Communication method and apparatus

Abstract

A communication method and apparatus, which are applied to the technical field of communications. The method comprises: a first apparatus determining first information of a first model, wherein the first information comprises information of model training and / or information of model inference, the information of model training comprises at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the information of model inference comprises inference interpretability information; and sending the first information to a second apparatus, wherein the second apparatus is a consumer of the first model. By means of the method provided in the present application, the consumer of the first model can better understand and trust the model on the basis of the first information, thereby helping to improve the availability of the model.

Description

A communication method and apparatus

[0001] Cross-reference to related applications

[0002] This application claims priority to Chinese Patent Application No. 202411600398.2, filed on November 8, 2024, entitled "A Communication Method and Apparatus", the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of communication technology, and in particular to a communication method and apparatus. Background Technology

[0004] Artificial intelligence / machine learning (AI / ML) technologies and related applications are being increasingly adopted by a wider range of industries. Currently, AI / ML technologies introduce interpretability into AI / ML models, aiming to help users understand and trust these models. How to achieve interpretability in AI / ML models so that consumers (or users) of the models can better understand and trust them is a current research hotspot. Summary of the Invention

[0005] This application provides a communication method and apparatus to enable consumers of a model to better understand and trust the model.

[0006] Firstly, this application provides a communication method applicable to a first device. Optionally, the first device can be a producer of a first model. Optionally, the first model can be an AI / ML model, and the naming and implementation of the first model are not limited in the embodiments of this application.

[0007] In one example, the first device may be a domain management functional unit, or a device within a domain management functional unit (e.g., a module, communication module, circuitry or chip responsible for communication functions (e.g., a modem chip, or a system-on-a-chip (SoC) chip or system-in-package (SIP) chip containing a modem core), a chip system, or a processor), or a logical node, logical module, or software capable of implementing all or part of the domain management functional unit. In another example, the first device may also be a network element, or a device within a network element (e.g., a module, communication module, circuitry or chip responsible for communication functions (e.g., a modem chip, or a SoC chip or SIP chip containing a modem core), a chip system, or a processor), or a logical node, logical module, or software capable of implementing all or part of the network element. The network element may be, for example, an access network element or a core network element, and is not limited to any particular type. In another example, the first device may also be a cross-domain management function unit, or a device within the cross-domain management function unit (e.g., a module, communication module, circuit or chip responsible for communication functions (such as a modem chip, or a SoC chip or SIP chip containing a modem core), chip system or processor), or a logical node, logical module or software that can implement all or part of the cross-domain management function unit.

[0008] The method may include: a first device determining first information of a first model, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the model inference information includes inference interpretability information; and sending the first information to a second device, the second device being a consumer of the first model.

[0009] Alternatively, the method may include: a first device sending first information to a second device, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the model inference information includes inference interpretability information, and the second device is a consumer of the first model.

[0010] In this application, the first device sends first information to the second device. The first information includes at least one of data interpretability information, process interpretability information, model interpretability information, and reasoning interpretability information. The first information explains the first model to the second device from multiple perspectives, including training data (or data samples), training process, training model, and model reasoning. This enables the second device to better understand and trust the model based on the first information, which is beneficial to improving the usefulness of the model.

[0011] In one possible implementation, the data interpretability information may include the distribution pattern of the data samples (e.g., uniform or non-uniform distribution) and / or the characteristics of the data samples.

[0012] For example, the characteristics of the data sample may include at least one of the following: mobility, coverage, energy saving, load, fault, or service experience.

[0013] Data samples are the basic input for training models, and the quality of data samples is crucial to the accuracy of training models. In the above implementation, data interpretability information is defined, and the training data of the first model is interpreted through the distribution and / or features of the data samples. In this way, consumers of the first model can determine whether to perform retraining based on the data interpretability information, so that consumers can better understand and trust the model, which is conducive to improving the usability of the model.

[0014] In one possible implementation, the process interpretability information may include training duration and / or number of iterations.

[0015] The above implementation defines process interpretability information, which explains the training process of the first model by training duration and / or number of iterations. In this way, consumers of the first model can determine whether to perform testing, simulation, deployment or retraining based on the process interpretability information, so that consumers can better understand and trust the model, which is conducive to improving the usability of the model.

[0016] In one possible implementation, the model interpretability information may include at least one of the following: complexity, simulation environment, or the reason for the prediction result.

[0017] The above implementation defines model interpretability information and explains the first model through complexity, simulation environment, and reasons corresponding to prediction results. In this way, consumers of the first model can determine whether to perform testing, simulation, deployment, or retraining based on the model interpretability information, so that consumers can better understand and trust the model, which is conducive to improving the usability of the model.

[0018] In one possible implementation, the reasoning interpretability information may include at least one of the following: a first reasoning result, information on the performance affected by the first reasoning result, a first interference result related to the first reasoning result, or information on the performance affected by the first interference result.

[0019] Optionally, the performance information affected by the first inference result may include the performance affected by the first inference result (or the performance metric affected by the first inference result) and / or the value of the performance affected by the first inference result.

[0020] Optionally, the performance affected by the first inference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0021] Optionally, the information on the performance affected by the first interference result may include the performance affected by the first interference result (or the performance index affected by the first interference result) and / or the value of the performance affected by the first interference result.

[0022] Optionally, the performance affected by the first interference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0023] The above implementation defines inference interpretability information. The inference behavior of the first model is explained by the interpretability information of the inference result (including the inference result and / or the performance information affected by the inference result) and the interpretability information of the interference result (including the interference result and / or the performance information affected by the interference result). In this way, the consumer of the first model can determine whether to perform retraining based on the inference interpretability information, so that the consumer can better understand and trust the model, which is conducive to improving the usability of the model.

[0024] In one possible implementation, the first information can be carried in the information object class of the ML training report; or, the model training information can be carried in the information object class of the ML training report, and the model inference information can be carried in the information object class of the ML inference report.

[0025] In one possible implementation, the method may further include: the first device sending second information to the second device, the second information being used to indicate the reasoning type to which the first information applies.

[0026] For example, the inference type may include at least one of the following: management data analysis, access network intelligence, or network data analysis functions.

[0027] Through the above implementation, the interpretability information of the first model can be presented differently for different types of reasoning, which helps to improve the usability of the model.

[0028] In one possible implementation, the first device determining the first information of the first model may include: the first device receiving third information from the second device, the third information including at least one of the following: first indication information, first indicator, second indicator, third indicator, fourth indicator, or first inference type; wherein, the first indication information is used to indicate the execution of at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability, the first indicator including the distribution of data samples and / or the characteristics of data samples, the second indicator including training duration and / or number of iterations, the third indicator including at least one of the following: complexity, simulation environment, or the reason corresponding to the prediction result, the fourth indicator including the performance indicator affected by the inference result and / or the performance indicator affected by the interference result; and determining the first information based on the third information.

[0029] Through the above implementation method, the first device can actively acquire the first information, or it can acquire the first information in response to the third information of the second device.

[0030] In one possible implementation, the third information includes the first indication information, which is used to indicate the execution of at least one of the following: data interpretability, process interpretability, or model interpretability; the first device determines the first information based on the third information, which may include: the first device performs model training interpretability based on the third information to obtain the first information, wherein the first information includes information about the model training.

[0031] In one possible implementation, the third information includes the first inference type; the first inference type includes management data analysis, and the information trained by the model includes interpretability information applicable to the management data analysis; or, the first inference type includes access network intelligence, and the information trained by the model includes interpretability information applicable to the access network intelligence; or, the first inference type includes network data analysis function, and the information trained by the model includes interpretability information applicable to the network data analysis function.

[0032] In one possible implementation, the third information includes the first indication information, which is used to indicate the interpretability of the reasoning; the first device determines the first information based on the third information, which may include: the first device performs the interpretability of the model reasoning based on the third information to obtain the first information, wherein the first information includes information of the model reasoning.

[0033] In one possible implementation, the third information includes the first inference type; the first inference type includes management data analysis, and the information of the model inference may include at least one of the following: a second inference result corresponding to the management data analysis, information on the performance affected by the second inference result, a second interference result related to the second inference result, or information on the performance affected by the second interference result; or, the first inference type includes access network intelligence, and the information of the model inference may include at least one of the following: a third inference result corresponding to the access network intelligence, information on the performance affected by the third inference result, a third interference result related to the third inference result, or information on the performance affected by the third interference result; or, the first inference type includes network data analysis function, and the information of the model inference may include at least one of the following: a fourth inference result corresponding to the network data analysis function, information on the performance affected by the fourth inference result, a fourth interference result related to the fourth inference result, or information on the performance affected by the fourth interference result.

[0034] In one possible implementation, the third information can be carried in the information object class of the ML training request; or part of the third information can be carried in the information object class of the ML training request, and the remaining information can be carried in the information object class of the ML inference request.

[0035] Secondly, this application provides a communication method adapted to a second device. Optionally, the second device can be a consumer of a first model. Optionally, the first model can be an AI / ML model, and the naming and implementation of the first model are not limited in the embodiments of this application.

[0036] In one example, the second device may be a cross-domain management functional unit, or a device within a cross-domain management functional unit (e.g., a module, communication module, circuitry or chip responsible for communication functions (such as a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the cross-domain management functional unit. In another example, the second device may also be a third-party device, or a device within a third-party device (e.g., a module, communication module, circuitry or chip responsible for communication functions (such as a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the third-party device. The third-party device may be, for example, operator equipment or a business operation unit, and is not limited thereto.

[0037] The method may include: a second device receiving first information from a first device, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the model inference information includes inference interpretability information; and performing corresponding processing based on the first information.

[0038] Alternatively, the method may include: a second device receiving first information from a first device, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the model inference information includes inference interpretability information.

[0039] In one possible implementation, the first information includes information about the model training, which includes data interpretability information; the second device performs corresponding processing based on the first information, which may include: the second device determining whether the first model needs to be retrained based on the data interpretability information.

[0040] In one possible implementation, the first information includes information about the model training, which includes process interpretability information; the second device performs corresponding processing based on the first information, which may include: the second device determining, based on the process interpretability information, whether at least one of the following needs to be performed on the first model: testing, simulation, deployment, or retraining.

[0041] In one possible implementation, the first information includes information about the model training, which includes model interpretability information; the second device performs corresponding processing based on the first information, which may include: the second device determining whether to perform at least one of the following on the first model based on the model interpretability information: testing, simulation, deployment, or retraining.

[0042] In one possible implementation, the first information includes the information of the model inference; the second device performs corresponding processing based on the first information, which may include: the second device determining whether the first model needs to be retrained based on the inference interpretability information.

[0043] In one possible implementation, the data interpretability information may include the distribution pattern of the data samples (e.g., uniform or non-uniform distribution) and / or the characteristics of the data samples.

[0044] For example, the characteristics of the data sample may include at least one of the following: mobility, coverage, energy saving, load, fault, or service experience.

[0045] In one possible implementation, the process interpretability information may include training duration and / or number of iterations.

[0046] In one possible implementation, the model interpretability information may include at least one of the following: complexity, simulation environment, or the reason for the prediction result.

[0047] In one possible implementation, the reasoning interpretability information may include at least one of the following: a first reasoning result, information on the performance affected by the first reasoning result, a first interference result related to the first reasoning result, or information on the performance affected by the first interference result.

[0048] Optionally, the performance information affected by the first inference result may include the performance affected by the first inference result (or the performance metric affected by the first inference result) and / or the value of the performance affected by the first inference result.

[0049] Optionally, the performance affected by the first inference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0050] Optionally, the information on the performance affected by the first interference result may include the performance affected by the first interference result (or the performance index affected by the first interference result) and / or the value of the performance affected by the first interference result.

[0051] Optionally, the performance affected by the first interference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0052] In one possible implementation, the first information can be carried in the information object class of the ML training report; or, the model training information can be carried in the information object class of the ML training report, and the model inference information can be carried in the information object class of the ML inference report.

[0053] In one possible implementation, the method may further include: a second device receiving second information from the first device, the second information being used to indicate the reasoning type to which the first information applies.

[0054] For example, the inference type may include at least one of the following: management data analysis, access network intelligence, or network data analysis functions.

[0055] In one possible implementation, the method may further include: the second device sending third information to the first device, the third information including at least one of the following: first indication information, first indicator, second indicator, third indicator, fourth indicator, or first inference type; wherein, the first indication information is used to indicate the execution of at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability, the first indicator including the distribution mode of data samples and / or the characteristics of data samples, the second indicator including training duration and / or number of iterations, the third indicator including at least one of the following: complexity, simulation environment, or the reason corresponding to the prediction result, and the fourth indicator including the performance indicator affected by the inference result and / or the performance indicator affected by the interference result.

[0056] In one possible implementation, the first information includes information about the model training, and the third information includes the first inference type; the first inference type includes management data analysis, and the information about the model training includes interpretability information applicable to the management data analysis; or, the first inference type includes access network intelligence, and the information about the model training includes interpretability information applicable to the access network intelligence; or, the first inference type includes network data analysis functions, and the information about the model training includes interpretability information applicable to the network data analysis functions.

[0057] In one possible implementation, the first information includes the model inference information, and the third information includes the first inference type; the first inference type includes management data analysis, and the model inference information may include at least one of the following: a second inference result corresponding to the management data analysis, performance information affected by the second inference result, a second interference result related to the second inference result, or performance information affected by the second interference result; or, the first inference type includes access network intelligence, and the model inference information may include at least one of the following: a third inference result corresponding to access network intelligence, performance information affected by the third inference result, a third interference result related to the third inference result, or performance information affected by the third interference result; or, the first inference type includes network data analysis function, and the model inference information may include at least one of the following: a fourth inference result corresponding to the network data analysis function, performance information affected by the fourth inference result, a fourth interference result related to the fourth inference result, or performance information affected by the fourth interference result.

[0058] In one possible implementation, the third information can be carried in the information object class of the ML training request; or part of the third information can be carried in the information object class of the ML training request, and the remaining information can be carried in the information object class of the ML inference request.

[0059] Thirdly, this application provides a communication device that can be used to execute the methods described in the first aspect and any possible implementation thereof. The communication device can be a first device. The communication device may include modules, units, or means corresponding to the methods described in the first aspect and any possible implementation thereof. These modules, units, or means can be implemented in hardware, software, or by hardware executing corresponding software implementations. The hardware or software includes one or more modules or units corresponding to the above functions.

[0060] In one possible implementation, the communication device may include a baseband device and a radio frequency device.

[0061] In another possible implementation, the communication device may include a processing module (sometimes also called a processing unit) and a transceiver module (sometimes also called a transceiver unit). The transceiver module is capable of both sending and receiving functions. When the transceiver module performs the sending function, it may be called a sending module (sometimes also called a sending unit), and when it performs the receiving function, it may be called a receiving module (sometimes also called a receiving unit). The sending module and the receiving module may be the same functional module, referred to as the transceiver module, which performs both sending and receiving functions; or, the sending module and the receiving module may be different functional modules, with "transceiver module" being a collective term for these functional modules.

[0062] Fourthly, this application provides a communication device that can be used to execute the methods described in the second aspect and any possible implementation thereof. The communication device can be a second device. The communication device may include modules, units, or means corresponding to the methods described in the second aspect and any possible implementation thereof. These modules, units, or means may be implemented in hardware, software, or by hardware executing corresponding software implementations. The hardware or software includes one or more modules or units corresponding to the aforementioned functions.

[0063] In one possible implementation, the communication device may include a baseband device and a radio frequency device.

[0064] In another possible implementation, the communication device may include a processing module (sometimes also called a processing unit) and a transceiver module (sometimes also called a transceiver unit). The transceiver module is capable of both sending and receiving functions. When the transceiver module performs the sending function, it may be called a sending module (sometimes also called a sending unit), and when it performs the receiving function, it may be called a receiving module (sometimes also called a receiving unit). The sending module and the receiving module may be the same functional module, referred to as the transceiver module, which performs both sending and receiving functions; or, the sending module and the receiving module may be different functional modules, with "transceiver module" being a collective term for these functional modules.

[0065] Fifthly, this application provides a communication system that may include at least one of the following: the communication device provided in the third aspect above, or the communication device provided in the fourth aspect above.

[0066] Sixthly, this application also provides a communication device. The communication device may include one or more processors. Optionally, the communication device may further include a memory. The memory is used to store one or more computer programs or instructions. The one or more processors are used to execute the one or more computer programs or instructions stored in the memory, causing the communication device to perform the methods described in any of the first or second aspects and any possible implementations thereof.

[0067] In a seventh aspect, this application also provides a communication device, comprising: a processor and an interface circuit; the interface circuit is configured to receive signals from other communication devices outside the communication device and transmit them to the processor, or to send signals from the processor to other communication devices outside the communication device. The processor is configured to implement the methods described in any of the first or second aspects and any possible implementations thereof through logic circuits or by executing computer programs or instructions.

[0068] In some possible designs, when the device is a chip system, it can be composed of chips or contain chips and other discrete components.

[0069] Eighthly, this application also provides a chip system comprising at least one chip and a memory, wherein the at least one chip is configured to read and execute a program stored in the memory to implement the method described in any of the first or second aspects and any possible implementation thereof.

[0070] Ninthly, this application also provides a computer-readable storage medium for storing a computer program or instructions that, when executed, cause the method described in any of the first or second aspects and any possible implementation thereof to be implemented.

[0071] In a tenth aspect, this application also provides a computer program product comprising a computer program or instructions that, when executed on a computer, cause the method described in any of the first or second aspects and any possible implementation thereof to be implemented.

[0072] The technical effects achievable by the second to tenth aspects and any of their possible implementations are described in the same manner as the technical effects achievable by the first aspect and any of its possible implementations, and will not be repeated here. Attached Figure Description

[0073] Figure 1 is a schematic diagram of the architecture of a service system provided in this application;

[0074] Figure 2 is a schematic diagram of the architecture of another service system provided in this application;

[0075] Figure 3 is a flowchart illustrating a communication method provided in this application;

[0076] Figure 4 is a flowchart illustrating another communication method provided in this application;

[0077] Figure 5 is a schematic diagram of the structure of a communication device provided in this application;

[0078] Figure 6 is a schematic diagram of another communication device provided in this application;

[0079] Figure 7 is a schematic diagram of another communication device provided in this application. Detailed Implementation

[0080] The relevant terms used in the embodiments of this application will be explained below. It should be noted that these explanations are for the purpose of making the embodiments of this application easier to understand, and should not be regarded as a limitation on the scope of protection claimed by this application.

[0081] I. The term “explainability / explanation” in this application may be replaced with “explainable”, “explanation”, or “local explanation”, or may be used interchangeably.

[0082] The term "AI / ML" in this application may be replaced by "AI", "ML", or "AIML", or they may be used interchangeably. AI / ML may be understood as AI and ML, or AI or ML, or AI and / or ML.

[0083] The term "reasoning" in this application may be replaced with "model reasoning", "AI reasoning", "ML reasoning", "AI / ML reasoning", "AIML reasoning", "AI model reasoning", "ML model reasoning", "AI / ML model reasoning", or "AIML model reasoning", etc., or may be used interchangeably.

[0084] II. The term "multiple" in this application can refer to two or more. Therefore, "multiple" can also be understood as "at least two" in this application. "At least one" can be understood as one or more, such as one, two, or more. For example, "including at least one" means including one, two, or more. For instance, including at least one of A, B, and C could mean including A, B, C, A and B, A and C, B and C, or A, B, and C. "And / or" describes the relationship between related objects. Specifically, there can be three relationships. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / ", unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0085] 3. The terms “system” and “network” in this application may be used interchangeably, as may “according to” and “based on”.

[0086] IV. The ordinal numbers such as "first" and "second" mentioned in this application are generally used to distinguish different objects, and are not used to limit the order, sequence, priority, or importance of multiple objects. For example, the first device and the second device mentioned in this application are used to distinguish different devices, and do not limit the order, sequence, priority, or importance of these two devices.

[0087] V. The terms “comprising” and “having” and any variations thereof in this application are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product or device.

[0088] VI. The term "predefined" in this application may include predefined terms, such as protocol definitions. "Predefined" can be implemented by pre-storing corresponding codes, tables, or other means of indicating relevant information in the device (e.g., including various network elements), and this application does not limit the specific implementation method.

[0089] VII. The arrows or boxes indicated by dashed lines in the schematic diagrams in the accompanying drawings of this application represent optional steps or optional modules.

[0090] 8. In this application, "instruction" may include direct instruction, indirect instruction, explicit instruction, and implicit instruction. When describing a certain instruction information for the purpose of instructing A, it can be understood that the instruction information carries A, directly instructs A, or indirectly instructs A.

[0091] In this application, the information indicated by the instruction information is called the information to be instructed. In specific implementations, there are many ways to indicate the information to be instructed, such as, but not limited to, directly indicating the information to be instructed, such as the information to be instructed itself or its index. It can also indirectly indicate the information to be instructed by indicating other information, where there is a relationship between the other information and the information to be instructed. It can also indicate only a part of the information to be instructed, while the other parts are known or pre-agreed upon. For example, the instruction of specific information can be achieved by using a pre-agreed (e.g., protocol-defined) arrangement of various pieces of information, thereby reducing instruction overhead to some extent. Furthermore, the information to be instructed can be sent as a whole or divided into multiple sub-information pieces, and the sending period and / or timing of these sub-information pieces can be the same or different.

[0092] 9. In this application, "send" and "receive" indicate the direction of signal transmission. For example, "send information to XX" can be understood as the destination of the information being XX, which may include direct transmission via the air interface or indirect transmission by other units or modules via the air interface. "Receive information from YY" can be understood as the source of the information being YY, which may include direct reception from YY via the air interface or indirect reception from YY by other units or modules via the air interface. "Send" can also be understood as the "output" of the chip interface, and "receive" can also be understood as the "input" of the chip interface. In other words, sending and receiving can occur between devices, such as between access network devices and access and mobility management function network elements, or within a device, such as between components, modules, chips, software modules, or hardware modules within the device via buses, wiring, or interfaces.

[0093] 10. In this application, the terms "exemplarily," "for example," "e.g.," are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as an "example" in this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the term "example" is intended to present concepts in a concrete manner. In the embodiments of this application, "of," "corresponding, relevant," and "corresponding" may sometimes be used interchangeably, and it should be noted that their intended meanings are consistent unless the distinction is emphasized. In the embodiments of this application, "identifier" may sometimes be replaced with "identity," and it should be noted that their intended meanings are consistent unless the distinction is emphasized.

[0094] XI. The embodiments of this application will be presented in the context of a system including multiple devices, components, modules, etc. It should be understood that the system may include other unmentioned devices, components, modules, etc., or may only include some of the devices, components, or modules mentioned in the embodiments. Optionally, the terms "component" and "part" in this application can be used interchangeably.

[0095] The following describes the communication system to which the embodiments of this application are applicable.

[0096] The embodiments of this application can be applied to various mobile communication systems, such as New Radio (NR) systems, Long Term Evolution (LTE) systems, Advanced Long Term Evolution (LTE-A) systems, Future Communication Systems, and other communication systems, without limitation. Exemplarily, the various embodiments in this application can be used in the network management architecture of NR. The NR network management architecture may include management functions (MnFs). An MnF is a management entity defined by the 3rd Generation Partnership Project (3GPP), and its externally visible behavior and interfaces are defined as management services (MnS). In a service-providing management architecture, an MnF acts as an MnS producer or an MnS consumer. The management services produced by an MnF's MnS producer may have multiple MnS consumers. An MnF can consume multiple management services from one or more management service producers. As shown in Figure 1, the MnS provided by MnF can be used to provide services to other MnFs (such as MnF#A). In this case, MnF can act as an MnS producer, and MnF#A can act as an MnS consumer. Furthermore, MnF can also obtain MnS services provided by other MnFs (denoted as MnF#B, which may be the same as or different from MnF#A). That is, the MnF shown in Figure 1 can also act as an MnS consumer of the MnS services provided by MnF#B. In other words, the same MnF can act as both a consumer and a producer of MnS services.

[0097] Figure 2 is a schematic diagram of a service-oriented management architecture adapted to an embodiment of this application. This service-oriented management architecture includes a business support system (BSS), a cross-domain management function (CD-MnF) unit, a domain management function (Domain-MnF) unit, and a network element (NE). It should be understood that the cross-domain management function unit can be nodes such as a network management system (NMS), an MnS Producer, and an MnS Consumer, while the domain management function unit can be nodes such as a wireless automation engine (MBB automation engine, MAE), an element management system (EMS), an MnS Producer, and an MnS Consumer.

[0098] In this application, the cross-domain management function unit is used to manage one or more domain management function units. A domain management function unit can be used to manage one or more network elements. The following is a brief introduction to each unit.

[0099] If the management service is a management service provided by the cross-domain management functional unit, then the cross-domain management functional unit is the management service producer, and the business support system is the management service consumer.

[0100] If the management service is a management service provided by a domain management function unit, then the domain management function unit is the management service producer, and the cross-domain management function unit is the management service consumer.

[0101] When the management service is a management service provided by the network element, the network element is the management service producer, and the domain management functional unit is the management service consumer.

[0102] A business support system (BSS) is oriented towards communication services, providing functions and management services such as billing, settlement, accounting, customer service, sales, network monitoring, communication service lifecycle management, and service intent translation. The BSS can be an operator's operating system or a vertical industry operating system (vertical OT system).

[0103] The cross-domain management function unit can also be called a cross-domain management function node or a network management function (NMF) unit, such as an NMS or a network function management service consumer (NFMS_C). The cross-domain management function unit provides one or more of the following management functions or services: network lifecycle management, network deployment, network fault management, network performance management, network configuration management, network assurance, network optimization, and translation of network intents from communication service providers (Intent-CSP).

[0104] The network referred to in the aforementioned management functions or services may include one or more network elements or sub-networks, or it may be a network slice. In other words, the network management function unit may be a network slice management function (NSMF) unit, a management data analytical function (MDAF) unit, a self-organization network function (SON Function) unit, or an intent-driven management service (Intent Driven MnS) unit.

[0105] Optionally, in certain deployment scenarios, the cross-domain management function unit can also provide sub-network lifecycle management, sub-network deployment, sub-network fault management, sub-network performance management, sub-network configuration management, sub-network assurance, sub-network optimization functions, and translation of network intents from communication service consumers (Intent-CSC) for sub-network service producers or service consumers. Here, a sub-network consists of multiple smaller sub-networks, which can be network slice sub-networks.

[0106] Domain management function units can also be called domain management function nodes, NMFs, or network element management function units. For example, a domain management function unit can be a network element management entity such as a MAE, EMS, or a network function management service provider (NFMS_P).

[0107] The domain management function unit provides one or more of the following functions or management services: lifecycle management of subnetworks or network elements, deployment of subnetworks or network elements, fault management of subnetworks or network elements, performance management of subnetworks or network elements, assurance of subnetworks or network elements, optimization functions of subnetworks or network elements, and translation of intents from network operators (Intent-NOPs) of subnetworks or network elements. Here, a subnetwork includes one or more network elements. A subnetwork can also include other subnetworks, meaning one or more subnetworks can form a larger subnetwork.

[0108] It should be understood that in this application, "intent" can refer to the expectations of the intent producer (such as a network element) and the system in which the intent producer resides (such as a network or sub-network), and may include requirements, goals, or constraints. The translation of intent refers to the process of determining the strategy for the intent. For example, a strategy can be used to indicate conditions that do not satisfy the intent. For instance, when the intent is energy saving, strategy A could be: when energy consumption exceeds threshold 1, energy consumption is abnormal (i.e., not energy-efficient); strategy B could be: when energy consumption exceeds threshold 2, energy consumption is abnormal (i.e., not energy-efficient). It is understandable that even for the same intent, different strategies may determine different solutions that satisfy the intent.

[0109] Alternatively, a subnet can also be a network slice subnet. The domain management system can be a network slice subnet management function (NSSMF), a domain management data analytical function (domain MDAF), a domain self-organizing network function (SON Function), a domain intent management function, etc.

[0110] The domain management functional units can be classified as follows:

[0111] Based on network type, functions can be categorized into: Radio Access Network (RAN) Domain Management Function (RAN Domain MnF) units, Core Network Domain Management Function (CN Domain MnF) units, and Transport Network Domain Management Function (TN Domain MnF) units. It's important to note that a domain management function unit can also be a domain network management system, capable of managing one or more of the access network, core network, or transport network.

[0112] According to administrative regions, they can be divided into: regional management functional units of a certain region, such as regional management functional units of city A, regional management functional units of city B, etc.

[0113] A network element is an entity that provides network services, including core network elements, access network elements, or transport network elements. For example, core network elements may include, but are not limited to, access and mobility management functions (AMF), session management functions (SMF), policy control functions (PCF), network data analytical functions (NWDAF), network repository functions (NRF), or gateways. For example, access network elements may include, but are not limited to, base stations (e.g., next-generation node B (gNB), evolved Node B (eNB), central unit control plane (CUCP), central unit (CU), distribution unit (DU), central unit user plane (CUUP), etc. In this application, network function (NF) is also referred to as network element (NE).

[0114] Among them, network elements can provide one or more of the following management functions or services: network element lifecycle management, network element deployment, network element fault management, network element performance management, network element assurance, network element optimization functions, and network element intent translation, etc.

[0115] It should be understood that the network architecture and business scenarios described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0116] The relevant technical features of this application will be described below.

[0117] AI / ML technologies and related applications are being adopted by a wider range of industries, and AI / ML capabilities are also being used in various areas of 5GS, including intelligent optimization use cases for RAN base stations, such as mobility load balancing (MLB), mobility robustness optimization (MRO), and energy saving (ES); management data analytics services for network management, such as management data analytics (MDA); and network data analytics services for the core network, such as network data analytics function (NWDAF).

[0118] For example, the management workflow of AI / ML includes training, testing, simulation, deployment, and inference phases.

[0119] The training phase involves training one or more ML models, including initial training and retraining. It also includes the validation of ML entities to evaluate their performance on both training and validation data.

[0120] During the testing phase, if the validation results do not meet expectations, such as unacceptable variance, the ML model associated with that ML entity needs to be retrained. The training phase is the initial stage of the AI / ML management workflow.

[0121] During the simulation phase, ML entities used for inference are run in a simulation environment. The purpose is to evaluate the inference performance of the ML entities in the simulation environment before applying them to the target network or system.

[0122] The deployment phase is the process of enabling trained ML entities to be used in the target AI / ML inference function.

[0123] The reasoning phase is the process of using ML entities to perform reasoning through AI / ML reasoning functions.

[0124] Currently, AI / ML technology introduces interpretability into AI / ML models, aiming to understand and trust these models. Interpretability can include, but is not limited to, global interpretability and local interpretability. Global interpretability refers to the ability to explain and understand model decisions based on the conditional interactions between dependent (response) variables and independent (predictor) features across the complete dataset. Global interpretability requires the complete model structure. Local interpretability refers to focusing on a single data point and examining a local sub-region in the feature space surrounding that data point to understand its predictive decision, attempting to understand the model's decision based on that local sub-region. Local interpretability does not concern itself with the model's structure or assumptions.

[0125] In layman's terms, the interpretability of an AI / ML model can be understood as the ability of the AI / ML model to be explained and presented in terms that are understandable to humans. Its purpose is to explain the individual output provided by the AI / ML model, that is, to focus on explaining why the AI / ML model generates a specific output for a specific input data sample.

[0126] Improving the interpretability of AI / ML models so that consumers (or users) can better understand and trust them is a current research hotspot. Therefore, embodiments of this application provide a communication method and apparatus to enable consumers to better understand and trust the model, thereby improving its usability. The method and apparatus are based on the same technical concept. Since the principles underlying the problems solved by the method and apparatus are similar, their implementations can be mutually referenced, and repeated details will not be elaborated upon.

[0127] In the embodiments of this application, the first device can be a producer of the first model. For example, the first device can be an MnS producer, or a device within an MnS producer (such as a chip or chip system). The second device can be a consumer of the first model, or a user of the first model. For example, the second device can be an MnS consumer, or a device within an MnS consumer (such as a chip or chip system). The first device and the second device can be deployed in different entities, or they can be deployed in the same entity, without limitation.

[0128] Optionally, the first model can be an AI / ML model. For details on AI / ML, please refer to the terminology description; further elaboration is unnecessary. This application does not limit the naming or implementation of the first model.

[0129] In one embodiment, the first device may be a domain management functional unit as shown in FIG2, or a device within a domain management functional unit (e.g., a module, communication module, circuit or chip responsible for communication functions (e.g., a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the domain management functional unit. Alternatively, the first device may also be a network element as shown in FIG2, or a device within a network element (e.g., a module, communication module, circuit or chip responsible for communication functions (e.g., a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the network element. Alternatively, the first device may also be a cross-domain management functional unit as shown in FIG2, or a device within a cross-domain management functional unit (e.g., a module, communication module, circuit or chip responsible for communication functions (e.g., a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the cross-domain management functional unit.

[0130] In one embodiment, the second device may be the cross-domain management functional unit shown in Figure 2, or a device within the cross-domain management functional unit (e.g., a module, communication module, circuit or chip responsible for communication functions (e.g., a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the cross-domain management functional unit. Alternatively, the second device may also be a third-party device, or a device within a third-party device (e.g., a module, communication module, circuit or chip responsible for communication functions (e.g., a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the cross-domain management functional unit. The third-party device may be any device other than those in the cross-domain management functional unit, domain management functional unit, or network element, and is not limited in scope. For example, the third-party device may be operator equipment or a service operation unit, and is not limited in scope.

[0131] For example, in the architecture of a single-domain autonomous network, the first device can be a domain management functional unit or a network element, and the second device can be a cross-domain management functional unit. As another example, in the architecture of a cross-domain autonomous network, the first device can be a cross-domain management functional unit, and the second device can be carrier equipment or a service operation unit.

[0132] It should be understood that the embodiments of this application do not limit the implementation of the first device and the second device.

[0133] Figure 3 is a flowchart illustrating a communication method provided in an embodiment of this application. As shown in Figure 3, the method may include the following:

[0134] S301: The first device determines (or generates, or acquires) the first information of the first model.

[0135] S301 is an optional step, indicated by a dashed line in Figure 3.

[0136] The first model can be an AI / ML model. For a description of AI / ML, please refer to the aforementioned terminology explanation; it will not be repeated here. This application does not limit the naming or implementation of the first model.

[0137] The first information may include model training information, or model inference information, or both model training information and model inference information.

[0138] The information about model training and model inference will be introduced separately below.

[0139] 1. Information about model training:

[0140] The information used in model training refers to the information used in training the first model. In this application, the information used in model training can be used to explain the training of the first model. Optionally, the information used in model training can also be called: training explanation information, training explanation result, training explanation report, training-related explanation information, training-related explanation result, or training-related explanation report, etc. The naming of the information used in the embodiments of this application is not limited.

[0141] In one implementation, the information for model training may include at least one of the following: data interpretability information, process interpretability information, or model interpretability information. These will be described in detail below.

[0142] 1) Data explanation information can be understood as the interpretability information corresponding to the training data of the first model, which can be used to explain the training data of the first model. Here, training data can be replaced with data samples.

[0143] In one implementation, data interpretability information may include the distribution pattern of data samples and / or the characteristics of data samples. That is, data interpretability information includes the distribution pattern of data samples in the first model and / or the characteristics of data samples in the first model. Here, the data samples are the data used to train the model, also known as training data. The distribution pattern of the data samples can be, for example, uniformly distributed or non-uniformly distributed, and is not limited. The characteristics of the data samples refer to the features (or properties, or attributes, etc.) possessed by the data samples. Optionally, the characteristics of the data samples can be replaced by attributes or properties of the data samples, and is not limited.

[0144] For example, the characteristics of a data sample may include at least one of the following: mobility, coverage, energy efficiency, load, fault, or business experience.

[0145] For example, if the data sample is data on mobility-related measurements, then the data sample can be a data sample with mobility characteristics. Optionally, the mobility-related measurements may include, but are not limited to, at least one of the following: the number of successful handovers, or the number of failed handovers, etc.

[0146] For example, if the data sample is data on signal strength-related measurements, then the data sample can be a data sample with coverage characteristics. Optionally, the signal strength-related measurements may include, but are not limited to, at least one of the following: reference signal receiving power (RSRP), or reference signal receiving quality (RSRQ), etc.

[0147] For example, if the data sample is data on energy-saving related measurements, then the data sample can be a data sample with energy-saving characteristics. Optionally, energy-saving related measurements may include, but are not limited to, at least one of the following: energy consumption, or energy efficiency, etc.

[0148] For example, if the data sample is data of load-related measurements, then the data sample can be a data sample with load characteristics. Optionally, load-related measurements may include, but are not limited to, at least one of the following: physical radio bear (PRB) utilization, radio resource control (RRC) connection count, user throughput, or resource utilization rate, etc.

[0149] For example, if the data sample is data on fault-related measurements, then the data sample can be a data sample with fault characteristics. Optionally, fault-related measurements may include, but are not limited to, at least one of the following: alarm rate, or number of transport block (TB) errors, etc.

[0150] For example, if the data sample is data related to service experience measurements, then the data sample can be a data sample with service experience characteristics. Optionally, service experience related measurements may include, but are not limited to, at least one of the following: channel quality indicator (CQI) measurements, or quality of service (QoS) measurements, etc.

[0151] For specific measurements related to mobility, coverage, energy efficiency, load, fault, and service experience, please refer to the relevant descriptions in 3GPP TS28.552. This application does not limit these measurements.

[0152] Optionally, the distribution of data samples can be applied to all inference types; or it can be defined separately based on different inference types.

[0153] Optionally, the features of the data samples can be applied to all inference types and can be called general features; or they can be defined separately based on different inference types.

[0154] The reasoning type may include at least one of the following: management data analytical (MDA), access network intelligence (e.g., RAN intelligence), or network data analysis function (NWDAF).

[0155] For example, an MDA may include at least one of the following: coverage analysis or mobility analysis, as detailed in the types of MDAs defined in 3GPP TS28.104.

[0156] For example, RAN intelligence may include at least one of the following: random access channel optimization, mobile load balancing, mobile robustness optimization, or energy saving, etc. For details, please refer to the use cases defined in 3GPP TS28.313 or TS28.310.

[0157] For example, NWDAF may include at least one of the following: slice load analysis or user data congestion analysis, etc., as detailed in the use cases defined in 3GPP TS23.288.

[0158] For example, features of data samples such as mobility, coverage, and load can be applied to the inference type of MDA. For example, features of data samples such as energy saving and load can be applied to the inference type of RAN intelligence. For example, features of data samples such as faults and service experience can be applied to the inference type of NWDAF.

[0159] Taking energy-saving optimization as an example, the characteristics of the data sample may include at least one of the following: energy saving, or load, etc.

[0160] Taking fault management as an example, the characteristics of the data can include at least one of the following: fault, or business experience, etc.

[0161] Taking coverage optimization management as an example, the characteristics of the data may include at least one of the following: mobility, coverage, or load.

[0162] It should be understood that the embodiments of this application do not limit the characteristics of the data samples adapted to different inference types.

[0163] In this application, the training data of the first model may be provided by the second device, i.e., by the consumer of the first model; or it may be provided by the first device, i.e. by the producer of the first model; or it may be provided jointly by the first device and the second device; or it may be provided by other devices besides the first device and the second device. The embodiments of this application do not limit the source of the training data of the first model.

[0164] Optionally, the training data is provided by a second device, which may send interpretability information of the training data to the first device. Correspondingly, the first device receives the interpretability information of the training data and determines whether to use the training data based on this information. The interpretability information of the training data can be found in the description of data interpretability information, and will not be repeated here.

[0165] Optionally, data interpretability information may also be referred to as data interpretation information, data interpretation report, data interpretation result, training data interpretation information, training data interpretation report, training data interpretation result, data interpretability report, data interpretability result, training data interpretability information, training data interpretability report, or training data interpretability result, etc. The naming of data interpretability information is not limited in the embodiments of this application.

[0166] Training data (or data samples) is the basic input for training a model. The quality of the training data is crucial to the accuracy of the training model. In this embodiment, the training data of the first model is explained by the distribution of data samples and / or the characteristics of data samples. In this way, consumers of the first model can better understand and trust the model based on the interpretability information of the data, which is conducive to improving the usability of the model.

[0167] 2) Process explanation information can be understood as the explainability information corresponding to the training process of the first model, which can be used to explain the training process of the first model.

[0168] In one implementation, process interpretability information may include training duration, or the number of iterations, or both. That is, process interpretability information may include the number of iterations of the first model and / or the training duration of the first model.

[0169] Optionally, the training duration can be applied to all inference types; or it can be defined separately based on different inference types without restriction.

[0170] Optionally, the number of iterations can be applied to all inference types; or it can be defined separately based on different inference types without restriction.

[0171] Optionally, process interpretability information may also be referred to as: process interpretation information, process interpretation report, process interpretation result, training process interpretation information, training process interpretation report, training process interpretation result, process interpretability report, process interpretability result, training process interpretability information, training process interpretability report, or training process interpretability result, etc. The naming of process interpretability information is not limited in the embodiments of this application.

[0172] In this embodiment, the training process of the first model is explained by the training duration and / or number of iterations. In this way, consumers of the first model can determine whether to perform testing, simulation, deployment or retraining based on the interpretability information of the process, so that consumers can better understand and trust the model, which is beneficial to improving the usability of the model.

[0173] 3) Model interpretability information, which can be understood as the interpretability information corresponding to the training model of the first model, and can be used to explain the training model of the first model.

[0174] In one implementation, model interpretability information may include at least one of the following: complexity (or model complexity, or the complexity of training the model, etc.), simulation environment, or the reason for the prediction result. Optionally, complexity may also be referred to as model complexity, or the complexity of training the model, etc.

[0175] The simulation environment can be used to explain in which environments the training model has been simulated. The simulation environment can be, for example, a digital twin network (NDT), without limitation. Alternatively, the simulation environment can also be referred to as the model's simulation environment, the simulation environment for training the model, the simulation environment for model execution, or the simulation environment for training the model, etc., without limitation.

[0176] The cause (or factor) corresponding to the prediction result can be used to explain which factors caused a certain prediction result of the training model. For example, the cause corresponding to the prediction result can be the reason (or explanation) for the potential deterioration of the training model. For example, if the prediction result of the training model is an increase in energy consumption (or an increase in load) in cell A, the cause corresponding to the prediction result can include at least one of the following: a surge in traffic in cell A, or a change in coverage, thereby causing an increase in energy consumption in cell A. Optionally, the cause corresponding to the prediction result can also be called the cause corresponding to the model prediction result, or the cause corresponding to the training model prediction result, etc., without limitation.

[0177] Optionally, complexity can be applied to all types of reasoning; or it can be defined separately based on different types of reasoning, without restriction.

[0178] Optionally, the simulation environment can be applied to all inference types; or it can be defined separately based on different inference types without restriction.

[0179] Optionally, the reason corresponding to the prediction result can be applied to all inference types; or it can be defined separately based on different inference types without restriction.

[0180] Optionally, model interpretability information may also be referred to as: model interpretation information, model interpretation report, model interpretation result, training model interpretation information, training model interpretation report, training model interpretation result, model interpretability report, model interpretability result, training model interpretability information, training model interpretability report, or training model interpretability result, etc. The naming of model interpretability information is not limited in this application embodiment.

[0181] In this embodiment, the training model of the first model is explained by the complexity, simulation environment, and reasons corresponding to the prediction results. In this way, consumers of the first model can determine whether to perform testing, simulation, deployment, or retraining based on the interpretability information of the model, so that consumers can better understand and trust the model, which is conducive to improving the usability of the model.

[0182] 2. Information from model reasoning:

[0183] The information in model reasoning refers to the information in the first model's reasoning, or the information of the first model's reasoning result. In this application, the information in model reasoning can be used to explain the reasoning of the first model, or can be used to explain the reasoning result of the first model. For example, the information in model reasoning may include inference explanation information, or the information in model reasoning may be inference explanation information.

[0184] In one embodiment, the reasoning interpretability information may include at least one of the following: interpretability information of at least one reasoning result, or interpretability information of at least one interfering result related to at least one reasoning result. Optionally, the at least one reasoning result includes a first reasoning result, and correspondingly, the reasoning interpretability information may include at least one of the following: interpretability information of the first reasoning result, or interpretability information of a first interfering result related to the first reasoning result.

[0185] The following section uses the first inference result and the first interference result as examples for introduction.

[0186] 1) The first reasoning result is the reasoning result of the first model.

[0187] Interpretability information of the first inference result can be used to explain the performance impact that may result from the behavior that adopts the first inference result. For example, interpretability information of the first inference result may include the first inference result (e.g., it may be indicated by a string, without limitation) and / or information about the performance impact of the first inference result.

[0188] The performance impact information of the first inference result may include a performance metric (e.g., indicated by a string, without limitation) and / or the value of the performance impact of the first inference result. Optionally, the performance metric impacted by the first inference result may also be expressed as: the performance impacted by the first inference result. Optionally, the value of the performance impacted by the first inference result may also be expressed as: the value of the performance metric impacted by the first inference result.

[0189] It should be noted that the performance affected by the first inference result can be a specific value of the performance affected by the first inference result, or it can be a percentage increase / decrease in the performance affected by the first inference result, or it can be an increase / decrease in the performance affected by the first inference result. The implementation method of the performance affected by the inference result in this application embodiment is not limited.

[0190] For example, the performance metrics affected by the first inference result may include at least one of the following: cell energy consumption, cell energy efficiency, cell throughput, cell resource utilization (e.g., cell physical resource block (PRB) utilization), radio resource control (RRC) connection count, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution, etc. For specific performance metrics, please refer to the relevant descriptions in 3GPP TS28.552.

[0191] Optionally, the interpretability information of the first reasoning result can be applied to all reasoning types; or it can be defined separately based on different reasoning types without restriction.

[0192] Optionally, the interpretability information of the first reasoning result may also be referred to as: the interpretability report of the first reasoning result, the interpretability result of the first reasoning result, the explanation information of the first reasoning result, the explanation report of the first reasoning result, or the explanation result of the first reasoning result, etc. The embodiments of this application do not limit the interpretability information of the first reasoning result.

[0193] 2) The first interference result can be understood as the inference result adopting a behavior other than that of the first inference result. For example, the first inference result of the first model is to shut down cell A, cell B and cell C to take over the traffic of cell A; the first interference result related to the first inference result may include shutting down other cells (such as non-cell A) and / or non-cell B and cell C (i.e. cells other than cell B and cell C) to take over the traffic of cell A.

[0194] The interpretability information of the first interference result can be used to explain the performance impact that may result from actions that adopt the first interference result, or it can be used to explain the performance impact that may result from actions that adopt actions other than the first inference result. For example, the interpretability information of the first interference result may include the first interference result (e.g., indicated by a string, without limitation) and / or information about the performance impact of the first interference result.

[0195] The information regarding the performance impact of the first interference result may include a performance metric affected by the first interference result (e.g., indicated by a string, without limitation) and / or the value of the performance impacted by the first interference result. Optionally, the performance metric affected by the first interference result may also be expressed as: the performance impacted by the first interference result. Optionally, the value of the performance impacted by the first interference result may also be expressed as: the value of the performance metric affected by the first interference result.

[0196] It should be noted that the value of the performance affected by the first interference result can be a specific value of the performance affected by the first interference result, or it can be a percentage increase / decrease in the performance affected by the first interference result, or it can be an increase / decrease in the performance affected by the first interference result. The implementation method of the value of the performance affected by the interference result is not limited in the embodiments of this application.

[0197] For example, the performance metrics affected by the first interference result may include at least one of the following: cell energy consumption, cell energy efficiency, cell throughput, cell resource utilization (e.g., cell PRB utilization), number of RRC connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution, etc. For specific performance metrics, please refer to the relevant descriptions in 3GPP TS28.552.

[0198] It should be understood that the performance metrics affected by the first inference result and the performance metrics affected by the first interference result may be the same or different, without restriction.

[0199] Optionally, the performance of the first interference result can be applied to all inference types; or it can be defined separately based on different inference types without restriction.

[0200] Optionally, the interpretability information of the first interference result may also be referred to as: the interpretability report of the first interference result, the interpretability result of the first interference result, the explanation information of the first interference result, the explanation report of the first interference result, or the explanation result of the first interference result, etc. The embodiments of this application do not limit the interpretability information of the first interference result.

[0201] In another implementation, the inference interpretability information may include at least one of the following: a first inference result, information on the performance affected by the first inference result, a first interference result, or information on the performance affected by the first interference result, as described above, and will not be repeated here.

[0202] Taking energy-saving optimization as an example, the inference result of energy-saving optimization (i.e., the first inference result) is to shut down cell A, and have cell B and cell C take over the traffic of cell A. The performance information affected by the first inference result may include at least one of the following performance indicators: cell energy consumption, cell energy efficiency, cell throughput, cell PRB utilization, or RRC connection count, and the value of that at least one performance indicator. The value of that at least one performance indicator includes at least one of the following: percentage increase / decrease in regional energy consumption, percentage increase / decrease in energy efficiency of individual cells within the region, percentage increase / decrease in cell throughput change rate, percentage increase / decrease in cell PRB utilization, or that the energy consumption of individual cells within the region does not exceed a first threshold. Optionally, the first threshold may be pre-configured, and this application does not limit it.

[0203] Assume that the first interference result related to the first inference result may include shutting down other cells (such as non-cell A) and / or non-cell B and cell C (i.e., cells other than cell B and cell C) to take over the traffic of cell A. Information regarding the performance impact of the first interference result may include at least one performance indicator selected from cell energy consumption, cell energy efficiency, cell throughput, cell PRB utilization, or RRC connection count, and the value of that at least one performance indicator. The value of that at least one performance indicator may include at least one of the following: percentage increase / decrease in regional energy consumption, percentage increase / decrease in energy efficiency of individual cells within the region, percentage increase / decrease in cell throughput change rate, percentage increase / decrease in cell PRB utilization, or that the energy consumption of individual cells within the region does not exceed a first threshold, etc.

[0204] Taking coverage optimization management as an example, the inference result of coverage optimization includes adjusted antenna parameters such as downtilt angle, azimuth angle, and cell transmit power. Assuming the first inference result is to lower the downtilt angle and / or increase the transmit power, the performance information affected by the first inference result may include at least one performance indicator, such as the number of successful cell handovers, the number of failed cell handovers, or the cell signal strength distribution, and the value of that at least one performance indicator. The value of that at least one performance indicator may include at least one of the following: a percentage increase / decrease in the number of successful cell handovers, a percentage increase / decrease in the number of failed cell handovers, or a percentage of cell signal strength greater than a second threshold. Optionally, the second threshold can be -110 dB, without limitation. Optionally, the second threshold can be pre-configured, and this application does not limit it.

[0205] Assuming that the first interference result related to the first inference result may include increasing the downtilt angle and / or decreasing the transmit power, the performance information affected by the first interference result may include at least one performance indicator such as the number of successful cell handovers, the number of failed cell handovers, or the cell signal strength distribution, and the value of the at least one performance indicator. The value of the at least one performance indicator may include at least one of the following: percentage increase / decrease in the number of successful cell handovers, percentage increase / decrease in the number of failed cell handovers, or percentage of cell signal strength greater than a second threshold, etc.

[0206] Optionally, the reasoning interpretability information may also be referred to as: reasoning explanation information, reasoning explanation report, reasoning explanation result, model reasoning explanation information, model reasoning explanation report, model reasoning explanation result, reasoning result explanation information, reasoning result explanation report, reasoning result explanation result, reasoning interpretability report, reasoning interpretability result, reasoning result interpretability information, reasoning result interpretability report, or reasoning result interpretability result, etc. The naming of the reasoning interpretability information is not limited in the embodiments of this application.

[0207] In this embodiment, the reasoning behavior of the first model is explained by the interpretability information of the reasoning results and / or the interpretability information of the interference results. In this way, consumers of the first model can determine whether to perform retraining based on the reasoning interpretability information, so that consumers can better understand and trust the model, which is conducive to improving the usability of the model.

[0208] The aforementioned first information may include model training information and / or model inference information. In another embodiment, the first information may include at least one of the following: data interpretability information, process interpretability information, model interpretability information, or inference interpretability information. For details on data interpretability information, process interpretability information, model interpretability information, and inference interpretability information, please refer to the foregoing content; they will not be repeated here. This application does not limit the implementation form of the first information.

[0209] In S301, the first device determines (or generates, or acquires) first information. In one embodiment, the first device can determine the first information based on predefined information (or pre-configured information, or pre-agreed information). For example, the predefined information includes the interpretability of performing (or carrying out) model training; the first device performs the interpretability of model training and obtains model training information. For instance, the predefined information includes the interpretability of performing training data; the first device performs the interpretability of training data and obtains data interpretability information. Another example is that the predefined information includes the interpretability of performing the training process; the first device performs the interpretability of the training process and obtains process interpretability information. Yet another example is that the predefined information includes the interpretability of performing training model training; the first device performs the interpretability of training model training and obtains model interpretability information. Yet another example is that the predefined information includes the interpretability of performing (or carrying out) model inference; the first device performs the interpretability of model inference and obtains model inference information, such as obtaining inference interpretability information.

[0210] In another implementation, the first device may determine (or generate, or acquire) the third information based on the third information. For example, the second device may send the third information to the first device; correspondingly, the first device receives the third information from the second device and determines the first information based on the third information. Exemplarily, the second device may generate (or determine, or acquire) the third information and send it to the first device. For example, the second device may generate the third information based on model training requirements and / or model inference requirements, and send it to the first device. This application does not limit the method of generating the third information.

[0211] The third information may include at least one of the following: first indication information, first indicator (denoted as indicator #1), second indicator (denoted as indicator #2), third indicator (denoted as indicator #3), fourth indicator (denoted as indicator #4), or first inference type (denoted as inference type #1). Optionally, the indicator may also be replaced with: parameter, standard, or measurement standard, etc.

[0212] The following sections will introduce them separately.

[0213] 1) The first indication information may be used to indicate whether to perform (or allow) at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability. Alternatively, the first indication information may be used to enable or disable at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability. Alternatively, the first indication information may be used to indicate whether local interpretation of at least one of the following is required: data samples (or training data), training process, training model, or model inference.

[0214] Optionally, the first indication information can be implemented using a Boolean type (e.g., yes or no), without restriction.

[0215] For ease of understanding, embodiments of this application use first instruction information to instruct the execution of at least one of the following: data interpretability, process interpretability, model interpretability, or reasoning interpretability. Furthermore, the instruction information instructing the execution of data interpretability will be referred to as instruction information #1, the instruction information instructing the execution of process interpretability as instruction information #2, the instruction information instructing the execution of model interpretability as instruction information #3, and the instruction information instructing the execution of reasoning interpretability as instruction information #4. In other words, the first instruction information includes at least one of instruction information #1, instruction information #2, instruction information #3, and instruction information #4.

[0216] In one example, the third information includes at least one of the following: instruction information #1, instruction information #2, or instruction information #3. The first device performs interpretability testing of model training based on the third information to obtain the first information, such as model training information. For example, if the third information includes instruction information #1, the first device performs interpretability testing of training data based on instruction information #1 to obtain data interpretability information. As another example, if the third information includes instruction information #2, the first device performs interpretability testing of the training process based on instruction information #2 to obtain process interpretability information. As yet another example, if the third information includes instruction information #3, the first device performs interpretability testing of the trained model based on instruction information #3 to obtain model interpretability information.

[0217] In another example, the third information includes instruction information #4, and the first device performs interpretability testing of the model reasoning based on this third information to obtain the first information, such as model reasoning information. For example, the third information includes instruction information #4, and the first device performs interpretability testing of the model reasoning based on instruction information #4 to obtain reasoning interpretability information.

[0218] Optionally, the indication information #1 may also be called data interpretability indication, training data interpretability indication, data interpretation indication, or training data interpretation indication, etc. The naming of the indication information #1 is not limited in this application embodiment.

[0219] Optionally, the instruction information #2 may also be called process interpretability instruction, training process interpretability instruction, process explanation instruction, or training process explanation instruction, etc. The naming of the instruction information #2 is not limited in this application embodiment.

[0220] Optionally, the indication information #3 may also be called model interpretability indication, training model interpretability indication, model explanation indication, or training model explanation indication, etc. The naming of the indication information #3 is not limited in this application embodiment.

[0221] Optionally, the instruction information #4 may also be called the reasoning interpretability instruction, the reasoning result interpretability instruction, the reasoning explanation instruction, or the reasoning result explanation instruction, etc. The naming of the instruction information #4 is not limited in this application embodiment.

[0222] 2) Indicator #1 can be understood as a measure of the interpretability of the data sample.

[0223] For example, indicator #1 may include the distribution pattern and / or characteristics of the data samples, as detailed above. For instance, if the third information includes indicator #1, which represents the distribution pattern of the data samples, then the first information includes data interpretability information, which may include the distribution pattern of the data samples in the first model (e.g., uniform distribution, non-uniform distribution, etc.). As another example, if the third information includes indicator #1, which represents a mobility feature, and the data samples in the first model include mobility-related data samples, then the first information includes data interpretability information, which may include mobility. The implementation methods for the characteristics of other data samples are similar and will not be listed individually.

[0224] Optionally, indicator #1 may also be called a data interpretability indicator, a data sample interpretability indicator, a training data interpretability indicator, a data interpretation indicator, a data sample interpretation indicator, or a training data interpretation indicator, etc. The naming of indicator #1 is not limited in this application embodiment.

[0225] 3) Indicator #2 can be understood as a metric for the interpretability of the training process.

[0226] For example, indicator #2 may include training duration and / or number of iterations. For instance, if the third information includes indicator #2, where indicator #2 is training duration, then the first information includes process interpretability information, which may include the training duration of the first model. As another example, if the third information includes indicator #2, where indicator #2 is the number of iterations, then the first information includes process interpretability information, which may include the number of iterations of the first model.

[0227] Optionally, metric #2 may also be called process interpretability metric, training process interpretability metric, process explanation metric, or training process explanation metric, etc. The naming of metric #2 is not limited in this application embodiment.

[0228] 4) Indicator #3 can be understood as a metric for the interpretability of the trained model.

[0229] For example, indicator #3 may include at least one of the following: complexity, simulation environment, or the reason corresponding to the prediction result. For instance, if the third information includes indicator #3, and indicator #3 is complexity, then the first information includes model interpretability information, which may include the complexity of the first model. As another example, if the third information includes indicator #3, and indicator #3 is the simulation environment, then the first information includes model interpretability information, which may include the simulation environment in which the first model was executed. As yet another example, if the third information includes indicator #3, and indicator #3 is the reason corresponding to the prediction result, then the first information includes model interpretability information, which may include the reason corresponding to the prediction result of the first model.

[0230] Optionally, metric #3 can also be called model interpretability metric, training model interpretability metric, model explanation metric, or training model explanation metric, etc. The naming of metric #3 is not limited in this application embodiment.

[0231] 5) Indicator #4 can be understood as a metric for the interpretability of model reasoning.

[0232] For example, metric #4 may include performance metrics affected by the inference result and / or performance metrics affected by interference results, as described above. For instance, if the third information includes metric #4, and metric #4 includes performance metrics affected by the inference result (e.g., denoted as performance #1), then the first information includes inference interpretability information, which may include information about the first inference result of the first model and the performance #1 affected by the first inference result. As another example, if the third information includes metric #4, and metric #4 includes performance metrics affected by interference results (e.g., denoted as performance #2), then the first information includes inference interpretability information, which may include information about the first interference result and the performance #2 affected by the first interference result.

[0233] Optionally, the performance metrics indicated by metric #4 can be applied to interpretability information of inference results and interpretability information of interference results to reduce information transmission overhead.

[0234] Optionally, indicator #4 may also be called the reasoning interpretability indicator, the reasoning result interpretability indicator, the reasoning explanation indicator, or the reasoning result explanation indicator, etc. The naming of indicator #4 is not limited in this application embodiment.

[0235] 6) Inference type #1 can be used to indicate the interpretability information of the model training corresponding to (or related to) inference type #1, and / or to report the interpretability information of the model inference corresponding to (or related to) inference type #1.

[0236] For example, inference type #1 may include at least one of the following: MDA, RAN intelligence, or NWDAF. MDA, RAN intelligence, and NWDAF are described above and will not be repeated here.

[0237] In one example, the third information includes inference type #1, and the first information includes model training information applicable to inference type #1. For example, inference type #1 includes MDA, and the model training information includes interpretability information applicable to MDA. Another example is inference type #1 including RAN intelligence, and the model training information includes interpretability information applicable to RAN intelligence. Yet another example is inference type #1 including NWDAF, and the model training information includes interpretability information applicable to NWDAF.

[0238] In another example, the third information includes inference type #1, and the first information includes model inference information applicable to inference type #1. For example, inference type #1 includes MDA, and the model inference information (or inference interpretability information) may include at least one of the following: a second inference result corresponding to MDA, information on the performance impact of the second inference result, a second interference result related to the second inference result, or information on the performance impact of the second interference result. As another example, inference type #1 includes RAN intelligence, and the model inference information (or inference interpretability information) may include at least one of the following: a third inference result corresponding to RAN intelligence, information on the performance impact of the third inference result, a third interference result related to the third inference result, or information on the performance impact of the third interference result. As yet another example, inference type #1 includes NWDAF, and the model inference information (or inference interpretability information) may include at least one of the following: a fourth inference result corresponding to NWDAF, information on the performance impact of the fourth inference result, a fourth interference result related to the fourth inference result, or information on the performance impact of the fourth interference result.

[0239] The descriptions of the second, third, and fourth inference results can be referenced from those of the first inference result. Information regarding the performance impact of the second, third, and fourth inference results can also be referenced from those of the first inference result. Similarly, the descriptions of the second, third, and fourth interference results can be referenced from those of the first interference result. Further details are omitted here.

[0240] The aforementioned third information includes at least one of the following: first instruction information, indicator #1, indicator #2, indicator #3, indicator #4, and reasoning type #1. In another embodiment, the third information may include at least one of the following: first instruction information, indicator information, or reasoning type #1; or, the third information may include at least one of the following: instruction information #4, instruction information #5, indicator #1, indicator #2, indicator #3, indicator #4, or reasoning type #1; or, the third information may include at least one of the following: instruction information #4, instruction information #5, indicator information, or reasoning type #1; or, the third information may include at least one of the following: instruction information #4, instruction information #5, indicator #4, indicator #5, or reasoning type #1.

[0241] The indicator information may include at least one of the following: indicator #1, indicator #2, indicator #3, indicator #4.

[0242] Specifically, instruction information #5 can be used to instruct the execution of at least one of the following: data interpretability, process interpretability, or model interpretability. Optionally, instruction information #5 can also be called training interpretability instruction, training interpretability instruction, training interpretation instruction, or training interpretation instruction, etc. The naming of instruction information #5 is not limited in this embodiment. In addition, instruction information #5 and instruction information #4 can be carried in the same information, or they can be carried by different information.

[0243] Specifically, indicator #5 may include at least one of the following: indicator #1, indicator #2, and indicator #3. Optionally, indicator #5 may also be called a training interpretability indicator, a training interpretability indicator, a training interpretability indicator, or a training interpretability indicator, etc. The naming of indicator #5 is not limited in this embodiment. In addition, indicator #4 and indication information #5 may be carried in the same information, or they may be carried by different information.

[0244] Optionally, the third information can be carried in the information object class (IOC) of the ML training request; or, some of the third information can be carried in the IOC of the ML training request, and the remaining information can be carried in the IOC of the ML inference request. This application does not limit the method of carrying the third information.

[0245] In one implementation, the third information includes at least one of the following: indication information #5, indicator #5, or inference type #2. The third information can be carried in the IOC of the ML training request; or, in other words, the third information is included in the IOC of the ML training request. For example, a second device sends an ML training request to a first device, which requests the first device to create (or build) a first model. The IOC of the ML training request includes the third information. Accordingly, the first device receives the ML training request from the second device, creates the first model, and determines the first information based on the third information. Inference type #2 can be used to indicate the interpretability information of the model training corresponding to inference type #2. This inference type #2 belongs to inference type #1.

[0246] In one example, the IOC of the ML training request includes a training explanation indicator, i.e., indicator information #5, as shown in Table 1. The training explanation indicator can be a data type and can include at least one of the following: data explainability indicator (i.e., indicator information #1), process explainability indicator (i.e., indicator information #2), or model explainability indicator (i.e., indicator information #3), as shown in Table 2. Optionally, the data explainability indicator can be Boolean (e.g., yes or no), without restriction. Optionally, the process explainability indicator can be Boolean, without restriction. Optionally, the model explainability indicator can be Boolean, without restriction. Optionally, the IOC of the ML training request may also include training explainability metrics, i.e., metric #1, metric #2, and metric #3, as shown in Table 3. Optionally, the IOC of the ML training request may also include inference type #2.

[0247] Table 1

[0248] Table 2

[0249] Table 3

[0250] The "Support Qualifier" can take the following values: mandatory (M), optional (O), condition optional (CO), and conditional mandatory (CM). M indicates the attribute is mandatory, O indicates it is optional, CM indicates it is conditionally mandatory (meaning it is mandatory only if a certain condition is met), and CO indicates it is optional only if a certain condition is met. "Readability" can take the values ​​true (T) and false (F). T indicates the attribute is readable, and F indicates it is not readable. "Writability" can take the values ​​TRUE and FALSE. T indicates the attribute is writable, and F indicates it is not writable. This explanation also applies to the tables below and will not be repeated hereafter.

[0251] The attributes in Table 1, other than the training interpretability indicator, can be found in 3GPP TS28.105 and will not be repeated here. It should be understood that the values ​​of support qualifiers, readability, and writability for each attribute in Tables 1, 2, and 3 are merely examples, and this application embodiment does not limit them.

[0252] In one implementation, the third information includes at least one of the following: indication information #4, indicator #4, or inference type #3. The third information may be carried in the IOC of the ML inference request; or, in other words, the third information is included in the IOC of the ML inference request. For example, a second device sends an ML inference request to a first device, which requests the first device to perform (or create) inference for a first model. The IOC of the ML inference request includes the third information. Accordingly, the first device receives the ML inference request from the second device, performs model inference based on the first model, and determines the first information according to the third information. Inference type #3 can be used to indicate the interpretability information of the model inference corresponding to inference type #3. This inference type #3 belongs to inference type #1.

[0253] In one example, the IOC of an ML inference request includes an inference explanation indicator, i.e., indicator information #4, as shown in Table 4. Optionally, the inference explanation indicator can be Boolean and is not limited. Optionally, the IOC of an ML inference request may also include an inference explanation metric, i.e., metric #4, as shown in Table 5. Optionally, the IOC of an ML training request may also include inference type #3.

[0254] Table 4

[0255] Table 5

[0256] It should be understood that the values ​​of the support qualifiers, readability, and writability of each attribute in Tables 4 and 5 are merely examples, and the embodiments of this application do not limit them.

[0257] S302: The first device sends first information to the second device; correspondingly, the second device receives the first information from the first device.

[0258] In one implementation, the first information may be carried in the IOC of the ML training report; or, the first information may be included in the IOC of the ML training report.

[0259] In another implementation, model training information is carried in the IOC of the ML training report (or in other words, the model training information is included in the IOC of the ML training report), and / or, model inference information is carried in the IOC of the ML inference report (or in other words, the model inference information is included in the IOC of the ML inference report). For example, the first device sends an ML training report to the second device, which includes model training information. For example, the first device sends an ML inference report to the second device, which includes model inference information.

[0260] In one example, the IOC of the ML training report includes a training explanation indicator, which includes information about the model training, as shown in Table 6. Other attributes in Table 6 besides the training explanation indicator can be found in 3GPP TS28.105, and will not be elaborated further here.

[0261] Table 6

[0262] Optionally, the training interpretability results can be a data type, which may include at least one of the following: data interpretability report (i.e., data interpretability information), process interpretability report (i.e., process interpretability information), or model interpretability report (i.e., model interpretability information), as shown in Table 7.

[0263] It should be noted that the support qualifier for the data interpretability report in Table 7 is CM, indicating that the condition is mandatory. For example, if the data explanation indicator in Table 2 is "yes", the first device needs to generate (or determine, or acquire) a data interpretability report. Similarly, the support qualifier for the process interpretability report in Table 7 is CM, indicating that the condition is mandatory. For example, if the process explanation indicator in Table 2 is "yes", the first device needs to generate a process interpretability report. The support qualifier for the model interpretability report in Table 7 is CM, indicating that the condition is mandatory. For example, if the model explanation indicator in Table 2 is "yes", the first device needs to generate a model interpretability report.

[0264] Table 7

[0265] Optionally, the data interpretability report can be a data type that may include data interpretability metrics (such as metric #1) and the corresponding values ​​of the data interpretability metrics (which may be referred to as data interpretability values), as shown in Table 8.

[0266] Table 8

[0267] Optionally, the process interpretability report can be a data type, which may include process interpretability metrics (such as metric #2) and the corresponding values ​​of the process interpretability metrics (which may be called process interpretability values), as shown in Table 9. For example, if the process interpretability metric is training time, the process interpretability value is the training time of the first model. Another example is if the process interpretability metric is the number of iterations, and the process interpretability value is the number of iterations of the first model.

[0268] Table 9

[0269] Optionally, the model interpretability report can be a data type that may include model interpretability indicators (such as indicator #3) and factors predicted by the model (i.e., the reasons corresponding to the prediction results), as shown in Table 10. For example, the model interpretability report includes the reasons corresponding to the prediction results. The prediction results of the first model may be an increase or decrease in cell energy consumption or an increase or decrease in cell energy efficiency. The reasons corresponding to the prediction results may include changes in traffic and / or changes in coverage, etc., without limitation.

[0270] Table 10

[0271] It should be understood that the values ​​of the support qualifiers, readability, and writability of each attribute in Tables 6, 7, 8, 9, and 10 are merely examples, and the embodiments of this application do not limit them.

[0272] In another example, the IOC of the ML inference report includes an inference explanation indicator, i.e., it includes inference interpretability information, as shown in Table 11. Optionally, the inference interpretability indicator can be a data type, which may include at least one of the following: inference result, performance metric affected by the inference result, performance value affected by the inference result (i.e., the value of the performance affected by the inference result), interference result, performance metric affected by the interference result, or performance value affected by the interference result (i.e., the value of the performance affected by the interference result), as shown in Table 12.

[0273] Table 11

[0274] Table 12

[0275] It should be understood that the values ​​of the support qualifiers, readability, and writability of each attribute in Tables 11 and 12 are merely examples, and the embodiments of this application do not limit them.

[0276] In one embodiment, the first device may further send second information to the second device, and correspondingly, the second device receives the second information from the first device. This second information can be used to indicate the reasoning type to which the first information applies; or, in other words, the second information can be used to indicate the reasoning type to which the interpretability included in the first information applies. The implementation of this second information can refer to the implementation of reasoning type #1, and will not be repeated here. Optionally, the second information and the first information can be carried in the same message, or they can be carried in different messages, without limitation.

[0277] S303: The second device performs corresponding processing based on the first information.

[0278] S303 is an optional step, indicated by a dashed line in Figure 3.

[0279] For example, the second device processes the first model according to the first information. For instance, the first information includes data interpretability information, and the second device can determine (or ascertain) whether retraining of the first model is necessary based on the data interpretability information. Alternatively, the first information includes process interpretability information, and the second device can determine (or ascertain) whether one of the following actions needs to be performed on the first model: testing, simulation, deployment, or retraining. Another example is that the first information includes model interpretability information, and the second device can determine (or ascertain) whether at least one of the following actions needs to be performed on the first model: testing, simulation, deployment, or retraining. Yet another example is that the first information includes inference interpretability information, and the second device can determine (or ascertain) whether retraining of the first model is necessary based on the inference interpretability information. The embodiments of this application do not limit the implementation method of the second device processing the first information accordingly.

[0280] In this application, the first device sends first information to the second device. The first information includes at least one of data interpretability information, process interpretability information, model interpretability information, and reasoning interpretability information. The first information explains the first model to the second device from multiple perspectives, including training data (or data samples), training process, training model, and model reasoning. This enables the second device to better understand and trust the model based on the first information, which is beneficial to improving the usefulness of the model.

[0281] Figure 4 is a flowchart illustrating another communication method provided in an embodiment of this application. This embodiment is adapted to the architecture of a single-domain autonomous network, and is illustrated using Domain-MnF as the first device and NMS as the second device. Domain-MnF and NMS are described in the foregoing and will not be repeated here. As shown in Figure 4, the method may include the following:

[0282] S401: NMS sends an ML training request to Domain-MnF. Correspondingly, Domain-MnF receives the ML training request from NMS.

[0283] Among them, the ML training request can be used to request the creation of the first model.

[0284] Optionally, the ML training request may include information #1. For example, the IOC of the ML training request includes information #1. This information #1 may include at least one of the following: indication information #5, indicator #5, or inference type #2. Please refer to the description in S301 for indication information #5, indicator #5, and inference type #2, which will not be repeated here.

[0285] For example, NMS can determine information #1 and send information #1 to Domain-MnF via the ML training request. For instance, NMS can determine information #1 based on model training requirements and write it into the ML training request. For example, when requesting the creation of an object for the first model training, NMS can add information #1 to the creation operation, such as adding information #1 as a writable attribute to the IOC (Initial Code) of the ML model training request.

[0286] S402: Domain-MnF executes model training based on ML training requests to obtain information #2.

[0287] Information #2 refers to information about model training. Information #2 may include at least one of the following: data interpretability information, process interpretability information, or model interpretability information. For details on data interpretability information, process interpretability information, and model interpretability information, please refer to the relevant descriptions in S301; they will not be repeated here.

[0288] For example, Domain-MnF receives an ML training request from NMF, performs model training in response to the ML training request, and can obtain a first model and information #2. For example, the ML training request includes information #1, Domain-MnF configuration information #1, performs model training, and obtains the first model and information #2. For example, after receiving the ML training request, Domain-MnF creates an object for the first model training, such as creating an IOC for the ML model training request, and configures information #1 into that object.

[0289] For example, information #1 includes indication information #1 and / or indicator #1, and information #2 includes data interpretability information. As another example, information #1 includes indication information #2 and / or indicator #2, and information #2 includes process interpretability information. Yet another example, information #1 includes indication information #3 and / or indicator #3, and information #2 includes model interpretability information.

[0290] S403: Domain-MnF sends message #2 to NMS. NMS receives message #2 from Domain-MnF.

[0291] Optionally, information #2 can be carried in the ML training report (or response), without restriction.

[0292] S404: NMS will process information #2 accordingly.

[0293] Step S404 is optional and is represented by a dashed line in Figure 4. The implementation process of S404 can be referred to the content of S303, and will not be repeated here.

[0294] S405: The NMS sends an ML inference request to the Domain-MnF. Correspondingly, the Domain-MnF receives the ML inference request from the NMS.

[0295] Among them, the ML inference request can be used to request the execution of inference for the first model.

[0296] Optionally, the ML inference request may include information #3. For example, the IOC of the ML inference request includes information #3. This information #3 may include at least one of the following: indication information #4, indicator #4, or inference type #3. Please refer to the description in S301 for indication information #4, indicator #4, and inference type #3; they will not be repeated here.

[0297] For example, NMS can determine information #3 and send information #3 to Domain-MnF via an ML inference request. For instance, NMS can determine information #3 based on model inference requirements and write information #3 into the ML inference request. For example, NMS can add information #3 when requesting the object to perform the first model inference, such as adding information #3 as a writable attribute to the IOC (Inference Request Object) of the ML model inference request.

[0298] S406: Domain-MnF obtains information #4 by performing model inference based on ML inference request.

[0299] Information #4 refers to information related to model inference. Information #4 may include information on the interpretability of the inference; please refer to the relevant description in S301 for details, which will not be repeated here.

[0300] For example, Domain-MnF receives an ML inference request from NMF, performs model inference in response to the ML inference request, and can obtain information #4. For example, the ML inference request includes information #3, Domain-MnF configuration information #3, performs model inference, and obtains information #4. For example, after receiving the ML inference request, Domain-MnF creates a first model inference object, such as creating an IOC for the ML model inference request, and configures information #3 into that object.

[0301] S407: Domain-MnF sends message #4 to NMS. NMS receives message #4 from Domain-MnF.

[0302] Optionally, information #4 can be carried in the ML inference report (or response), without restriction.

[0303] S408: NMS performs the corresponding processing based on information #4.

[0304] Step S408 is optional and is represented by a dashed line in Figure 4. The implementation process of S408 can be referred to the content of S303, and will not be repeated here.

[0305] It should be understood that ML training requests and ML inference requests may come from the same device or from different devices, and this application embodiment does not limit this.

[0306] In another implementation, for the architecture of a single-domain autonomous network, the first device can also be an NE, and the second device can be an NMS. The NE can implement the content implemented by Domain-MnF in Figure 4. The NE and NMS interact through Domain-MnF. For example, an ML training request sent by the NMS can be transmitted to Domain-MnF and forwarded to the NE; information #2 sent by the NE can be transmitted to Domain-MnF and forwarded to the NMS. Similarly, an ML inference request sent by the NMS can be transmitted to Domain-MnF and forwarded to the NE; information #4 sent by the NE can be transmitted to Domain-MnF and forwarded to the NMS.

[0307] In another implementation, for the architecture of the cross-domain autonomous network, the first device can also be an NMS, and the second device can be a service operation unit. The NMS can implement the content implemented by Domain-MnF in Figure 4, and the service operation unit can implement the content implemented by the NMS in Figure 4. For the specific implementation process, please refer to the description in Figure 4, which will not be repeated here.

[0308] Based on the same technical concept as the above-described method embodiments, this application provides a corresponding communication device that can be used to perform the functions of the relevant steps in the above-described method embodiments. This function can be implemented in hardware, software, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. The communication device can be a terminal or access network device, or a device within the terminal or access network device (e.g., a module, communication module, circuit or chip responsible for communication functions (such as a modem chip, or a SoC chip or SIP chip containing a modem core), chip system, or processor), or a logical node, logical module, or software capable of implementing all or part of the terminal or functions.

[0309] Figure 5 illustrates a schematic diagram of a communication device 500 provided in an embodiment of this application. This communication device 500 can implement the functions or steps performed by the first device or the second device in the various method embodiments described above.

[0310] For example, when the communication device 500 is used to implement the functions or steps implemented by the first device in the above-described method embodiments, the communication device 500 may be a domain management function unit or a component in a domain management function unit, or a network element or a component in a network element, or a cross-domain management function unit or a component in a cross-domain management function unit, etc.

[0311] For example, when the communication device 500 is used to implement the functions or steps implemented by the second device in the above method embodiments, the communication device 500 may be a cross-domain management function unit or a component in the cross-domain management function unit, or a third-party device or a component in a third-party device, etc.

[0312] In one embodiment, the communication device 500 may include a processing module 501 and a transceiver module 502; or it may include a processing module 501 but not a transceiver module 502; or it may include a transceiver module 502 but not a processing module 501. Wherein:

[0313] The processing module 501 can be used to support the communication device 500 in performing the processing actions in the above method embodiments. The processing module 501 can be implemented using one or more processors. For example, the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontroller units (MCUs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0314] In this application, the processing module 501 may also be referred to as a processing unit, etc., without limitation.

[0315] Transceiver module 502 is used for inputting and / or outputting information. Input information can be replaced by received information, and output information can be replaced by transmitted information. When outputting information, transceiver module 502 can output information to other devices outside of communication device 500, or to other units within communication device 500. In some embodiments, transceiver module 502 can be implemented through at least one of a physical interface, a communication module, a communication interface, and an input / output interface. In other embodiments, transceiver module 502 can be implemented through interface circuitry, such as a mobile communication module. The mobile communication module may include one or more of at least one antenna, at least one filter, a switch, a power amplifier, and a low-noise amplifier (LNA).

[0316] Optionally, the transceiver module 502 may include a sending module and / or a receiving module. The sending module is used to perform the sending operation in the above method embodiments. The receiving module is used to perform the receiving operation in the above method embodiments. It should be noted that the communication device 500 may include a sending module but not a receiving module. Alternatively, the communication device 500 may include a receiving module but not a sending module. Specifically, it depends on whether the above scheme performed by the communication device 500 includes both sending and receiving actions.

[0317] In this application, the transceiver module 502 may also be referred to as a communication interface, a communication module, a transceiver unit, an interface module, an interface unit, or a communication unit, etc., without limitation.

[0318] It should be noted that the communication device 500 may include a processing module 501, but not a transceiver module 502. Alternatively, the communication device 500 may include a transceiver module 502, but not a processing module 501. Specifically, it depends on whether the above-described scheme executed by the communication device 500 includes processing and transceiver actions.

[0319] Optionally, the communication device 500 may further include a storage module, not shown in FIG5. The storage module may be used to store instructions and / or data, and the processing module 501 may read the instructions and / or data in the storage module to enable the communication device 500 to implement the aforementioned method embodiment.

[0320] Optionally, the communication device 500 may be a chip system, the transceiver module 502 may be the input / output interface of the chip (e.g., a baseband chip), and the processing module 501 may be the processor of the chip system.

[0321] In one possible design, when the communication device 500 is a communication equipment or a communication module within a communication equipment, the functionality of the processing module 501 can be implemented by one or more processors. Exemplarily, the processor may include a modem chip (also known as a baseband chip), or a system-on-a-chip (SoC) chip or system-in-package (SIP) chip containing a modem core. The functionality of the transceiver module 502 can be implemented by transceiver circuitry. Optionally, the communication equipment may be a domain management function unit, a network element, or a cross-domain management function unit.

[0322] In one possible design, when the communication device 500 is a circuit or chip responsible for communication functions within a communication device, such as a modem chip or a SoC chip or SIP chip containing a modem core, the function of the processing module 501 can be implemented by a circuit system in the aforementioned chip that includes one or more processors or processor cores. The function of the transceiver module 502 can be implemented by interface circuits or data transceiver circuits on the aforementioned chip. Optionally, the communication device can be a cross-domain management function unit or a third-party device.

[0323] In the first implementation, the communication device 500 can perform the functions of the first device and execute the following: a processing module 501 is used to determine the first information of the first model, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the model inference information includes inference interpretability information; a transceiver module 502 is used to send the first information to a second device, the second device being a consumer of the first model.

[0324] In one possible implementation, the data interpretability information may include the distribution pattern of the data samples (e.g., uniform or non-uniform distribution) and / or the characteristics of the data samples.

[0325] For example, the characteristics of the data sample may include at least one of the following: mobility, coverage, energy saving, load, fault, or service experience.

[0326] In one possible implementation, the process interpretability information may include training duration and / or number of iterations.

[0327] In one possible implementation, the model interpretability information may include at least one of the following: complexity, simulation environment, or the reason for the prediction result.

[0328] In one possible implementation, the reasoning interpretability information may include at least one of the following: a first reasoning result, information on the performance affected by the first reasoning result, a first interference result related to the first reasoning result, or information on the performance affected by the first interference result.

[0329] Optionally, the performance information affected by the first inference result may include the performance affected by the first inference result (or the performance metric affected by the first inference result) and / or the value of the performance affected by the first inference result.

[0330] Optionally, the performance affected by the first inference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0331] Optionally, the information on the performance affected by the first interference result may include the performance affected by the first interference result (or the performance index affected by the first interference result) and / or the value of the performance affected by the first interference result.

[0332] Optionally, the performance affected by the first interference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0333] In one possible implementation, the first information can be carried in the information object class of the ML training report; or, the model training information can be carried in the information object class of the ML training report, and the model inference information can be carried in the information object class of the ML inference report.

[0334] In one possible implementation, the transceiver module 502 is further configured to send second information to the second device, the second information being used to indicate the reasoning type to which the first information applies.

[0335] For example, the inference type may include at least one of the following: management data analysis, access network intelligence, or network data analysis functions.

[0336] In one possible implementation, when determining the first information of the first model, the transceiver module 502 is used to receive third information from the second device. The third information includes at least one of the following: first indication information, first indicator, second indicator, third indicator, fourth indicator, or first inference type; wherein, the first indication information is used to indicate the execution of at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability; the first indicator includes the distribution mode of data samples and / or the characteristics of data samples; the second indicator includes training duration and / or number of iterations; the third indicator includes at least one of the following: complexity, simulation environment, or the reason corresponding to the prediction result; the fourth indicator includes the performance indicator affected by the inference result and / or the performance indicator affected by the interference result; the processing module 501 is used to determine the first information based on the third information.

[0337] In one possible implementation, the third information includes the first indication information, which is used to indicate the execution of at least one of the following: data interpretability, process interpretability, or model interpretability; when determining the first information based on the third information, the processing module 501 is used to perform model training interpretability based on the third information to obtain the first information, which includes information about the model training.

[0338] In one possible implementation, the third information includes the first inference type; the first inference type includes management data analysis, and the information trained by the model includes interpretability information applicable to the management data analysis; or, the first inference type includes access network intelligence, and the information trained by the model includes interpretability information applicable to the access network intelligence; or, the first inference type includes network data analysis function, and the information trained by the model includes interpretability information applicable to the network data analysis function.

[0339] In one possible implementation, the third information includes the first indication information, which is used to indicate the interpretability of the reasoning; when determining the first information based on the third information, the processing module 501 is used to perform the interpretability of the model reasoning based on the third information to obtain the first information, which includes the information of the model reasoning.

[0340] In one possible implementation, the third information includes the first inference type; the first inference type includes management data analysis, and the information of the model inference may include at least one of the following: a second inference result corresponding to the management data analysis, information on the performance affected by the second inference result, a second interference result related to the second inference result, or information on the performance affected by the second interference result; or, the first inference type includes access network intelligence, and the information of the model inference may include at least one of the following: a third inference result corresponding to the access network intelligence, information on the performance affected by the third inference result, a third interference result related to the third inference result, or information on the performance affected by the third interference result; or, the first inference type includes network data analysis function, and the information of the model inference may include at least one of the following: a fourth inference result corresponding to the network data analysis function, information on the performance affected by the fourth inference result, a fourth interference result related to the fourth inference result, or information on the performance affected by the fourth interference result.

[0341] In one possible implementation, the third information can be carried in the information object class of the ML training request; or part of the third information can be carried in the information object class of the ML training request, and the remaining information can be carried in the information object class of the ML inference request.

[0342] In the second implementation, the communication device 500 can perform the functions of the second device, executing the following: a transceiver module 502, used to receive first information from the first device, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information; the model inference information includes inference interpretability information; and a processing module 501, used to perform corresponding processing based on the first information.

[0343] In one possible implementation, the first information includes information about the model training, which includes data interpretability information; when processing according to the first information, the processing module 501 is used to determine whether the first model needs to be retrained based on the data interpretability information.

[0344] In one possible implementation, the first information includes information about the model training, which includes process interpretability information; when performing corresponding processing based on the first information, the processing module 501 is used to determine, based on the process interpretability information, whether at least one of the following needs to be performed on the first model: testing, simulation, deployment, or retraining.

[0345] In one possible implementation, the first information includes information about the model training, which includes model interpretability information; when processing according to the first information, the processing module 501 is used to determine whether at least one of the following needs to be performed on the first model: testing, simulation, deployment, or retraining, based on the model interpretability information.

[0346] In one possible implementation, the first information includes the information of the model inference; when performing corresponding processing based on the first information, the processing module 501 is used to determine whether the first model needs to be retrained based on the inference interpretability information.

[0347] In one possible implementation, the data interpretability information may include the distribution pattern of the data samples (e.g., uniform or non-uniform distribution) and / or the characteristics of the data samples.

[0348] For example, the characteristics of the data sample may include at least one of the following: mobility, coverage, energy saving, load, fault, or service experience.

[0349] In one possible implementation, the process interpretability information may include training duration and / or number of iterations.

[0350] In one possible implementation, the model interpretability information may include at least one of the following: complexity, simulation environment, or the reason for the prediction result.

[0351] In one possible implementation, the reasoning interpretability information may include at least one of the following: a first reasoning result, information on the performance affected by the first reasoning result, a first interference result related to the first reasoning result, or information on the performance affected by the first interference result.

[0352] Optionally, the performance information affected by the first inference result may include the performance affected by the first inference result (or the performance metric affected by the first inference result) and / or the value of the performance affected by the first inference result.

[0353] Optionally, the performance affected by the first inference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0354] Optionally, the information on the performance affected by the first interference result may include the performance affected by the first interference result (or the performance index affected by the first interference result) and / or the value of the performance affected by the first interference result.

[0355] Optionally, the performance affected by the first interference result may include at least one of the following: cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

[0356] In one possible implementation, the first information can be carried in the information object class of the ML training report; or, the model training information can be carried in the information object class of the ML training report, and the model inference information can be carried in the information object class of the ML inference report.

[0357] In one possible implementation, the transceiver module 502 is further configured to receive second information from the first device, the second information being used to indicate the inference type to which the first information applies.

[0358] For example, the inference type may include at least one of the following: management data analysis, access network intelligence, or network data analysis functions.

[0359] In one possible implementation, the transceiver module 502 is further configured to send third information to the first device, the third information including at least one: first indication information, first indicator, second indicator, third indicator, fourth indicator, or first inference type; wherein, the first indication information is used to indicate the execution of at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability, the first indicator including the distribution mode of data samples and / or the characteristics of data samples, the second indicator including training duration and / or number of iterations, the third indicator including at least one of the following: complexity, simulation environment, or the reason corresponding to the prediction result, and the fourth indicator including the performance indicator affected by the inference result and / or the performance indicator affected by the interference result.

[0360] In one possible implementation, the first information includes information about the model training, and the third information includes the first inference type; the first inference type includes management data analysis, and the information about the model training includes interpretability information applicable to the management data analysis; or, the first inference type includes access network intelligence, and the information about the model training includes interpretability information applicable to the access network intelligence; or, the first inference type includes network data analysis functions, and the information about the model training includes interpretability information applicable to the network data analysis functions.

[0361] In one possible implementation, the first information includes the model inference information, and the third information includes the first inference type; the first inference type includes management data analysis, and the model inference information may include at least one of the following: a second inference result corresponding to the management data analysis, performance information affected by the second inference result, a second interference result related to the second inference result, or performance information affected by the second interference result; or, the first inference type includes access network intelligence, and the model inference information may include at least one of the following: a third inference result corresponding to access network intelligence, performance information affected by the third inference result, a third interference result related to the third inference result, or performance information affected by the third interference result; or, the first inference type includes network data analysis function, and the model inference information may include at least one of the following: a fourth inference result corresponding to the network data analysis function, performance information affected by the fourth inference result, a fourth interference result related to the fourth inference result, or performance information affected by the fourth interference result.

[0362] In one possible implementation, the third information can be carried in the information object class of the ML training request; or part of the third information can be carried in the information object class of the ML training request, and the remaining information can be carried in the information object class of the ML inference request.

[0363] Detailed descriptions of the above-mentioned processing module 501 and transceiver module 502 can be obtained directly from the relevant descriptions in the foregoing method embodiments, and will not be repeated here.

[0364] Figure 6 illustrates a schematic diagram of another communication device 600 provided in an embodiment of this application. The communication device 600 may include a processor 620, used to implement or support the communication device 600 in implementing the functions of the first or second device in the foregoing method embodiments. For details, please refer to the detailed descriptions in the foregoing method embodiments, which will not be repeated here. For example, the processor 620 is used to read and execute program instructions through the communication interface 610, so that the communication device 600 implements the corresponding method. The processor 620 may include one or more processors, without limitation.

[0365] It should be noted that the aforementioned functional modules can be implemented by hardware or by a combination of hardware and software, without limitation. Furthermore, when the communication device 600 includes only the processor 620, the communication device 600 can be a chip or a chip system.

[0366] For example, the communication device 600 can be a chip system. The chip system can be composed of chips or may include chips and other discrete components, without limitation.

[0367] For example, when the communication device 600 is a chip, the communication interface 610 can be the chip's input / output interface, where input corresponds to receiving operations and output corresponds to sending operations.

[0368] Optionally, the communication device 600 may further include a memory 630 for storing program instructions and / or data. The memory 630 is coupled to the processor 620. This coupling can be understood as an indirect coupling or communication connection between devices, units, or modules, and can be electrical, mechanical, or other forms, used for information exchange between devices, units, or modules. The processor 620 may operate in conjunction with the memory 630; the processor 620 and the memory 630 may be integrated together or disposed separately.

[0369] Furthermore, the processor 620 is used to execute program instructions stored in the memory 630 so that the communication device 600 implements the corresponding method.

[0370] One or more of the memories in memory 630 may be included in the processor, or memory 630 may exist independently, such as off-chip memory, and be connected to processor 620 via a communication bus (represented by thick line 640 in Figure 6). Memory 630 and processor 620 may also be integrated together.

[0371] Optionally, the communication device 600 further includes a communication interface 610 (shown as dashed lines in FIG. 6) for communicating with other devices via a transmission medium, thereby enabling the devices in the communication device 600 to communicate with other devices.

[0372] For example, when the communication device 600 is the first device, other devices can be second devices, etc. The processor 620 can use the communication interface 610 to send and receive data. For example, the processor 620 can be used to control the communication interface 610 to receive and / or send signals.

[0373] Specifically, the communication interface 610 can be a transceiver. In terms of hardware implementation, the transceiver can be used to implement the functions of the transceiver module 502 mentioned above, and the transceiver is integrated into the communication device 600 to form the communication interface 610.

[0374] Optionally, the transceiver may include a transmitter and / or a receiver to respectively implement the sending and receiving operations in the method embodiment; other operations besides sending and receiving may be implemented by the processor 620.

[0375] It should be noted that the communication interface 610 may have both sending and receiving functions, enabling the transmission and reception of signals; or it may have a sending function but no receiving function, used to transmit signals; or it may have a receiving function but no sending function, used to receive signals.

[0376] It should be noted that the specific connection medium between the communication interface 610, processor 620, and memory 630 is not limited in the embodiments of this application. Figure 6 shows the memory 630, processor 620, and communication interface 610 connected via a communication bus 640. The connection methods between other components are merely illustrative and not intended to be limiting. The communication bus 640 can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used in Figure 6, but this does not indicate that there is only one communication bus or one type of communication bus.

[0377] In the embodiments of this application, the processor 620 may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices. The general-purpose processor may be a microprocessor or any conventional processor. The methods disclosed in conjunction with the embodiments of this application may be executed by the hardware in the processor, or by a combination of hardware and software in the processor.

[0378] In this embodiment, the memory 630 can be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), or it can be volatile memory, such as random-access memory (RAM). The memory can also be any other medium used to carry or store program code in the form of instructions or data structures that can be accessed by a computer; or it can be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0379] In a first possible implementation, the communication device 600 may be a first device used to implement the relevant methods corresponding to the first device in the above embodiments. For specific functions, please refer to the descriptions in the above embodiments.

[0380] For example, the methods corresponding to the first device in the above embodiments include: determining first information of the first model, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the model inference information includes inference interpretability information; sending the first information to a second device, the second device being a consumer of the first model.

[0381] In a second possible implementation, the communication device 600 may be a second device used to implement the methods corresponding to the second device in the above embodiments. For specific functions, please refer to the descriptions in the above embodiments.

[0382] For example, the methods corresponding to the second device in the above embodiments include: receiving first information from the first device, the first information including model training information and / or model inference information, wherein the model training information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the model inference information includes inference interpretability information; and performing corresponding processing based on the first information.

[0383] For the specific implementation process, please refer to the relevant content in the aforementioned embodiments; it will not be repeated here.

[0384] Figure 7 exemplarily illustrates a structural schematic diagram of another communication device 700 provided in an embodiment of this application. It is understood that the communication device 700 includes means of necessary forms, such as modules, units, elements, circuits, or interfaces, to be appropriately configured together to execute this solution. The communication device 700 shown in Figure 7 can be an access network device or a component within an access network device (e.g., a chip or communication module), or a core network device or a component within a core network device (e.g., a chip or communication module), or a terminal device or a component within a terminal device (e.g., a chip or communication module), and can be used to perform the operation of the first or second device in the above method embodiments. The communication device 700 includes one or more processors 701. The processor 701 can be a general-purpose processor or a dedicated processor, etc. Optionally, the processor 701 can include a baseband processor and / or a central processing unit; or, the processor 701 can integrate the functions of a baseband processor and a central processing unit. The specific details of the processor 701 can be found in the description of the processor 620 above, and will not be repeated here.

[0385] In one possible design, processor 701 may include program 703. Program 703 may be run on processor 701 to cause communication device 700 to perform the methods described in the above method embodiments. In another possible design, communication device 700 includes circuitry (not shown in FIG. 7) for performing the methods in the above method embodiments; or, in other words, the circuitry may be used to indicate the function of the first or second device in the above method embodiments.

[0386] Optionally, the communication device 700 may include one or more memories 702. The memories 702 store a program 704, which can be executed on the processor 701 to cause the communication device 700 to perform the methods described in the above method embodiments.

[0387] Optionally, processor 701 may include AI module 707, and / or memory 702 may include AI module 708. The AI ​​module can be used to implement AI-related functions. For example, the AI ​​module can be used to perform model training and / or model inference. The AI ​​module can be implemented through software, hardware, or a combination of both. For example, the AI ​​module may include a radio intelligence control (RIC) module. For example, the AI ​​module can be a near real-time RIC or a non-real-time RIC.

[0388] Optionally, data may also be stored in the processor 701 and / or the memory 702. The processor and memory may be configured separately or integrated together.

[0389] Optionally, the communication device 700 may further include a transceiver 705 and / or an antenna 706. The transceiver 705 may also be referred to as a transceiver module, transceiver unit, transceiver, transceiver circuit, or transceiver, etc., and can be used to realize the transmission and reception functions of the communication device through the antenna 706.

[0390] It should be noted that the module division in the above embodiments of this application is illustrative and only represents a logical functional division. In actual implementation, there may be other division methods. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, exist as separate physical units, or have two or more units integrated into one unit. The integrated units can be implemented in hardware, as software functional units, or in a combination of hardware and software. Whether a function is executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0391] For example, the functional unit in any of the above devices may be one or more integrated circuits configured to implement the above methods, such as one or more ASICs, one or more CPUs, one or more MCUs, one or more DSPs, or one or more FPGAs, or a combination of at least two of these integrated circuit forms.

[0392] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0393] This application also provides a communication system, which may include one or more of the following: a first device or a second device. The first device or the second device can be found in the descriptions of the foregoing method embodiments, and will not be repeated here.

[0394] This application also provides a computer-readable storage medium for storing computer programs or instructions, which, when run, enable the methods or steps executed by the first or second device in the foregoing embodiments to be implemented.

[0395] This application also provides a computer program product, including a computer program, which, when run on a computer, causes the methods or steps executed by the first or second device in the foregoing embodiments to be implemented.

[0396] This application provides a chip system including a processor for implementing the functions of the first or second device in the aforementioned method (e.g., executing corresponding methods or steps). The chip system may be composed of a chip or may include a chip and other discrete devices.

[0397] Optionally, the chip system also includes a memory for storing program instructions that the processor can read and execute to implement the corresponding method.

[0398] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0399] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0400] In the various embodiments of this application, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of different embodiments are consistent and can be referenced by each other. The technical features of different embodiments can be combined to form new embodiments according to their inherent logical relationship.

[0401] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0402] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0403] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0404] It is understood that the various numerical designations used in the embodiments of this application are merely for descriptive convenience and are not intended to limit the scope of the embodiments of this application. The order of the process numbers described above does not imply the order of execution; the execution order of each process should be determined by its function and internal logic.

[0405] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A communication method applied to a first device, characterized in that, The method includes: Determine the first information of the first model, wherein the first information includes at least one of the following: data interpretability information, process interpretability information, model interpretability information, or reasoning interpretability information; The first information is sent to a second device, which is a consumer of the first model.

2. The method according to claim 1, characterized in that, The method further includes: Send a second message to the second device, the second message being used to indicate the type of reasoning to which the first message applies.

3. The method according to claim 2, characterized in that, The inference type to which the first information applies includes at least one of the following: Management data analysis, access network intelligence, or network data analysis functions.

4. The method according to any one of claims 1 to 3, characterized in that, The first information for determining the first model includes: The system receives third information from the second device, the third information including at least one of the following: first indication information, first indicator, second indicator, third indicator, fourth indicator, or first inference type; wherein, the first indication information is used to indicate the execution of at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability, the first indicator including the distribution mode of data samples and / or the characteristics of data samples, the second indicator including training duration and / or number of iterations, the third indicator including at least one of the following: complexity, simulation environment, or the reason corresponding to the prediction result, and the fourth indicator including the performance indicator affected by the inference result and / or the performance indicator affected by the interference result; The first information is determined based on the third information.

5. The method according to claim 4, characterized in that, The third information includes the first indication information, which is used to indicate the execution of at least one of the following: data interpretability, process interpretability, or model interpretability; Determining the first information based on the third information includes: The first information is obtained by performing model training interpretability based on the third information, wherein the first information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information.

6. The method according to claim 5, characterized in that, The third information includes the first reasoning type; The first reasoning type includes management data analysis, and the first information includes interpretability information applicable to the management data analysis; or, The first reasoning type includes access network intelligence, and the first information includes interpretability information applicable to the access network intelligence; or, The first reasoning type includes network data analysis functionality, and the first information includes interpretability information applicable to the network data analysis functionality.

7. The method according to any one of claims 4 to 6, characterized in that, The third information includes the first indication information, which is used to indicate the interpretability of the reasoning. Determining the first information based on the third information includes: The first information is obtained based on the interpretability of the third information execution model, and the first information includes the reasoning interpretability information.

8. The method according to claim 7, characterized in that, The third information includes the first reasoning type; The first inference type includes management data analysis, and the first information includes at least one of the following: a second inference result corresponding to the management data analysis, performance information affected by the second inference result, a second interference result related to the second inference result, or performance information affected by the second interference result; or, The first inference type includes access network intelligence, and the first information includes at least one of the following: a third inference result corresponding to the access network intelligence, performance information affected by the third inference result, a third interference result related to the third inference result, or performance information affected by the third interference result; or, The first inference type includes a network data analysis function, and the first information includes at least one of the following: a fourth inference result corresponding to the network data analysis function, information on the performance affected by the fourth inference result, a fourth interference result related to the fourth inference result, or information on the performance affected by the fourth interference result.

9. The method according to any one of claims 4 to 8, characterized in that, The third information is carried in the information object class of the machine learning (ML) training request; or part of the third information is carried in the information object class of the ML training request, and the remaining information is carried in the information object class of the ML inference request.

10. The method according to any one of claims 1 to 9, characterized in that, The data interpretability information includes the distribution pattern of the data samples and / or the characteristics of the data samples.

11. The method according to claim 10, characterized in that, The data sample features include at least one of the following: mobility, coverage, energy saving, load, fault, or service experience.

12. The method according to any one of claims 1 to 11, characterized in that, The process interpretability information includes training duration and / or number of iterations.

13. The method according to any one of claims 1 to 12, characterized in that, The model interpretability information includes at least one of the following: complexity, simulation environment, or the reason for the prediction result.

14. The method according to any one of claims 1 to 13, characterized in that, The reasoning interpretability information includes at least one of the following: a first reasoning result, information on the performance affected by the first reasoning result, a first interference result related to the first reasoning result, or information on the performance affected by the first interference result.

15. The method according to claim 14, characterized in that, The performance affected by the first inference result includes at least one of the following: cell energy consumption, cell energy efficiency, cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution; and / or, the performance affected by the first interference result includes at least one of the following: cell energy consumption, cell energy efficiency, cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

16. A communication method applied to a second device, characterized in that, The method includes: Receive first information from a first device, the first information including at least one of the following: data interpretability information, process interpretability information, model interpretability information, or reasoning interpretability information; Perform corresponding processing based on the first information.

17. The method according to claim 16, characterized in that, The method further includes: Receive second information from the first device, the second information being used to indicate the type of reasoning to which the first information applies.

18. The method according to claim 17, characterized in that, The inference type to which the first information applies includes at least one of the following: Management data analysis, access network intelligence, or network data analysis functions.

19. The method according to any one of claims 16 to 18, characterized in that, The method further includes: Send a third message to the first device, the third message including at least one of the following: a first instruction message, a first indicator, a second indicator, a third indicator, a fourth indicator, or a first inference type; wherein, the first instruction message is used to instruct the execution of at least one of the following: data interpretability, process interpretability, model interpretability, or inference interpretability, the first indicator including the distribution mode of data samples and / or the characteristics of data samples, the second indicator including training duration and / or number of iterations, the third indicator including at least one of the following: complexity, simulation environment, or the reason corresponding to the prediction result, and the fourth indicator including the performance indicator affected by the inference result and / or the performance indicator affected by the interference result.

20. The method according to claim 19, characterized in that, The first information includes at least one of the following: data interpretability information, process interpretability information, or model interpretability information, and the third information includes the first reasoning type; The first reasoning type includes management data analysis, and the first information includes interpretability information applicable to the management data analysis; or, The first reasoning type includes access network intelligence, and the first information includes interpretability information applicable to the access network intelligence; or, The first reasoning type includes network data analysis functionality, and the first information includes interpretability information applicable to the network data analysis functionality.

21. The method according to claim 19 or 20, characterized in that, The first information includes the reasoning interpretability information, and the third information includes the first reasoning type; The first inference type includes management data analysis, and the first information includes at least one of the following: a second inference result corresponding to the management data analysis, performance information affected by the second inference result, a second interference result related to the second inference result, or performance information affected by the second interference result; or, The first inference type includes access network intelligence, and the first information includes at least one of the following: a third inference result corresponding to the access network intelligence, performance information affected by the third inference result, a third interference result related to the third inference result, or performance information affected by the third interference result; or, The first inference type includes a network data analysis function, and the first information includes at least one of the following: a fourth inference result corresponding to the network data analysis function, information on the performance affected by the fourth inference result, a fourth interference result related to the fourth inference result, or information on the performance affected by the fourth interference result.

22. The method according to any one of claims 19 to 21, characterized in that, The third information is carried in the information object class of the machine learning (ML) training request; or part of the third information is carried in the information object class of the ML training request, and the remaining information is carried in the information object class of the ML inference request.

23. The method according to any one of claims 16 to 22, characterized in that, The data interpretability information includes the distribution pattern of the data samples and / or the characteristics of the data samples.

24. The method according to claim 22, characterized in that, The data sample features include at least one of the following: mobility, coverage, energy saving, load, fault, or service experience.

25. The method according to any one of claims 16 to 24, characterized in that, The process interpretability information includes training duration and / or number of iterations.

26. The method according to any one of claims 16 to 25, characterized in that, The model interpretability information includes at least one of the following: complexity, simulation environment, or the reason for the prediction result.

27. The method according to any one of claims 16 to 26, characterized in that, The reasoning interpretability information includes at least one of the following: a first reasoning result, information on the performance affected by the first reasoning result, a first interference result related to the first reasoning result, or information on the performance affected by the first interference result.

28. The method according to claim 27, characterized in that, The performance affected by the first inference result includes at least one of the following: cell energy consumption, cell energy efficiency, cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution; and / or, the performance affected by the first interference result includes at least one of the following: cell energy consumption, cell energy efficiency, cell throughput, cell resource utilization, number of radio resource control connections, number of successful cell handovers, number of failed cell handovers, or cell signal strength distribution.

29. A communication device, characterized in that, Includes modules for performing the method as described in any one of claims 1 to 28.

30. A communication device, characterized in that, It includes at least one processor, said at least one processor being used to perform the method as described in any one of claims 1 to 28.

31. A communication system, characterized in that, It includes a first device and / or a second device, wherein the first device is used to perform the method as described in any one of claims 1 to 15, and the second device is used to perform the method as described in any one of claims 16 to 28.

32. A computer-readable storage medium, characterized in that, The device stores a computer program or instructions that, when executed, cause the method as described in any one of claims 1 to 28 to be implemented.

33. A computer program product, characterized in that, The computer program product includes a computer program that, when run on a computer, causes the method as described in any one of claims 1 to 28 to be implemented.

Images