Communication method and related device

Abstract

The invention provides a communication method and a communication device, which are applied to the technical field of communication. In the technical scheme provided by the invention, a receiving node acquires an output result of a first artificial intelligence (AI) model, and adds interpretability analysis on the output result, and the receiving node acquires an interpretable result of the output result, and performs reasonability judgment on the output result of an inference model. Therefore, the model interpretability analysis and interaction are added in the communication network AI application, the reasonability of the AI reasoning result is improved, and the reliability of the AI algorithm applied to the communication network is ensured.

Description

Technical Field

[0001] This application relates to the field of communication technologies, and in particular, to communication methods and related devices. Background Art

[0002] With the development of artificial intelligence (AI) and the continuous improvement of the performance requirements of communication systems, AI technologies are applied in more and more scenarios in communication systems.

[0003] Currently, the technical solutions for applying AI technologies in communication systems are as follows: A receiving node sends an AI task request to a model management node or provides data required for AI inference; the model management node completes AI inference locally or jointly with a computing network element; the model management node feeds back the inference result to the receiving node; the receiving node returns a result confirmation or feedback information to the model management node. The first step and the fourth step among them can be optionally executed steps respectively. Among them, the receiving node is the requester of AI inference, for example, it can be a terminal; the model management node is the node that provides and executes AI services in the communication network.

[0004] However, when applying AI technologies in communication systems with the above technical solutions, the following problems will occur: The inference result of the AI model fed back by the model management node to the receiving node lacks reliability judgment, resulting in the inability to guarantee the reliability of the communication performance of the communication system. Summary of the Invention

[0005] This application provides a communication method and related devices, which can realize the reliable feedback of the inference result of the AI model in the communication system, thereby improving the communication performance.

[0006] In a first aspect, this application provides a communication method, which is used for a communication node, or is applied to a device, module, circuit or chip in the communication node, or is a device that can be used in matching with the communication node. The method includes: obtaining the inference result of a first artificial intelligence (AI) model; obtaining the interpretable result of the inference result.

[0007] In this method, an AI model interpretable mechanism is added to the communication node to obtain the interpretable result of the inference result of the AI model. Since the interpretable result can be used to interpret the AI model, it can be used to judge the rationality of the inference result of the AI model, and thus can be used to evaluate the reliability after the inference result is used. It can be seen that obtaining the interpretable result can provide technical support for judging the credibility and rationality of the inference result of the AI model, which helps to ensure the reliability of the AI algorithm in the communication system, and further improves the communication performance.

[0008] It can be understood that obtaining the interpretable result in this method may include: receiving the interpretable result, or the communication node itself determining the interpretable result.

[0009] In some possible implementation manners, this method may further include: sending a first piece of information, where the first piece of information indicates first interpretable-related information, and the first interpretable-related information is interpretable-related information of an AI model. The interpretable-related information of the AI model includes at least one of the following pieces of information: an interpretable method of the AI model, an output type, an interpretation content of the interpretable method of the AI model, or an interpretation accuracy of the interpretable method of the AI model. The output type is the type of the output result of the AI model. Among them, the first piece of information is used for the communication node that receives the first piece of information to determine the interpretable result of the inference result of the AI model.

[0010] In this implementation manner, optionally, obtaining the inference result of the AI model may include: receiving the inference result of the AI model; obtaining the interpretable result may include: receiving the interpretable result. Among them, sending the first piece of information may be before receiving the interpretable result, and further, may also be before receiving the inference result.

[0011] In this implementation manner, the current communication node provides a model interpretability metric for the communication node that provides the inference result of the AI model, so that the communication node that receives the first piece of information can obtain an interpretable result that better meets the requirements of the current communication node, thereby improving the rationality of model interpretability analysis.

[0012] In some possible implementation manners, the first piece of information is one of multiple interpretable-related information, where the first piece of information includes an index of the first interpretable-related information in the multiple interpretable-related information.

[0013] Or rather, the current communication node uses the index to indicate to the communication node that determines the interpretable result the interpretable-related information required to determine the interpretability result.

[0014] In some possible implementation manners, this method further includes: receiving a second piece of information, where the second piece of information indicates the interpretable-related information associated with each output type in at least one output type.

[0015] Or rather, the current communication node learns about the interpretable-related information supported by the communication node that determines the interpretable result through the second piece of information. In this case, when the current communication node sends the first piece of information, the interpretable-related information indicated by the first piece of information may be determined by the second piece of information. For example, the interpretable-related information indicated by the first piece of information may be the interpretable-related information supported by the communication node that determines the interpretable result.

[0016] In some possible implementations, the method further includes: receiving third information, where the third information indicates information required to obtain an interpretable result of the inference result of the first AI model; wherein, obtaining the interpretable result of the inference result includes: determining the interpretable result according to the third information.

[0017] Or rather, input data required for interpretability analysis is obtained through the third information, and the input data is used for interpretability analysis to obtain an interpretable result. For example, when the current communication node determines the interpretable result by itself, it can receive the third information.

[0018] In some possible implementations, the method further includes: sending fourth information, where the fourth information is used to request information required to obtain an interpretable result of the inference result of the first AI model.

[0019] Or rather, the current communication node requests the input data required for interpretability analysis through the fourth information. For example, when the current communication node determines the interpretable result by itself, it can send the fourth information.

[0020] In some possible implementations, the method further includes: sending the interpretable result.

[0021] For example, when the interpretable result is used by other communication nodes to analyze the rationality of the inference result, the current communication node can send the interpretable result determined by itself to other communication nodes.

[0022] In some possible implementations, obtaining the interpretable result of the inference result includes: receiving the interpretable result. For example, when the interpretable result is analyzed by other communication nodes, the current communication node can receive the interpretable result from other communication nodes.

[0023] In some possible implementations, the method further includes: sending fifth information, where the fifth information indicates the rationality of the inference result of the first AI model.

[0024] For example, when the current communication node is a node that uses the interpretable result, the current communication node can determine or rather judge the rationality of the inference result based on the interpretable result, and feedback the rationality of the inference result to other communication nodes that determine the inference result, so that other communication nodes can perform processing such as adjusting the AI model to improve the reliability of the inference result, thereby improving communication reliability.

[0025] In a second aspect, the present application provides a communication device, including modules or units for implementing the method in the first aspect and any possible implementation manner of the first aspect. It should be understood that each module or unit can implement the corresponding function by executing a computer program.

[0026] In a third aspect, the present application provides a communication device, including a processor configured to execute the communication method according to the first aspect or any one of the possible implementation manners in the first aspect. The communication device may be a chip or a chip system applied to a terminal device.

[0027] The device may further include a memory for storing instructions and data. The memory is coupled to the processor, and when the processor executes the instructions stored in the memory, the method described in the above first aspect or any one of the possible implementation manners thereof can be implemented. The device may further include a communication interface for communicating the device with other devices. Exemplarily, the communication interface may be a transceiver, a circuit, a bus, a module, or other types of communication interfaces.

[0028] In a fourth aspect, the present application provides a computer-readable storage medium storing program code for a communication device to execute, where the program code includes instructions for implementing the method according to the first aspect and any one of the possible implementation manners in the first aspect.

[0029] In a fifth aspect, the present application provides a computer program product including instructions, which, when the computer program product runs on a communication device, causes the communication device to implement the method according to the first aspect and any one of the possible implementation manners in the first aspect.

[0030] In a sixth aspect, the present application provides a communication system including a communication device for implementing the method according to the first aspect and any one of the possible implementation manners in the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a schematic diagram of a communication system applicable to the method of the embodiment of the present application;

[0032] Figure 2 is a schematic diagram of another communication system applicable to the method of the embodiment of the present application;

[0033] Figure 3 is a schematic flowchart of the communication method provided by an embodiment of the present application;

[0034] Figure 4 is a schematic flowchart of the communication method provided by an embodiment of the present application;

[0035] Figure 5 is a schematic structural diagram of the communication device provided by an embodiment of the present application;

[0036] Figure 6 is a schematic structural diagram of the communication device provided by an embodiment of the present application;

[0037] Figure 7Schematic structural diagram of a communication device provided by an embodiment of the present application;

[0038] Figure 8 Schematic structural diagram of a communication device provided by another embodiment of the present application. Detailed implementation manners

[0039] Next, the technical solutions in the embodiments of the present application will be described with reference to the accompanying drawings in the embodiments of the present application.

[0040] For the convenience of clearly describing the technical solutions in the embodiments of the present application, in the embodiments of the present application, terms such as "first" and "second" are used to distinguish identical or similar items with basically the same functions and roles. Those skilled in the art can understand that the terms "first", "second", etc. do not limit the quantity and execution order, and the terms "first", "second", etc. do not necessarily limit being different.

[0041] It should be noted that in the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "exemplary" or "for example" aims to present relevant concepts in a specific manner.

[0042] In the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone, where A and B may be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after. "At least one (item)" or its similar expression below refers to any combination of these items, including any combination of single item (item) or plural items (items). For example, at least one (item) of a, b, and (or) c may represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c may be single or multiple.

[0043] To facilitate the understanding of the communication method provided by the embodiments of the present application, the system architecture and application scenarios of the communication method provided by the embodiments of the present application will be described below. It can be understood that the system architecture and application scenarios described in the embodiments of the present application are for more clearly explaining the technical solutions in the embodiments of the present application, and do not constitute a limitation to the technical solutions provided by the embodiments of the present application.

[0044] In recent years, due to the rapid development of machine learning, especially deep learning technology, the capabilities of AI in fields such as prediction, fitting, classification, and online autonomous decision-making have been continuously improving, bringing more emerging scenarios while strengthening many traditional applications. For mobile systems, machine learning technology has been considered for application in the physical layer to replace traditional modules such as channel prediction. In addition, in application scenarios involving interactions between nodes, AI has also brought more methods for predicting communication events or behaviors such as load balancing and mobility optimization. Facing problems such as high power consumption introduced by 5G, various AI methods based on deep learning (DL) or reinforcement learning (RL) have been introduced to formulate reasonable energy-saving strategies.

[0045] It should be noted that the AI application interactions of multiple nodes involved in the communication system of this application include, but are not limited to, channel prediction, mobility optimization, load balancing, AI prediction information interaction between base stations, etc. This application does not focus on describing the AI application scenarios in the communication system. Any communication system that can implement the functional set of AI applications in the communication system should be included in the protection scope of this application.

[0046] In the application of the communication field combined with AI technology, the technical solution provided by this application can be applied to various communication systems, such as: the fifth generation (5G) or new radio (NR) system, the long term evolution (LTE) system, the LTE frequency division duplex (FDD) system, the LTE time division duplex (TDD) system, the wireless local area network (WLAN) system, the satellite communication system, future communication systems, such as the sixth generation (6G) mobile communication system, or a fusion system of multiple systems, etc. The technical solution provided by this application can also be applied to device-to-device (D2D) communication, vehicle-to-everything (V2X) communication, machine-to-machine (M2M) communication, machine type communication (MTC), and the Internet of Things (IoT) communication system or other communication systems, such as in the sixth generation (6G).

[0047] A device in a communication system can send signals to another device or receive signals from another device. The signals can include information, signaling, data, etc. Herein, the device can also be replaced by an entity, a network entity, a communication device, a communication module, a node, a communication node, etc. In this application, a communication node is taken as an example for description. For example, a communication system can include multiple communication nodes, such as at least one receiving node and a model management node. The model management node can send signals to the receiving node, and / or the receiving node can send signals to the model management node. Among them, the receiving node can be the requester of AI inference, responsible for receiving, analyzing, and using the AI inference results; the model management node can be the providing and executing node of AI services in the communication network, responsible for storing, training, and managing the AI models corresponding to tasks, and performing AI task inference in response to the requests of the receiving nodes.

[0048] In an embodiment of this application, the receiving node can be a terminal device or a network device.

[0049] Among them, the terminal device can also be referred to as a user equipment (UE), an access terminal, a user unit, a user station, a mobile station, a mobile unit, a remote station, a remote terminal, a mobile device, a user terminal, a terminal, a wireless communication device, a user agent, or a user device.

[0050] The terminal device can be a device that provides voice / data. For example, it can be a handheld device, a vehicle-mounted device, etc. with wireless connection capabilities. Currently, some examples of terminals are: mobile phone, tablet computer, laptop computer, palmtop computer, mobile internet device (MID), wearable device, virtual reality (VR) device, augmented reality (AR) device, wireless terminal in industrial control, wireless terminal in self-driving, wireless terminal in remote medical surgery, wireless terminal in smart grid, wireless terminal in transportation safety, wireless terminal in smart city, wireless terminal in smart home, cellular phone, cordless phone, session initiation protocol (SIP) phone, wireless local loop (WLL) station, personal digital assistant (PDA), a handheld device with wireless communication capabilities, a computing device, or other processing devices connected to a wireless modem, wearable device, terminal device in a 5G network, or terminal device in a future evolved public land mobile network (PLMN), etc. The embodiments of this application are not limited thereto.

[0051] By way of example and not limitation, in the embodiments of this application, the terminal device can also be a wearable device. A wearable device can also be referred to as a wearable intelligent device, which is a general term for devices developed by applying wearable technology to the intelligent design of daily wear, such as glasses, gloves, watches, clothing, and shoes, etc. A wearable device is a portable device that is either directly worn on the body or integrated into the user's clothes or accessories. A wearable device is not just a hardware device, but more importantly, it realizes powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable intelligent devices include those with complete functions and large sizes that can achieve complete or partial functions without relying on a smartphone, such as smart watches or smart glasses, etc., and those that only focus on a certain type of application function and need to cooperate with other devices such as smartphones, such as various smart bracelets and smart jewelry for monitoring physical signs.

[0052] In the embodiments of the present application, the device for implementing the functions of the terminal device may be the terminal device itself, or a device capable of supporting the terminal device to implement such functions, such as a chip system. This device may be installed in the terminal device or used in combination with the terminal device. In the embodiments of the present application, the chip system may be composed of chips, or may include chips and other discrete devices. In the embodiments of the present application, only the case where the device for implementing the functions of the terminal device is the terminal device is used as an example for illustration, which does not limit the solutions of the embodiments of the present application.

[0053] The model management node in the embodiments of this application can be a network device. Among them, the network device is a device used to communicate with terminal devices, and this network device can also be referred to as an access network device or a radio access network device. For example, the network device can be a base station. The network device in the embodiments of this application can refer to a radio access network (RAN) node (or device) that connects a terminal device to a wireless network. The base station can generally cover various names in the following, or be replaced with the following names. For example: Node B, evolved Node B (eNB), next generation Node B (gNB), relay station, access point, transmitting and receiving point (TRP), transmitting point (TP), master station, secondary station, multi standard radio (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), radio unit (RU), positioning node, etc. The base station can be a macro base station, a micro base station, a relay node, a donor node or the like, or a combination thereof. The base station can also refer to a communication module, a modem or a chip used in the aforementioned devices or apparatuses. The base station can also be a mobile switching center and devices that undertake the functions of a base station in D2D, V2X, M2M communications, network-side devices in 6G networks, devices that undertake the functions of a base station in future communication systems, etc. The base station can support networks with the same or different access technologies. Optionally, the RAN node can also be a server, a wearable device, a vehicle or an in-vehicle device, etc. For example, the access network device in vehicle to everything (V2X) technology can be a road side unit (RSU). The embodiments of this application do not limit the specific technologies and specific device forms adopted by the network device.In some deployments, the network device mentioned in the embodiments of this application may be a device including a CU, or a DU, or a device including a CU and a DU, or a control plane CU node (Central Unit-Control Plane (CU-CP)) and a user plane CU node (Central Unit-User Plane (CU-UP)) and a DU node. For example, the network device may include a gNB-CU-CP, a gNB-CU-UP, and a gNB-DU.

[0054] In some deployments, multiple RAN nodes cooperate to assist a terminal in achieving wireless access, and different RAN nodes respectively implement some functions of a base station. For example, the RAN node may be a CU, a DU, a CU-CP, a CU-UP, or an RU, etc. The CU and the DU may be separately provided, or may also be included in the same network element, such as a BBU. The RU may be included in a radio frequency device or a radio frequency unit, such as included in an RRU, an AAU, or an RRH.

[0055] The RAN node may support one or more types of fronthaul interfaces. Different fronthaul interfaces respectively correspond to DUs and RUs with different functions. If the fronthaul interface between the DU and the RU is a Common Public Radio Interface (CPRI), the DU is configured to implement one or more of the baseband functions, and the RU is configured to implement one or more of the radio frequency functions. If the fronthaul interface between the DU and the RU is another interface, compared with the CPRI, some of the downlink and / or uplink baseband functions, for example, for the downlink, one or more of precoding, digital beamforming (BF), or inverse fast Fourier transform (IFFT) / adding cyclic prefix (CP), are moved from the DU to the RU for implementation. For the uplink, one or more of digital beamforming (BF), or fast Fourier transform (FFT) / removing cyclic prefix (CP) are moved from the DU to the RU for implementation. In a possible implementation manner, this interface may be an Enhanced Common Public Radio Interface (eCPRI). In the eCPRI architecture, the splitting method between the DU and the RU is different, corresponding to different categories (Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, F.

[0056] Taking eCPRI Cat A as an example, for downlink transmission, with layer mapping as the division, the DU is configured to implement layer mapping and one or more functions before it (i.e., one or more of encoding, rate matching, scrambling, modulation, and layer mapping), while other functions after layer mapping (such as one or more of RE mapping, digital beamforming (BF), or inverse fast Fourier transform (IFFT) / adding cyclic prefix (CP)) are moved to the RU for implementation. For uplink transmission, with de-RE mapping as the division, the DU is configured to implement demapping and one or more functions before it (i.e., one or more of decoding, derate matching, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and de-RE mapping), while other functions after demapping (such as one or more of digital BF or fast Fourier transform (FFT) / removing CP) are moved to the RU for implementation. It can be understood that for the function descriptions of DU and RU corresponding to various types of eCPRI, reference can be made to the eCPRI protocol and will not be elaborated here.

[0057] In a possible design, the processing unit in the BBU for implementing baseband functions is called the baseband high (BBH) unit, and the processing unit in the RRU / AAU / RRH for implementing baseband functions is called the baseband low (BBL) unit.

[0058] In different systems, the CU (or CU-CP and CU-UP), DU, or RU may also have different names, but those skilled in the art can understand their meanings. For example, in the ORAN system, the CU can also be called O-CU (open CU), the DU can also be called O-DU, the CU-CP can also be called O-CU-CP, the CU-UP can also be called O-CU-UP, and the RU can also be called O-RU. Any one of the CU (or CU-CP, CU-UP), DU, and RU in this application can be implemented through software modules, hardware modules, or a combination of software modules and hardware modules.

[0059] In the embodiments of the present application, the device for implementing the functions of a network device may be a network device; it may also be a device capable of supporting the network device to implement such functions, such as a chip system, a hardware circuit, a software module, or a combination of a hardware circuit and a software module. This device may be installed in the network device or used in conjunction with the network device. In the embodiments of the present application, only the case where the device for implementing the functions of the model management node is a network device is taken as an example for illustration, which does not limit the solutions of the embodiments of the present application.

[0060] The network device and / or the terminal device may be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they may also be deployed on the water surface; they may also be deployed on aircraft, balloons, and satellites in the air. In the embodiments of the present application, the scenarios where the network device and the terminal device are located are not limited. In addition, the terminal device and the network device may be hardware devices, or software functions running on dedicated hardware or software functions running on general hardware. For example, they are virtualized functions instantiated on a platform (such as a cloud platform), or entities including dedicated or general hardware devices and software functions. The present application does not limit the specific forms of the terminal device and the network device.

[0061] Figure 1 It is a schematic diagram of a communication system applicable to the method of the embodiments of the present application. As Figure 1 shown, the communication system may include a receiving node and a model management node.

[0062] The model management node may be simply referred to as the management node, and may be a base station with computing power or a computing network element in the network; the receiving node may be a terminal or a base station.

[0063] The management node is responsible for training the model and performing inference on the requested tasks, and the receiving node initiates and receives the inference results of the AI tasks. When the management node completes the AI task inference and feeds back the inference results to the receiving node, the management node may perform model interpretability analysis and inform the receiving node of the interpretability analysis results, or the receiving node may perform model interpretability analysis by itself to obtain the interpretability analysis results, so as to assist the receiving node in judging the inference results and subsequent operations.

[0064] Figure 2 It is a schematic diagram of another communication system applicable to the method of the embodiments of the present application. As Figure 2 shown, compared with the communication system Figure 1 shown, the communication system may further include a third node.

[0065] The third node may be a dedicated model management network element for AI model evaluation. For example, the third node may be a base station with computing power or a computing network element in the network. One difference between the third node and the management node is that the third node is not the node performing the AI inference for this task.

[0066] Figure 2 In the communication system shown, the model interpretability analysis can be performed by a third-party node, and the results of the interpretability analysis can be informed to the receiving node to assist the receiving node in judging the inference results and subsequent operations.

[0067] Figure 1 or Figure 2 In different scenarios of the integration of the shown communication system and AI technology, the AI model can perform different functional operations. For example, when the receiving node is a UE and the model management node is a gNB, the corresponding UE will request AI tasks such as resource scheduling, prediction instructions, and sensing from the gNB; when the receiving node is a gNB and the model management node is a gNB, in this cross-site scenario, the AI model can be applied to mobile enhancement, handover prediction, terminal behavior prediction, etc.; when the receiving node is a gNB and the model management node is a NodeC, it can be applied to most RAN AI scenarios, and the base station requests the AI inference results from the computing network element.

[0068] It should be noted that Figure 1 and Figure 2 are only simplified schematic diagrams shown for ease of understanding. In actual applications, the communication system may include multiple receiving nodes and may also include multiple model management nodes. The embodiments of the present application do not limit the number of receiving nodes and model management nodes included in the communication system.

[0069] Most of the existing mainstream AI models adopt an end-to-end training and inference mode. The AI models have a "black box effect", and there is no unified and general explanation for how the training data guides the update of the AI model and how the AI model understands and processes the input data. In addition, complex models and large models with general capabilities are gradually becoming the mainstream, and there is a negative correlation between the complexity of the model and its interpretability. This also makes it more difficult to interpret more complex models, resulting in an increase in the uncertainty of the credibility of the output results.

[0070] Therefore, explainable AI (XAI) has also become a hot topic in academic and industrial circles. The interpretability method refers to the method of understanding and trusting the inference / training process and results of machine learning models, so as to help describe the accuracy, fairness, transparency, and rationality of the models.

[0071] Many existing XAI methods provide a large number of options for explaining models, such as triggering from dimensions like text, images, feature importance, etc., to generate AI interpretable solutions that can be understood by humans. Among them, XAI can be divided into inherently interpretable and post hoc interpretable from the inherent interpretability of the model. The former represents that the mechanism and design of the model itself have the ability to be interpretable, such as models like decision trees, feature decomposition, support vector machine (SVM), etc.; the latter requires additional interpretability analysis for the model, and is mostly applicable to deep learning models such as neural network (NN), transformer, etc. Among them, for post hoc model-agnostic XAI methods, popular methods include local interpretable model-agnostic explanations (Lime), that is, training an interpretable equivalent linear model through local samples, the explanation method based on the Shapley value of each feature (shapley additive explanations, SHAP), finding the range of the smallest interpretable subset of features (Anchors method), etc.

[0072] For the AI methods concerned by mobile networks, multiple use cases introduced by the current NG-RAN AI all involve interactions between different nodes. The content of the interaction includes inputs, outputs related to the AI model, and related inference strategies, etc. The AI models used by different nodes are usually based on internal implementations, that is, they will not directly interact between different device manufacturers. Therefore, XAI methods are needed to explain the model or the output results.

[0073] An exemplary application scenario of the communication method of this application is when interacting with the prediction and inference results of the AI model between communication nodes, using the XAI method to obtain the interpretable results of the prediction and inference results of the AI model, and then judging the rationality of the prediction and inference results of the AI model, so as to ensure the reliability of communication.

[0074] Next, specific embodiments will be combined to introduce the new communication method proposed in this application. By adapting the XAI method to the AI application in the mobile network, an interpretable analysis of the model in the interaction is given, so as to help the receiving node judge the reliability of the output of the inference node.

[0075] In the communication method provided by an embodiment of this application, the management node performs the inference of the AI task and the interpretable analysis of the model, and sends the inference result and the interpretable result to the receiving node, so that the receiving node can judge the credibility of the inference result and perform subsequent tasks. An exemplary flowchart of the communication method of this embodiment is as Figure 3As shown. The method may include S305, S310, S320, S330, and S340.

[0076] S305, the model management node sends second information to the receiving node, and the second information is used to indicate the interpretable relevant information associated with each output type in at least one output type. Accordingly, the receiving node receives the second information.

[0077] As an example, the output type may include at least one of the following types: the type of AI inference result that the model management node subsequently provides to the receiving node, for example, the prediction result or the execution measurement that the model management node subsequently provides to the receiving node; the specific data that the model management node needs to collect from the receiving node; the model management node instructs the receiving node to change the currently applied AI model policy; or, the black-box AI model.

[0078] As an example, the interpretable relevant information may include at least one of the following information: the interpretability method, which is used to indicate the method for performing interpretability analysis; the interpretation content of the interpretability method, which is used to indicate the interpretation content corresponding to different interpretability methods; the accuracy of the interpretability method, which is used to indicate the interpretation accuracy provided by different interpretability methods.

[0079] As an example, the output type may be included in the interpretable relevant information. In this case, it means that the interpretable relevant information is associated with the output type, or in other words, the other information included in the interpretable relevant information is associated with the output type.

[0080] The interpretable relevant information associated with the output type includes the interpretability method, which can be understood as: the interpretability method recommended or supported by the AI inference result of this output type, or in other words, the interpretability method that the model management node can support or recommend.

[0081] In some possible implementation manners, the second information may be understood as the model management node sending an inference type indication to the receiving node and requesting confirmation of the interpretability method. Or, the behavior of the management node sending the second information may be understood as the management node initiating an inference indication and an XAI confirmation.

[0082] As an example, the interpretability method may include but is not limited to the following methods: Lime, Shap, Anchors, etc.

[0083] As an example, the returned interpretation content corresponding to different interpretability methods is different. For example, the Shap method returns the importance of features, the Anchors method returns the smallest interpretable subset, and the Lime-based method returns a linear model, etc.

[0084] It should be noted that this application does not limit the specific interpretability method adopted and the corresponding interpretation content, and can be generally applicable to various innate or acquired interpretability methods.

[0085] As an example, the interpretation content of the interpretability method can be the input attribute feature value, the AI model, or the comparison of the interpretation logics before and after the model change, etc.

[0086] As an example, different interpretability methods can provide different interpretation accuracies and / or different numbers of feature attributes, etc. For example, different interpretability methods can provide the number of feature attributes corresponding to the inference result, and the accuracy of the provided feature values is different.

[0087] For example, the Sharp method can provide the number of features corresponding to the prediction result and whether it can provide the corresponding importance value and / or importance attribute value, etc.

[0088] The interpretability-related information can be represented in the form of index - content - interpretation method shown in the following table, and an index is established to indicate various category combinations.

[0089] Table 1 Composition and Corresponding Relationship of Model Interpretability Indication Information

[0090]

[0091] S310, the receiving node sends the first information to the model management node. Correspondingly, the model management node receives the first information. The first information indicates the interpretability-related information.

[0092] For the sake of convenient description, the interpretability-related information indicated by the first information can be called the first interpretability-related information.

[0093] In some possible implementation manners, the first information can be understood as the receiving node indicating the required interpretability-related information to the model management node. Or, the act of the receiving node sending the first information can be understood as the receiving node indicating the selected interpretability method.

[0094] In some possible implementation manners, when the receiving node sends the first information, the interpretability-related information indicated by the first information can be determined by the second information.

[0095] As an example, the interpretability-related information indicated by the first information can also be understood as the interpretability-related information supported or recommended by the model management node.

[0096] In some possible implementation manners, the receiving node may feedback the selected interpretability method to the model management node according to the interpretability method indicated by the model management node; or, the receiving node may also feedback its default interpretability method to the model management node without following the interpretability method indicated by the model management node.

[0097] As an example, when the receiving node specifies the interpretability-related information supported or recommended by the model management node, the first information may be one of multiple interpretability-related information, and the first information further includes the index of the first interpretability-related information among the multiple interpretability-related information.

[0098] In the steps of this embodiment, the interpretability-related information indicated by the first information may refer to the content indicated in Table 1, or may indicate the interpretability-related information that needs to be provided additionally in addition to Table 1.

[0099] As an example, the first information may include the interpretability method based on the index value and the parameter selection corresponding to different interpretability methods. For example, for local interpretability methods such as Shap, Lime, and Anchors, information such as the selected local sample points or sample sets needs to be indicated.

[0100] As an example, the first information may also directly feedback the first interpretability-related information without indicating the first interpretability-related information according to the index.

[0101] S320, the model management node performs AI model inference and conducts interpretability analysis on the inference result of the first AI model.

[0102] In some possible implementation manners, the inference result of the AI model may be understood as the output result obtained by the model management node using the AI model to train the prediction event.

[0103] As an example, for a general AI application case, such as using an AI model to perform inference and prediction on UE handover, the model management node may perform an optimal handover decision through the first AI model, adaptively adjust the optimal parameters of UE handover, etc., and feedback the inference result of the first AI model to the UE, that is, the receiving node. Finally, the UE performs a handover operation according to the inference result (or output result) of the AI model, which can reduce the handover delay of the UE and improve the handover success rate, etc.

[0104] It should be noted that the AI application interactions among multiple nodes involved in the communication system of this application include, but are not limited to, channel prediction, handover prediction, load balancing, AI prediction information interaction between base stations, etc. This application does not focus on describing the AI application scenarios in the communication system. Any communication system that can implement the function set of AI applications in the communication system should be included in the protection scope of this application.

[0105] In some possible implementations, the model management node completes the inference of the AI task according to the instructions returned by the receiving node, and performs model interpretability analysis based on the method agreed with the receiving node, or adds an interpretable explanation to the inference result of the model.

[0106] In some possible implementations, interpretability analysis can be understood as the interpretive analysis of the inference result of the AI model using the specified interpretability method.

[0107] S330. The model management node sends the inference result and the interpretable result of the first AI model to the receiving node. Correspondingly, the receiving node receives the inference result and the interpretable result of the first AI model.

[0108] In some possible implementations, the model management node performs interpretability analysis on the inference result of the first AI model to obtain an interpretable result.

[0109] In this application, the interpretable result of the inference result of the AI model can be the score of the interpretability test or the model obtained through the interpretability method, etc. The interpretable result of the inference result of the AI model in this application can be referred to as the interpretability of the AI model.

[0110] As an example, when the model management node is based on the agreed interpretability method Lime, the interpretable result is a locally approximated simple linear model, which is used to indicate that the features of some dimensions in the input data play a major role in the inference result; when the interpretability method is Shap, the interpretable result can be the feature weight values of one or more inputs for the inference result, which are used to indicate to the receiving node the influence of one or more input items on the inference result, so as to assist the receiving node in confirming whether the inference result is logical; when the interpretability method is Anchors, the interpretable result is a set of feature / if-then rule sets for interpreting the model.

[0111] In the embodiment of this application, after the receiving node receives the inference result of the AI model, it is also necessary to judge the credibility or reasonableness of the inference result, so as to ensure the reliability of applying the AI model to the communication network. Then the obtained inference result of the first AI model here can be understood as the relevant content information for subsequent judgment of its credibility and reasonableness.

[0112] In this embodiment, the model management node sending the inference result of the first AI model and sending the interpretable result can be sent on different messages respectively, or can be sent in the same message.

[0113] In the embodiment of this application, after the receiving node obtains the interpretable result of the inference result of the first AI model, optionally, it further includes S340.

[0114] S340, the receiving node sends the fifth piece of information to the model management node to indicate the reasonableness of the inference result of the first AI model. Correspondingly, the model management node receives the fifth piece of information.

[0115] As an example, the reasonableness of the model inference result can be understood as the receiving node accepting, recognizing, or relying on the inference result, and the unreasonableness can be understood as not accepting, not recognizing, or being unreliable.

[0116] In some possible implementation manners, when the receiving node sends the fifth piece of information to the model management node, it indicates that the inference result is reasonable, and not sending it indicates unreasonableness; or, sending the fifth piece of information indicates unreasonableness, and not sending the fifth piece of information indicates reasonableness.

[0117] As an example, the receiving node determines whether to receive the output value of the relevant AI model or the operation indicated by the model policy according to the reasonableness of the inference result.

[0118] In some possible implementation manners, if the receiving node receives the inference result, it indicates that the inference result is reasonable, then it will receive the output value of the relevant AI model or the operation indicated by the model policy, and further perform subsequent tasks.

[0119] In some possible implementation manners, if the receiving node does not receive the inference result, it indicates the reason.

[0120] As an example, the indicated reason can be not recognizing the current interpretable result. Taking the Lime method as an example, the receiving node can indicate not recognizing the accuracy of the simple linear model and thus not recognizing the inference result; taking the Shap method as an example, the receiving node can indicate not recognizing the input item or its weight value, that is, not recognizing the influence of a specific input item on the inference result and thus not recognizing the inference result; taking the Anchors method as an example, the receiving node can indicate not recognizing its rule set or the sample points covered by the rule set.

[0121] As an example, the indicated reason can be not recognizing the reasonableness of the inference result, that is, not thinking that the current interpretable result is correct, or requiring an interpretable result with better accuracy. For example, for the Lime method, a more accurate linear model is required, and for the Shap method, a feature analysis result with more input parameters is required. Or in a general method, it is indicated to add more sample points to participate in the interpretability analysis, etc.

[0122] This embodiment presents an interaction process for interpretability analysis performed by the model management node, i.e., the AI prediction and inference node. The prediction and inference node is also responsible for providing the output of the AI model and the interpretability analysis, enabling the receiving node to judge the reasonableness and reliability of the result based on the interpretability analysis while obtaining the output of the AI model, and flexibly selecting whether the inference result can be used for subsequent tasks. In addition, the prediction and inference node of the AI model provides the model inference result and performs interpretability analysis, reducing the overhead of intermediate signaling interaction and being able to better protect the privacy of the AI model.

[0123] Specifically, this embodiment does not limit the specific AI tasks and corresponding node instances. The corresponding interaction channel can select the connection channel of the corresponding network element. For example: Nx interface: The receiving node is a gNB, and the model management node is a core network side network element such as a computing management network element CMF, etc. The base station requests the AI model inference result or configures the scheduling policy from the computing network element of the core network; Xn interface: Both the receiving node and the model management node are gNBs, such as the source station sending the prediction result based on the AI model to the target station; F1 interface: The receiving node is a gNB-DU, and the model management node is a centralized unit gNB-CU, which is uniformly responsible for the management of the AI model; Uu interface: The receiving node is a terminal UE, and the model management node is a base station gNB. The terminal obtains the air interface information or configuration policy predicted by the AI model from the base station.

[0124] In the communication method provided by an embodiment of this application, the management node performs the inference of the AI task and sends the inference result and the input required for interpretability analysis to the receiving node. The receiving node performs the interpretability analysis by itself for subsequent tasks. An exemplary flowchart of the communication method of this embodiment is as Figure 4 shown. This method may include S410 to S460.

[0125] S410, the model management node sends the inference result of the first AI model to the receiving node. Correspondingly, the receiving node receives the inference result of the first AI model.

[0126] In some possible implementation manners, before the model management node sends the inference result, the inference process of the AI model has been completed.

[0127] In this embodiment, this step can refer to the process of obtaining and sending the inference result of the AI model in S320 and S330, which will not be elaborated here.

[0128] S420, the receiving node sends the first information to the model management node. Correspondingly, the model management node receives the first information. The first information indicates the first interpretability-related information.

[0129] As an example, the interpretability-related information indicated by the first information may be the interpretability-related information supported by the receiving node.

[0130] In the steps of this embodiment, the interpretable relevant information indicated by the first information can be determined with reference to the content indicated in Table 1.

[0131] In some possible implementation manners, the interpretable relevant information may further include the relevant inputs required for interpretable analysis, or rather, the input data required for interpretable analysis.

[0132] In some possible implementation manners, after receiving the inference result of the first AI model, the receiving node requests the model management to confirm the relevant inputs required for interpretability. The specific content may include, but is not limited to: the input quantities corresponding to the relevant policies, the inference results, etc.; the logical architecture of the model management node, etc.; the obtained parameters are used for the receiving node to perform interpretability judgment.

[0133] S430, the model management node sends the third information to the receiving node. Correspondingly, the receiving node receives the third information. The third information is used to indicate the information required to obtain the interpretable result of the inference result of the first AI model.

[0134] In some possible implementation manners, the receiving node may obtain the input data required for interpretable analysis or the input required for the interpretable method through the third information, and perform interpretable analysis on the input data to obtain an interpretable result.

[0135] In some possible implementation manners, the model management node may indicate the inputs required for the interpretable methods supported by itself and feedback them to the receiving node.

[0136] As an example, the model management node calculates the inputs required for the corresponding interpretable method based on the local AI model and feeds them back to the receiving node. For local interpretable methods such as Shap, Lime, and Anchors, information such as the selected local sample points / sets and the corresponding model outputs needs to be indicated.

[0137] S440, the receiving node performs interpretable analysis on the inference result of the first AI model.

[0138] It can be understood that the receiving node performs interpretable analysis on the inference result or the policy based on the obtained model inference result and the inputs required for the interpretable method provided by the model management node.

[0139] In the embodiments of this application, the interpretable analysis method is not limited, and it may be Shap, Anchors, Lime or other methods. The specific steps of interpretable analysis are the same as those described in S320.

[0140] In the embodiment of the present application, after the receiving node obtains an interpretable result through interpretable analysis based on the inference result output by the model management node and locally, optionally, S450 is further included.

[0141] S450, the receiving node sends fifth information to the model management node, which is used to indicate the rationality of the inference result of the first AI model. Correspondingly, the model management node receives the fifth information.

[0142] In this embodiment, this step can refer to step S340, which will not be elaborated here.

[0143] This embodiment provides an interaction process for interpretable analysis by the receiving node. The model management node provides auxiliary information for the receiving node to perform interpretable analysis by itself, enabling the receiving node to flexibly perform interpretable analysis locally. The auxiliary information provided by the model management node can be reused in subsequent other interpretable analysis methods.

[0144] It should be noted that this embodiment does not limit the specific AI task and the corresponding node instance. The corresponding interaction channel can select the connection channel of the corresponding network element, as specifically shown in the above embodiment, which will not be elaborated here.

[0145] In the communication method provided by an embodiment of the present application, the management node performs inference on the AI task, and sends the inference result and the input required for interpretable analysis to the third node. The third node performs interpretable analysis and sends the interpretable result to the receiving node, so as to facilitate the receiving node to perform subsequent tasks. An exemplary flowchart of the communication method of this embodiment is as Figure 5 shown. This method may include S510 to S550.

[0146] S510, the model management node sends the inference result of the first AI model to the receiving node and the third node. Correspondingly, the receiving node and the third node receive the inference result of the first AI model.

[0147] It can be understood that the model management node can send the inference result of the first AI model to both the receiving node and the third node.

[0148] In this embodiment, the third node is a network node in the communication system that can uniformly be responsible for the interpretable analysis of the AI model.

[0149] In this embodiment, this step can refer to S410, which will not be elaborated here.

[0150] S520, the third node sends the first information to the model management node. Correspondingly, the model management node receives the first information. The first information indicates the first interpretable relevant information.

[0151] In some possible implementation manners, the interpretable relevant information indicated by the first information may be the interpretable relevant information supported by the third node.

[0152] In some possible implementation manners, the first information may be understood as the third node indicating the interpretable relevant information to the model management node.

[0153] In this embodiment, the interpretable relevant information indicated by this step may refer to S420, which will not be elaborated here.

[0154] In some possible implementation manners, it may further include: the third node sends fourth information to the model management node, and the fourth information is used to request the information required for obtaining the interpretability result of the inference result of the first AI model.

[0155] Or it can be said that the third node requests the input data required for interpretability analysis through the fourth information. For example, when the third node determines the interpretability result by itself, it can send the fourth information.

[0156] S530, the model management node sends third information to the third node. Correspondingly, the third node receives the third information. The third information is used to indicate the information required for obtaining the interpretability result of the inference result of the first AI model.

[0157] In some possible implementation manners, the third information may be understood as the model management node indicating the input data required for interpretability analysis or the input required for the interpretability method to the third node.

[0158] In some possible implementation manners, the model management node may indicate the input required for the interpretable method supported by itself and feedback it to the third node.

[0159] As an example, the model management node calculates the input required for the corresponding interpretable method based on the local AI model and feeds it back to the third node. For local interpretable methods such as Shap, Lime, and Anchors, information such as the selected local sample points / sets and the corresponding model outputs need to be indicated.

[0160] S540, the third node performs interpretability analysis on the inference result of the first AI model.

[0161] In this embodiment, this step may refer to step S440, which will not be elaborated here.

[0162] S550, the third node sends the interpretability result to the receiving node. Correspondingly, the receiving node receives the interpretability result.

[0163] In this embodiment, the third node determines whether to receive relevant output values or policy instructions based on the output result of the model management node and the local interpretability analysis result, and feeds back the interpretability analysis result to the receiving node.

[0164] S560. The receiving node sends the fifth piece of information to the model management node to indicate the reasonableness of the inference result of the first AI model. Correspondingly, the model management node receives the fifth piece of information.

[0165] In some possible implementation manners, when the receiving node sends the fifth piece of information to the model management node, it indicates that the inference result is reasonable, and not sending it indicates that it is unreasonable; or, sending the fifth piece of information indicates unreasonableness, and not sending the fifth piece of information indicates reasonableness.

[0166] In some possible implementation manners, the third node can directly send the indication information to the model management node, or the receiving node can send the indication information to the model management node.

[0167] In this embodiment, this step can refer to step S340 and will not be elaborated here.

[0168] This embodiment provides an interaction process for introducing a third management node for interpretability analysis. The model management node provides auxiliary information for the third node to perform interpretability analysis uniformly, making the deployment and management of mobile network interpretability analysis more centralized and convenient to modify, and not adding additional computing load to the original nodes for analyzing model interpretability.

[0169] It should be noted that this embodiment does not limit the specific AI tasks and corresponding node instances. The corresponding interaction channels can select the connection channels of the corresponding network elements. The specific model management node - receiving node - third management node instances can be: gNB - UE / gNB - model management network element, gNB(DU) - UE - gNB(CU), etc.

[0170] The embodiments of this application provide an AI application in a communication network with a model interpretability analysis and interaction solution, enabling the receiving node to have a judgment ability based on the credibility of the AI inference model. In addition, the differences between the three embodiments of this application lie in the different deployment locations of the three types of interpretability analysis, such as being deployed in the model management node, the receiving node, or the third - party management node.

[0171] Figure 6 It is a schematic structural diagram of a communication device according to an embodiment of this application. As Figure 6 shown, the device 600 may include a processing module 601 and a communication module 602.

[0172] As a first example, the device 600 can be used to implement Figures 3 to 5The communication method implemented by the model management node in any of the embodiments shown. For example, the processing module 601 is used to implement Figures 3 to 5 The steps related to the processing executed by the model management node in any of the embodiments shown, and the communication module 602 is used to implement Figures 3 to 5 The steps such as sending and / or receiving executed by the model management node in any of the embodiments shown.

[0173] As a second example, the device 600 can be used to implement Figures 3 to 5 The communication method implemented by the receiving node in any of the embodiments shown. For example, the processing module 601 is used to implement Figures 3 to 5 The steps related to the processing executed by the receiving node in any of the embodiments shown, and the communication module 602 is used to implement Figures 3 to 5 The steps such as sending and / or receiving executed by the receiving node in any of the embodiments shown.

[0174] As a third example, the device 600 can be used to implement Figures 3 to 5 The communication method implemented by the third node in any of the embodiments shown. For example, the processing module 601 is used to implement Figures 3 to 5 The steps related to the processing executed by the third node in any of the embodiments shown, and the communication module 602 is used to implement Figures 3 to 5 The steps such as sending and / or receiving executed by the third node in any of the embodiments shown.

[0175] Figure 7 is a schematic structural diagram of a communication device applicable to an embodiment of the present application. As Figure 7 shown, the communication device 700 includes a processor 711, a memory 712, and a transceiver 713. The transceiver 713 includes a transmitter 7131, a receiver 7132, and an antenna 7133.

[0176] Among them, the processor 711, the memory 712, and the transceiver 713 communicate with each other through an internal connection path.

[0177] The receiver 7132 can be used to receive transmission control information through the antenna 7133, and the transmitter 7131 can be used to send transmission feedback information to other communication nodes through the antenna 7133.

[0178] Optionally, the device 700 may further include a memory 712 for the instructions executed by the processor 711 or the input data required for the processor 711 to run the instructions or the data generated after the processor 711 runs the instructions.

[0179] As an example, the processor 711 is used to implement the functions of the above-mentioned processing module 601, and the transceiver 713 is used to implement the functions of the above-mentioned communication module 602.

[0180] Figure 8 The structural schematic diagram of the communication device provided by another embodiment of the present application. As Figure 8 shown, the device 800 includes a processor 801 and a communication circuit 802. The processor 801 and the communication circuit 802 are coupled to each other. It can be understood that the communication circuit 802 can be a transceiver or an input / output interface. Optionally, the device 800 may further include a memory 803, which is used to store the instructions executed by the processor 801 or the input data required for the processor 801 to run the instructions or the data generated after the processor 801 runs the instructions. It can be understood that the memory 803 can be located outside the processor 801, or inside the processor 801.

[0181] As an example, the processor 801 is used to implement the functions of the above-mentioned processing module 601, and the communication circuit 802 is used to implement the functions of the above-mentioned communication module 602.

[0182] The device 800 can be a communication device or a chip applied to a communication device. For example, the device 800 can be a communication node or a chip applied in a communication node. It can be understood that when the device 800 is a communication node, the communication circuit 802 can be a transceiver.

[0183] In some embodiments of the present application, a computer program product is further provided. When the computer program product runs on a processor, it can implement the method implemented by the communication node in any of the above embodiments.

[0184] In some embodiments of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium contains computer instructions, and when the computer instructions run on a processor, they can implement the method implemented by the communication node in any of the above embodiments.

[0185] In some embodiments of the present application, a communication system is further provided, and the system can implement the method implemented by the communication node in any of the above embodiments.

[0186] It can be understood that the processor in the embodiments of the present application may be the following devices or all or part of the circuits for processing functions in the following devices: a central processing unit (CPU), and may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0187] The method steps in the embodiments of the present application may be implemented in a hardware manner or by a processor executing software instructions. The software instructions may be composed of corresponding software modules, and the software modules may be stored in a random access memory, flash memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, registers, hard disks, removable hard disks, CD-ROMs, or any other form of storage medium well known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium may also be a component of the processor. The processor and the storage medium may be located in an ASIC. In addition, the ASIC may be located in a network device or a terminal device. Of course, the processor and the storage medium may also exist as discrete components in a network device or a terminal device.

[0188] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, or other programmable devices. The computer program or instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless manner. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it can also be an optical medium, such as a digital video disc; or it can be a semiconductor medium, such as a solid-state drive.

[0189] In various embodiments of the present application, if there is no special description and logical conflict, the terms and / or descriptions between different embodiments are consistent and can be cross-referenced. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0190] It can be understood that the various numerical numbers involved in the embodiments of the present application are only for the convenience of description and do not limit the scope of the embodiments of the present application. The magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic.

Claims

1. A communication method, It is characterized in that Applied to a communication node, the method comprises: Obtaining the inference result of the first artificial intelligence AI model; Obtain an interpretable result of the inference result.

2. The method according to claim 1, It is characterized in that The method further comprises: Send first information, wherein the first information indicates first explainable related information, wherein the first explainable related information is explainable related information, and the explainable related information includes at least one of the following information: an explainability method, an output type, an explanation content of the explainability method, or an explanation accuracy of the explainability method, and the output type is the type of reasoning result of the AI ​​model.

3. The method according to claim 2, It is characterized in that The first information is one of a plurality of interpretable related information, wherein the first information includes an index of the first interpretable related information in the plurality of interpretable related information.

4. The method according to any one of claims 1 to 3, It is characterized in that The method further comprises: Second information is received, wherein the second information indicates interpretable related information associated with each output type of at least one output type.

5. The method according to any one of claims 1 to 4, It is characterized in that The method further comprises: receiving third information indicating information required to obtain an interpretable result of the inference result of the first AI model; Wherein, obtaining an interpretable result of the inference result includes: The explainable result is determined according to the third information.

6. The method according to claim 5, It is characterized in that The method further comprises: Send fourth information, where the fourth information is used to request information required to obtain an explainable result of the reasoning result of the first AI model.

7. The method according to any one of claims 1 to 6, It is characterized in that The method further comprises: The interpretable result is transmitted.

8. The method according to any one of claims 1 to 4, It is characterized in that The obtaining of an interpretable result of the inferred result includes: The interpretable result is received.

9. The method according to any one of claims 1 to 8, It is characterized in that The method further comprises: Send fifth information, where the fifth information indicates the rationality of the reasoning result of the first AI model.

10. A communication device, It is characterized in that The method comprises a functional module for implementing the method according to any one of claims 1 to 9.

11. A communication device, It is characterized in that include: Memory and processor; The memory is used to store program instructions; The processor is configured to execute program instructions in the memory to implement the method according to any one of claims 1 to 9.

12. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a program code for computer execution, wherein the program code includes instructions for implementing the method according to any one of claims 1 to 9.

13. A computer program product, It is characterized in that The computer program product comprises instructions for implementing the communication method according to any one of claims 1 to 9.

Images