fourth node can be a 6G NF2 with training capabilities and / or a 6G AI training NF.

[0428] Based on the training request, AiTSF determines the training method for the AI ​​model and selects the fourth node to provide training services.

[0429] In some embodiments, the training request does not include training data, or the AF / AS explicitly instructs the use of network data for training the AI ​​model, and AiTSF will be triggered to collect training data in the 6G system or AF.

[0430] In step S8105, the fourth node sends the response result of the training service to the first node.

[0431] After the fourth node trains the AI ​​model, it can upload the trained AI model as the response result of the training service to AiTSF.

[0432] Step S8106: The first node sends the training response of the AI ​​model to the second node.

[0433] AiTSF sends the response results of the training service to 6G NEF, so that 6G NEF can send the response results of the training service to AF / AS.

[0434] In some embodiments, the response from the training service may include: the trained AI model and the metadata of the trained AI model.

[0435] In step S8107, the second node sends the training response of the AI ​​model to the third node.

[0436] This disclosure also provides apparatus for implementing any of the above methods. For example, an apparatus is provided that includes units or modules for implementing the steps performed by the terminal in any of the above methods. Alternatively, another apparatus is provided that includes units or modules for implementing the steps performed by a network device (e.g., an access network device, or a core network device) in any of the above methods.

[0437] It should be understood that the division of units or modules in the above device is only a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, the units or modules in the device can be implemented by a processor calling software: for example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules in the above device. The processor can be, for example, a general-purpose processor, such as a Central Processing Unit (CPU) or a microprocessor, and the memory can be internal or external to the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits. The functionality of some or all of the units or modules can be achieved through the design of these hardware circuits, which can be understood as one or more processors. For example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC). The functionality of some or all of the units or modules is achieved through the design of the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a programmable logic device (PLD). Taking a field-programmable gate array (FPGA) as an example, it can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files, thereby achieving the functionality of some or all of the units or modules. All units or modules of the above device can be implemented entirely through processor-called software, entirely through hardware circuits, or partially through processor-called software with the remaining parts implemented through hardware circuits.

[0438] In this embodiment, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction read and execute capabilities, such as a Central Processing Unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a type of microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. The logical relationships of the aforementioned hardware circuits are fixed or reconfigurable. For example, the processor is a hardware circuit implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a Neural Network Processing Unit (NPU), a Tensor Processing Unit (TPU), a Deep Learning Processing Unit (DPU), etc.

[0439] Figure 9A is a schematic diagram of the structure of a first node according to an exemplary embodiment. As shown in Figure 9A, the first node 101 includes: a first receiving module 1011, a processing module 1012, and a first sending module 1013; the first receiving module 1011 is configured to receive first information, which is used by a third node to request training services for an artificial intelligence (AI) model; the processing module 1012 is configured to respond to the training service requested by the first information; the first sending module 1013 is configured to send second information, which includes a response result provided to the third node. Optionally, the first receiving module 1011 is used to perform the steps related to information reception performed by the first node in any of the above model training methods, which will not be described in detail here. Optionally, the processing module 1012 is used to perform the steps related to information processing performed by the first node in any of the above model training methods, which will not be described in detail here. Optionally, the first sending module 1013 is used to perform the steps related to information transmission performed by the first node in any of the above model training methods, which will not be described in detail here.

[0440] Figure 9B is a schematic diagram of the structure of a second node according to an exemplary embodiment. As shown in Figure 9B, the second node 102 includes: a second receiving module 1021 and a second sending module 1022; the second receiving module 1021 is configured to receive first information sent by a third node, the first information being used by the third node to request training services for an AI model; the second sending module 1022 is configured to send the first information to a first node responding to the training service; the second receiving module 1021 is also configured to receive second information sent by the first node, the second information including a response result provided to the third node; the second sending module 1022 is also configured to send the second information to the third node. Optionally, the second receiving module 1021 is used to perform the steps related to information reception performed by the second node in any of the above model training methods, which will not be described in detail here. Optionally, the second sending module 1022 is used to perform the steps related to information transmission performed by the second node in any of the above model training methods, which will not be described in detail here. Optionally, the second node 102 further includes a processing module, which is used to perform the information processing-related steps performed by the second node in any of the above methods, which will not be described in detail here.

[0441] Figure 9C is a schematic diagram illustrating the structure of a third node according to an exemplary embodiment. As shown in Figure 9C, the third node 103 includes a third sending module 1031 and a third receiving module 1032. The third sending module 1031 is configured to send first information, which is used by the third node to request training services for an AI model; the third receiving module 1032 is configured to receive second information, which includes a response result provided to the third node. Optionally, the third sending module 1031 is used to perform the steps related to information sending performed by the third node in any of the above model training methods, which will not be described in detail here. Optionally, the third receiving module 1032 is used to perform the steps related to information receiving performed by the third node in any of the above model training methods, which will not be described in detail here. Optionally, the third node 103 also includes a processing module, which is used to perform the steps related to information processing performed by the third node in any of the above methods, which will not be described in detail here.

[0442] Figure 9D is a schematic diagram illustrating the structure of a fourth node according to an exemplary embodiment. As shown in Figure 9D, the fourth node 104 includes a fourth receiving module 1041 and a fourth sending module 1042. The fourth receiving module 1041 is configured to receive third information sent by the first node based on first information; the first information is used by the third node to request training services for the AI ​​model; the fourth sending module 1042 is configured to send the response result of the training service to the first node. Optionally, the fourth receiving module 1041 is used to perform the steps related to information reception performed by the fourth node in any of the above model training methods, which will not be described in detail here. Optionally, the fourth sending module 1042 is used to perform the steps related to information transmission performed by the fourth node in any of the above model training methods, which will not be described in detail here. Optionally, the fourth node 104 also includes a processing module, which is used to perform the steps related to information processing performed by the fourth node in any of the above methods, which will not be described in detail here.

[0443] Figure 10A is a schematic diagram of a communication device according to an exemplary embodiment. The communication device 1100 can be a first node, a second node, a third node, or a fourth node; it can also be a chip, chip system, or processor that supports network devices in implementing any of the above methods; or it can be a chip, chip system, or processor that supports terminals in implementing any of the above model training methods. The communication device 1100 can be used to implement the model training methods described in the above method embodiments, as detailed in the descriptions within the above method embodiments.

[0444] As shown in Figure 10A, the communication device 1100 includes one or more processors 1101. The processor 1101 can be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit (CPU). The baseband processor can be used to process communication protocols and communication data, while the CPU can be used to control communication devices (e.g., base stations, baseband chips, terminal devices, terminal device chips, DUs or CUs, etc.), execute programs, and process program data. The processor 1101 is used to invoke instructions to cause the communication device 1100 to execute any of the above communication methods.

[0445] In some embodiments, the communication device 1100 further includes one or more memories 1102 for storing instructions. Optionally, all or part of the memories 1102 may also be located outside the communication device 1100.

[0446] In some embodiments, the communication device 1100 further includes one or more transceivers 1103. When the communication device 1100 includes one or more transceivers 1103, the communication steps such as sending and receiving in the above method are performed by the transceivers 1103, and other steps are performed by the processor 1101.

[0447] In some embodiments, a transceiver may include a receiver and a transmitter, which may be separate or integrated. Optionally, the terms transceiver, transceiver unit, transceiver, transceiver circuit, etc., may be used interchangeably; the terms transmitter, transmitting unit, transmitter, transmitting circuit, etc., may be used interchangeably; and the terms receiver, receiving unit, receiver, receiving circuit, etc., may be used interchangeably.

[0448] Optionally, the communication device 1100 further includes one or more interface circuits 1104, which are connected to the memory 1102. The interface circuits 1104 can be used to receive signals from the memory 1102 or other devices, and can be used to send signals to the memory 1102 or other devices. For example, the interface circuits 1104 can read instructions stored in the memory 1102 and send the instructions to the processor 1101.

[0449] The communication device 1100 described in the above embodiments may be a network device or a terminal, but the scope of the communication device 1100 described in this disclosure is not limited thereto, and the structure of the communication device 1100 may not be limited by FIG10A. The communication device may be a standalone device or a part of a larger device. For example, the communication device may be: (1) a standalone integrated circuit IC, or chip, or chip system or subsystem; (2) a collection of one or more ICs, optionally, the IC collection may also include storage components for storing data and programs; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, terminal device, smart terminal device, cellular phone, wireless device, handheld device, mobile unit, vehicle device, network device, cloud device, artificial intelligence device, etc.; (6) others, etc.

[0450] Figure 10B is a schematic diagram of a chip structure according to an exemplary embodiment. For cases where the communication device 1100 can be a chip or a chip system, please refer to the schematic diagram of the chip 1200 shown in Figure 10B, but it is not limited thereto.

[0451] Chip 1200 includes one or more processors 1201, which are used to invoke instructions to cause chip 1200 to execute any of the above communication methods.

[0452] In some embodiments, chip 1200 further includes one or more interface circuits 1202 connected to memory 1203. Interface circuits 1202 can be used to receive signals from memory 1203 or other devices, and can also be used to send signals to memory 1203 or other devices. For example, interface circuit 1202 can read instructions stored in memory 1203 and send those instructions to processor 1201. Optionally, terms such as interface circuit, interface, transceiver pin, and transceiver can be used interchangeably.

[0453] In some embodiments, chip 1200 further includes one or more memories 1203 for storing instructions. Optionally, all or part of the memories 1203 may be located outside of chip 1200.

[0454] This disclosure also provides a storage medium storing instructions that, when executed on a communication device 1100, cause the communication device 1100 to perform any of the methods described above. Optionally, the storage medium is an electronic storage medium. Optionally, the storage medium is a computer-readable storage medium, but it can also be a storage medium readable by other devices. Optionally, the storage medium can be a non-transitory storage medium, but it can also be a temporary storage medium.

[0455] This disclosure also provides a program product, which, when executed by the communication device 1100, causes the communication device 1100 to perform any of the above communication methods. Optionally, the program product is a computer program product.

[0456] This disclosure also provides a computer program that, when run on a computer, causes the computer to perform any of the above communication methods.

[0457] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0458] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A model training method, wherein, Executed by the first node, the method includes: Receive the first information, which is used by the third node to request the training service of the artificial intelligence (AI) model; The training service responds to the first information request; Send a second message, which includes a response result provided to the third node.

2. The method according to claim 1, wherein, The receipt of the first information includes at least one of the following: If the third node is a trusted node, the first information sent by the third node is received; If the third node is not a trusted node, the first information sent by the second node is received.

3. The method according to claim 2, wherein, The training service in response to the first information request includes: Based on at least one of the request information for the training service and the status information of the candidate nodes, one or more fourth nodes are selected to respond to the training service, and the fourth nodes are used to provide the response results of the training service; based on the first information, third information is sent to the fourth nodes; Receive the response result sent by the fourth node based on the third information.

4. The method according to claim 3, wherein, The sending of the second information includes at least one of the following: Based on the response result provided by the fourth node, send the second information to the second node; Based on the response provided by the fourth node, a second message is sent to the trusted third node.

5. The method according to claim 3 or 4, wherein, The method further includes: Based on the first information, the request information for the training service is determined, and the request information includes: fourth information, which indicates the service parameters for providing the training service, and the service parameters include at least one of the following: the type of the AI ​​model and the training method of the AI ​​model.

6. The method according to claim 5, wherein, The step of sending the third information to the fourth node based on the first information further includes: Based on the determined training method of the AI ​​model, assign training tasks of the AI ​​model to one or more selected fourth nodes; Send third information to one or more fourth nodes, the third information including the assigned training task of the AI ​​model.

7. The method according to any one of claims 3 to 6, wherein, The method further includes: Receive the status information sent by the fourth node, the status information including at least one of the following: Capability information, used to indicate the processing capability of the fourth node; Load information is used to indicate the load intensity of the fourth node.

8. The method according to any one of claims 3 to 7, wherein, The operation of the training service includes at least one of the following: The training operation of the AI ​​model; The process of collecting training data; Delivery of training results.

9. The method according to any one of claims 3 to 8, wherein, The first information includes at least one of the following: The AI ​​model to be trained; The training data of the AI ​​model to be trained; The metadata of the AI ​​model to be trained, wherein the metadata is used to indicate the model attributes of the AI ​​model to be trained; The fifth piece of information is used to request the first node to collect the training data.

10. The method of claim 9, wherein the metadata comprises at least one of the following: The model version of the AI ​​model; The model identifier of the AI ​​model; The size of the AI ​​model; The supplier identifier of the AI ​​model; The training time information of the AI ​​model.

11. The method according to claim 9 or 10, wherein, The step of sending third information to the fourth node based on the first information includes: If the first information does not contain the training data or if the first information includes the fifth information, the training data of the AI ​​model is collected. The third information, which includes the collected training data, is sent to the fourth node.

12. The method according to any one of claims 9 to 11, wherein, The second information includes at least one of the following: The model parameters of the trained AI model; The metadata of the AI ​​model that has completed training.

13. The method according to any one of claims 3 to 12, wherein, The fourth node responding to the training service includes at least one of the following: The first network function is a core network function with first and second capabilities. The second network function is a core network function that has the first capability. The first capability is the training capability of the AI ​​model; the second capability is different from the first capability.

14. A model training method, wherein, Executed by the fourth node, the method includes: Receive third information sent by the first node based on the first information; the first information is used by the third node to request training services for the AI ​​model; Send the response result of the training service to the first node.

15. The method according to claim 14, wherein, The third information includes the training task of the AI ​​model assigned by the first node to the fourth node.

16. The method according to claim 14 or 15, wherein, The method further includes: Send status information to the first node, the status information including at least one of the following: Capability information, used to indicate the processing capability of the fourth node; Load information is used to indicate the load intensity of the fourth node.

17. The method according to any one of claims 14 to 16, wherein, The response results of the training service include at least one of the following: The model parameters of the trained AI model; The metadata of the AI ​​model that has completed training.

18. The method according to any one of claims 14 to 17, wherein, The fourth node includes at least one of the following: The first network function is a core network function with first and second capabilities. The second network function is a core network function that has the first capability. The first capability is the training capability of the AI ​​model; the second capability is different from the first capability.

19. A model training method, wherein, Executed by the second node, the method includes: Receive first information sent by a third node, the first information being used by the third node to request training services for an AI model; Send the first information to the first node responding to the training service; Receive second information sent by the first node, the second information including a response result provided to the third node; The second information is sent to the third node.

20. The method according to claim 19, wherein, Sending the first information to the first node responding to the training service includes: Based on the first information, it is determined whether the third node is authorized to obtain the training service; wherein, the first information includes a sixth information, which is used to instruct the third node; If it is determined that the third node is authorized to obtain the training service, the first information is sent to the first node.

21. The method according to claim 19 or 20, wherein, The first information includes at least one of the following: The AI ​​model to be trained; The training data of the AI ​​model to be trained; The metadata of the AI ​​model to be trained, wherein the metadata is used to indicate the model attributes of the AI ​​model to be trained; The fifth piece of information is used to request the first node to collect the training data.

22. The method according to claim 21, wherein, The metadata includes at least one of the following: The model version of the AI ​​model; The model identifier of the AI ​​model; The size of the AI ​​model; The supplier identifier of the AI ​​model; The training time information of the AI ​​model.

23. The method according to claim 21 or 22, wherein, The second information includes at least one of the following: The model parameters of the trained AI model; The metadata of the AI ​​model that has completed training.

24. A model training method, wherein, Executed by a third node, the method includes: Send a first message, which is used by the third node to request the training service of the AI ​​model; Receive second information, which includes a response result provided to the third node.

25. The method according to claim 24, wherein, The sending of the first information includes at least one of the following: If the third node is a trusted node, the first information is sent to the first node; If the third node is not a trusted node, the first information is sent to the second node.

26. The method of claim 25, wherein, The operation of the training service includes at least one of the following: The training operation of the AI ​​model; The process of collecting training data; Delivery of training results.

27. The method according to claim 25 or 26, wherein, The first information includes: The sixth information is used to indicate the third node, and the sixth information is also used to determine whether the third node is authorized to obtain the training service.

28. The method according to any one of claims 24 to 27, wherein, The first information includes at least one of the following: The AI ​​model to be trained; The training data of the AI ​​model to be trained; The metadata of the AI ​​model to be trained, wherein the metadata is used to indicate the model attributes of the AI ​​model to be trained; The fifth piece of information involves requesting the first node to collect the training data.

29. The method according to claim 28, wherein, The metadata includes at least one of the following: The model version of the AI ​​model; The model identifier of the AI ​​model; The size of the AI ​​model; The supplier identifier of the AI ​​model.

30. The method according to claim 28 or 29, wherein, The second information includes at least one of the following: The model parameters of the trained AI model; The metadata of the AI ​​model that has completed training.

31. The method according to any one of claims 24 to 30, wherein, The third node includes at least one of the following: Core network functions; Access network functions; Application functions; Application server.

32. A model training method, wherein, Performed by a communication system, the method includes: The first node receives first information sent by the second or third node, and the first information is used by the third node to request training services for the AI ​​model. The first node responds to the training service of the first information request; The first node sends second information to the second or third node, the second information including a response result provided to the third node.

33. A first node, wherein, The first node includes: The first receiving module is configured to receive first information, which is used by the third node to request training services for an artificial intelligence (AI) model. The processing module is configured to provide a training service in response to the first information request; The first sending module is configured to send second information, which includes a response result provided to the third node.

34. A second node, wherein, The second node includes: The second receiving module is configured to receive first information sent by the third node, the first information being used by the third node to request training services for the AI ​​model; The second sending module is configured to send the first information to the first node responding to the training service; The second receiving module is further configured to receive second information sent by the first node, the second information including a response result provided to the third node; The second sending module is also configured to send the second information to the third node.

35. A third node, wherein, The third node includes: The third sending module is configured to send first information, which is used by the third node to request training services for the AI ​​model. The third receiving module is configured to receive second information, which includes a response result provided to the third node.

36. A fourth node, wherein, The fourth node includes: The fourth receiving module is configured to receive third information sent by the first node based on the first information; the first information is used for the third... The node requests training services for the AI ​​model; The fourth sending module is configured to send the response result of the training service to the first node.

37. A communication system, wherein, The communication system includes a first node, a second node, a third node, and a fourth node; the first node is configured to implement the model training method of any one of claims 1 to 13, the fourth node is configured to implement the model training method of any one of claims 14 to 18, the second node is configured to implement the model training method of any one of claims 19 to 23, and the third node is configured to implement the model training method of any one of claims 24 to 31.

38. A communication device, wherein, The communication device includes: One or more processors; The processor is configured to invoke instructions to cause the communication device to execute the model training method according to any one of claims 1 to 13, 14 to 18, 19 to 23, and 24 to 31.

39. A storage medium, wherein, The storage medium stores instructions that, when executed on a communication device, cause the communication device to perform the model training method according to any one of claims 1 to 13, 14 to 18, 19 to 23, and 24 to 31.

40. A program product, wherein, When the program product is executed by a communication device, the communication device performs the model training method according to any one of claims 1 to 13, 14 to 18, 19 to 23, and 24 to 31.

Images