Data processing method and apparatus

Abstract

Disclosed in the present application are a data processing method and apparatus. The method relates to the technical field of computers. The method comprises: inputting a node input feature matrix into a first model, so as to obtain a node output feature matrix, wherein parameters in the first model comprise the correlation between two of a plurality of nodes, the association relationship between two of the plurality of nodes is determined by means of comparing the correlation between the two nodes and a first threshold value, and the first threshold value is an element, which corresponds to the two nodes, in a first threshold value vector, the first threshold value vector being obtained on the basis of a first function used by the first model, and the first function being differentiable. On the basis of the method, the association relationship between various nodes is determined by means of a first threshold value vector, such that the process of determining the adjacency relation between the nodes is more intelligent and accurate, which is conducive to improving the performance of a model.

Description

Data processing method and apparatus

[0001] This application claims priority to Chinese Patent Application No. 202410627114.2, filed on May 17, 2024, entitled “Data Processing Method and Apparatus”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of computer technology, and in particular to data processing methods and apparatus. Background Technology

[0003] Machine learning (ML) is an important technological approach to achieving artificial intelligence (AI). ML can be used to enhance positioning accuracy by using the responses of several channels as input to an AI model to obtain the final location of the terminal device. Graph Neural Network (GNN) models have attracted widespread attention from researchers due to their excellent learning ability to aggregate neighboring information of nodes, and are therefore frequently used in AI positioning.

[0004] The input data structure of a GNN model is a graph. In mathematical abstraction, a graph is a collection of nodes connected by edges. A graph can be represented in computation using a node feature matrix and an adjacency matrix. The node feature matrix characterizes the initial features of each node in the graph, while the adjacency matrix characterizes the graph's topological structure, i.e., the connections between the nodes. The adjacency matrix is ​​determined based on the correlation between nodes. For example, if the correlation between two nodes is greater than a preset threshold, they are considered connected, neighbors, and their corresponding elements in the adjacency matrix are set to 1. If the correlation is less than the preset threshold, they are considered disconnected, and their corresponding elements in the adjacency matrix are set to 0. Currently, this preset threshold is manually set, which is not precise enough to accurately reflect the relationships between nodes, thus affecting model performance. Summary of the Invention

[0005] This application provides a data processing method and apparatus that uses a trainable threshold vector to determine the association between nodes, thereby making the adjacency determination process of each node more intelligent and accurate, which helps to improve the performance of the model.

[0006] In a first aspect, this application provides a data processing method, which includes: inputting a node input feature matrix into a first model to obtain a node output feature matrix, wherein the node input feature matrix includes multiple elements representing the input features of multiple nodes, and the node output feature matrix includes multiple elements representing the output features of multiple nodes; wherein the parameters in the first model include the correlation between two nodes among the multiple nodes; when the correlation between two nodes among the multiple nodes is greater than a first threshold, the two nodes are correlated; when the correlation between two nodes among the multiple nodes is less than or equal to the first threshold, the two nodes are not correlated; the first threshold is the element corresponding to two nodes in a first threshold vector, and the first threshold vector is obtained based on a first function used by the first model, and the first function is differentiable.

[0007] Based on the method described in the first aspect, since the first function is differentiable, the first threshold vector can be trained using the first function through the first model. The first threshold vector used to determine the association between nodes is a trainable parameter in the first model, and each element in the first threshold vector represents the threshold required to determine the association between its corresponding specific two nodes. The first threshold vector can also be updated and optimized through a backpropagation process. Therefore, the first threshold vector makes the process of determining whether there is a correlation between nodes more intelligent and accurate, which is beneficial to improving the performance of the model.

[0008] In this application, the two nodes described can also be understood as a pair of nodes. For example, the explanation can be provided by replacing two nodes with a pair of nodes: the parameters in the first model include the correlation of multiple pairs of nodes. Each pair of nodes in these multiple pairs includes two nodes, and both nodes in each pair belong to the multiple pairs of nodes. The first node pair is one of these multiple pairs of nodes. When the correlation of the first node pair is greater than a first threshold, the two nodes in the first node pair are correlated; when the correlation of the first node pair is less than or equal to the first threshold, the two nodes in the first node pair are not correlated. The first threshold is the element corresponding to the first node pair in the first threshold vector.

[0009] In one possible implementation, the first function satisfies:

[0010] in, is the output parameter of the first function, x is the input parameter of the first function, ReLU function is the linear rectified function, tanh function is the hyperbolic tangent function, and γ is a preset coefficient.

[0011] In one possible implementation, the first function satisfies:

[0012] in, Let be the output parameter of the first function, x be the input parameter of the first function, the ReLU function be a linear rectified function, and a, b, and γ be preset coefficients. The sigmoid function satisfies...

[0013] In one possible implementation, the first function satisfies:

[0014] in, Let γ be the output parameter of the first function, x be the input parameter of the first function, ReLU be the linear rectified function, γ be the preset coefficient, and softsign be the function that satisfies γ = γ * γ * γ.

[0015] In one possible implementation, the value of each element in the first threshold vector is greater than or equal to 0 and less than or equal to the maximum distance between any two nodes.

[0016] In one possible implementation, the value of each element in the first threshold vector is greater than or equal to 0 and less than or equal to 1.

[0017] In one possible implementation, the first model is used for node localization, and the node's output features are used to indicate the node's location information. This implementation method is beneficial for improving the accuracy of node localization.

[0018] Secondly, embodiments of this application provide an apparatus for performing the method in any aspect or any possible implementation of the first aspect. The apparatus includes modules for performing the method in any aspect or any possible implementation of the first aspect. These modules can be implemented in software, in hardware, or a combination of software and hardware.

[0019] Thirdly, embodiments of this application provide an apparatus including a processing circuit for executing the method in any possible implementation of the first aspect. The processing circuit executes a program stored in a memory, and when the program is executed, the method shown in any possible implementation of the first aspect is performed.

[0020] In one possible implementation, the memory is located outside the aforementioned device.

[0021] In one possible implementation, the memory is located within the aforementioned device.

[0022] In this application, the processing circuit and the memory can also be integrated into a single device, that is, the processing circuit and the memory can be integrated together.

[0023] In one possible implementation, the above-described apparatus further includes a transceiver circuit for receiving information (or input information) or sending information (or output information).

[0024] Fourthly, embodiments of this application provide an apparatus including a processing circuit and a transceiver circuit. The processing circuit can be a logic circuit, and the transceiver circuit can be an interface circuit. The logic circuit and the interface are coupled. The interface circuit is used to input and / or output information, and the logic circuit is used to execute the method in any possible implementation of the first aspect.

[0025] Fifthly, embodiments of this application provide a computer-readable storage medium for storing a computer program that, when run on a computer, causes the method shown in any possible implementation of the first aspect to be executed.

[0026] Sixthly, embodiments of this application provide a computer program product that, when run on a computer, causes the methods shown in any possible implementation of the first aspect to be executed. Attached Figure Description

[0027] Figure 1 is a schematic diagram of a possible application framework in the communication system provided in an embodiment of this application;

[0028] Figure 2 is a schematic diagram of a possible application framework in the communication system provided in an embodiment of this application;

[0029] Figure 3A is a schematic diagram of a communication system applicable to the communication method of the embodiments of this application;

[0030] Figure 3B is a schematic diagram of another communication system applicable to the communication method of the present application embodiment;

[0031] Figure 4 is a schematic diagram of an AI application framework in NR provided by an embodiment of this application;

[0032] Figure 5 is a schematic diagram of a neuron structure provided in an embodiment of this application;

[0033] Figure 6 is a schematic diagram of a neural network structure provided in an embodiment of this application;

[0034] Figure 7A is a schematic diagram of an AI-based uplink positioning provided in an embodiment of this application;

[0035] Figure 7B is a schematic diagram of another AI-based uplink positioning provided in an embodiment of this application;

[0036] Figure 8 is a flowchart illustrating a data processing method provided in an embodiment of this application;

[0037] Figure 9 is a schematic diagram of a step function provided in an embodiment of this application;

[0038] Figure 10 is a schematic diagram of the first function provided in an embodiment of this application;

[0039] Figure 11 is a schematic diagram of the structure of a device provided in an embodiment of this application;

[0040] Figure 12 is a schematic diagram of the structure of a device provided in an embodiment of this application;

[0041] Figure 13 is a schematic diagram of the structure of a device provided in an embodiment of this application. Detailed Implementation

[0042] The technical solutions in this application will now be described with reference to the accompanying drawings.

[0043] The technical solutions provided in this application can be applied to various communication systems, such as: 5th generation (5G) or new radio (NR) systems, long term evolution (LTE) systems, LTE frequency division duplex (FDD) systems, LTE time division duplex (TDD) systems, wireless local area network (WLAN) systems, satellite communication systems, future communication systems such as 6th generation (6G) mobile communication systems, or integrated systems of multiple systems. The technical solutions provided in 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 Internet of Things (IoT) communication systems or other communication systems.

[0044] In a communication system, one network element can send signals to or receive signals from another network element. These signals can include information, signaling, or data. The term "network element" can also be replaced by an entity, network entity, device, communication equipment, communication module, node, communication node, etc. This disclosure uses a network element as an example. For instance, a communication system can include at least one terminal device and at least one network device. The network device can send downlink signals to the terminal device, and / or the terminal device can send uplink signals to the network device. It is understood that the terminal device in this disclosure can be replaced by a first network element, and the network device can be replaced by a second network element, both performing the corresponding communication methods described in this disclosure.

[0045] In the embodiments of this application, the terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent, or user apparatus.

[0046] Terminal devices can be devices that provide voice / data, such as handheld devices with wireless connectivity, in-vehicle devices, etc. Currently, examples of terminals include: mobile phones, tablets, laptops, PDAs, mobile internet devices (MIDs), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving vehicles, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, cellular phones, cordless phones, session initiation protocol (SIP) phones, wireless local loop (WLL) stations, personal digital assistants (PDAs), handheld devices with wireless communication capabilities, computing devices or other processing devices connected to wireless modems, wearable devices, terminal devices in 5G networks, or future public land mobile communication networks. Terminal devices in a network (PLMN), etc., are not limited to this in the embodiments of this application.

[0047] By way of example and not limitation, in this embodiment, the terminal device can also be a wearable device. Wearable devices, also known as wearable smart devices, are a general term for devices that utilize wearable technology to intelligently design and develop everyday wearables, such as glasses, gloves, watches, clothing, and shoes. Wearable devices are portable devices that are worn directly on the body or integrated into the user's clothing or accessories. Wearable devices are not merely hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are feature-rich, large in size, and can achieve complete or partial functions without relying on a smartphone, such as smartwatches or smart glasses, as well as those that focus on a specific type of application function and require the use of other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0048] In this embodiment, the device for implementing the functions of the terminal device can be the terminal device itself, or it can be any device capable of supporting the terminal device in implementing those functions, such as a chip system. This device can be installed in or used in conjunction with the terminal device. In this embodiment, the chip system can be composed of chips or may include chips and other discrete components. This embodiment only uses the terminal device as an example to illustrate the device for implementing the functions of the terminal device, and does not constitute a limitation on the solution of this embodiment.

[0049] The network device in this application embodiment can be a device for communicating with a terminal device. This network device can also be called an access network device or a wireless access network device, such as a base station. In this application embodiment, the network device can refer to a radio access network (RAN) node (or device) that connects the terminal device to the wireless network. A base station can broadly encompass, or be replaced by, various names including: NodeB, evolved NodeB (eNB), next-generation NodeB (gNB), relay station, access point, transmitting and receiving point (TRP), transmitting point (TP), master station, auxiliary station, motor slide retainer (MSR) node, home base station, network controller, access node, wireless node, access point (AP), transmission node, transceiver node, baseband unit (BBU), remote radio unit (RRU), active antenna unit (AAU), remote radio head (RRH), central unit (CU), distributed unit (DU), radio unit (RU), positioning node, etc. A base station can be a macro base station, micro base station, relay node, donor node, or similar entities, or combinations thereof. A base station can also refer to a communication module, modem, or chip installed within the aforementioned equipment or apparatus. A base station can also be a mobile switching center, equipment performing base station functions in D2D, V2X, and M2M communications, network-side equipment in 6G networks, and equipment performing base station functions in future communication systems. A base station can support networks using the same or different access technologies. Optionally, a RAN node can also be a server, wearable device, vehicle, or in-vehicle equipment. For example, the access network equipment in vehicle-to-everything (V2X) technology can be a roadside unit (RSU). The embodiments of this application do not limit the specific technologies or equipment forms used in the network equipment.

[0050] Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move depending on the location of the mobile base station. In other examples, a helicopter or drone can be configured as a device to communicate with another base station.

[0051] In some deployments, the network devices mentioned in the embodiments of this application may be devices including CU, DU, or CU and DU, or devices with control plane CU nodes (central unit-control plane (CU-CP)) and user plane CU nodes (central unit-user plane (CU-UP)) and DU nodes. For example, the network devices may include gNB-CU-CP, gNB-CU-UP, and gNB-DU.

[0052] In some deployments, multiple RAN nodes collaborate to assist terminals in achieving wireless access, with different RAN nodes each implementing some of the base station's functions. For example, RAN nodes can be CUs, DUs, CU-CPs, CU-UPs, or RUs. CUs and DUs can be configured separately or included in the same network element, such as a BBU. RUs can be included in radio frequency equipment or radio frequency units, such as RRUs, AAUs, or RRHs.

[0053] RAN nodes can support one or more types of fronthaul interfaces, each corresponding to a DU and RU with different functions. If the fronthaul interface between the DU and RU is a common public radio interface (CPRI), the DU is configured to implement one or more baseband functions, and the RU is configured to implement one or more radio frequency functions. If the fronthaul interface between the DU and RU is another type of interface, relative to CPRI, some downlink and / or uplink baseband functions, such as, for downlink, precoding, digital beamforming (BF), or one or more of inverse fast Fourier transform (IFFT) / cyclic prefix addition (CP), are moved from the DU to the RU; and for uplink, digital beamforming (BF), or one or more of fast Fourier transform (FFT) / cyclic prefix removal (CP), are moved from the DU to the RU. In one possible implementation, the interface can be an enhanced common public radio interface (eCPRI). Under the eCPRI architecture, the segmentation between DU and RU differs, corresponding to different categories (Cat) of eCPRI, such as eCPRI Cat A, B, C, D, E, F.

[0054] Taking eCPRI Cat A as an example, for downlink transmission, the DU is configured to implement one or more functions before and after layer mapping (i.e., rate matching, scrambling, modulation, and layer mapping), while other functions after layer mapping (e.g., resource element (RE) mapping, digital beamforming (BF), or one or more functions of inverse fast Fourier transform (IFFT) / adding cyclic prefix (CP)) are moved to the RU. For uplink transmission, the DU is configured to implement one or more functions before and after de-mapping (i.e., decoding, rate matching de-matching, descrambling, demodulation, inverse discrete Fourier transform (IDFT), channel equalization, and de-RE mapping), while other functions after de-mapping (e.g., digital BF or one or more functions of fast Fourier transform (FFT) / removing CP) are moved to the RU. It is understandable that the functional descriptions of the DU and RU corresponding to various types of eCPRI can be found in the eCPRI protocol, and will not be elaborated here.

[0055] In one possible design, the processing unit in the BBU used to implement baseband functions is called the baseband high (BBH) unit, and the processing unit in the RRU / AAU / RRH used to implement baseband functions is called the baseband low (BBL) unit.

[0056] In different systems, CU (or CU-CP and CU-UP), DU, or RU may have different names, but those skilled in the art will understand their meaning. For example, in an ORAN system, CU can also be called O-CU (open CU), DU can also be called O-DU, CU-CP can also be called O-CU-CP, CU-UP can also be called O-CU-UP, and RU can also be called O-RU. Any of the units among 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.

[0057] In this embodiment, the apparatus for implementing the functions of a network device can be a network device itself; it can also be an apparatus capable of supporting the network device in implementing those functions, such as a chip system, hardware circuit, software module, or a hardware circuit plus a software module. This apparatus can be installed in the network device or used in conjunction with the network device. In this embodiment, the example of a network device being used to implement the functions of a network device is provided only and does not constitute a limitation on the solutions described in this embodiment.

[0058] Network devices and / or terminal devices can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; they can also be deployed on water; and they can also be deployed in the air on airplanes, balloons, and satellites. This application does not limit the scenario in which the network devices and terminal devices are located. Furthermore, terminal devices and network devices can be hardware devices, or software functions running on dedicated hardware or general-purpose hardware, such as virtualization functions instantiated on a platform (e.g., a cloud platform), or entities that include dedicated or general-purpose hardware devices and software functions. This application does not limit the specific form of the terminal devices and network devices.

[0059] To support AI technology in wireless networks, AI nodes may also be introduced into the network.

[0060] Optionally, the AI ​​node can be deployed in one or more of the following locations within the communication system: access network devices, terminal devices, or core network devices, etc. Alternatively, the AI ​​node can be deployed independently, for example, in a location other than any of the aforementioned devices, such as in the host or cloud server of an over-the-top (OTT) system. The AI ​​node can communicate with other devices in the communication system, which can be one or more of the following: network devices, terminal devices, or core network elements, etc.

[0061] It is understood that this application does not limit the number of AI nodes. For example, when there are multiple AI nodes, they can be divided based on function, such as different AI nodes being responsible for different functions.

[0062] It can also be understood that AI nodes can be independent devices, or they can be integrated into the same device to achieve different functions. Alternatively, they can be network elements in hardware devices, software functions running on dedicated hardware, or virtualization functions instantiated on a platform (e.g., a cloud platform). This application does not limit the specific form of the aforementioned AI nodes.

[0063] AI nodes can be AI network elements or AI modules.

[0064] Figure 1 illustrates a possible application framework in a communication system. As shown in Figure 1, network elements in the communication system are connected via interfaces (e.g., NG, Xn) or air interfaces. These network element nodes, such as core network equipment, access network nodes (RAN nodes), terminals, or one or more devices in OAM, are equipped with one or more AI modules (only one is shown in Figure 1 for clarity). The access network node can be a single RAN node or can include multiple RAN nodes, for example, including CU and DU. The CU and / or DU can also be equipped with one or more AI modules. Optionally, the CU can be further divided into CU-CP and CU-UP. One or more AI models are configured in CU-CP and / or CU-UP.

[0065] The AI ​​module is used to implement corresponding AI functions. AI modules deployed in different network elements can be the same or different. Depending on the parameter configuration, the AI ​​module can implement different functions. The AI ​​module model can be configured based on one or more of the following parameters: structural parameters (e.g., at least one of the following: number of neural network layers, neural network width, inter-layer connections, neuron weights, neuron activation function, or bias in the activation function), input parameters (e.g., type and / or dimension of input parameters), or output parameters (e.g., type and / or dimension of output parameters). The bias in the activation function can also be referred to as the neural network bias.

[0066] An AI module can have one or more models. A model can infer an output, which includes one or more parameters. The learning, training, or inference processes of different models can be deployed on different nodes or devices, or they can be deployed on the same node or device.

[0067] Figure 2 illustrates a possible application framework in a communication system. As shown in Figure 2, the communication system includes a RAN intelligent controller (RIC). For example, the RIC can be the AI ​​module shown in Figure 1, used to implement AI-related functions. The RIC includes near-real-time RICs (near-RT RICs) and non-real-time RICs (non-RT RICs). Non-real-time RICs primarily process non-real-time information, such as data that is not sensitive to latency, with latency in the order of seconds. Real-time RICs primarily process near-real-time information, such as data that is relatively sensitive to latency, with latency in the order of tens of milliseconds.

[0068] The near real-time RIC is used for model training and inference. For example, it can be used to train an AI model and then use that AI model for inference. The near real-time RIC can obtain network-side and / or terminal-side information from RAN nodes (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminals. This information can be used as training data or inference data. Optionally, the near real-time RIC can deliver inference results to RAN nodes and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU. For example, the near real-time RIC delivers the inference result to the DU, and the DU sends it to the RU.

[0069] The non-real-time RIC is also used for model training and inference. For example, it can be used to train an AI model and then use that model for inference. The non-real-time RIC can obtain network-side and / or terminal-side information from RAN nodes (e.g., CU, CU-CP, CU-UP, DU, and / or RU) and / or terminals. This information can be used as training data or inference data, and the inference results can be delivered to the RAN nodes and / or terminals. Optionally, inference results can be exchanged between CU and DU, and / or between DU and RU. For example, the non-real-time RIC delivers the inference results to the DU, which then forwards them to the RU.

[0070] The near real-time RIC and non-real-time RIC can also be set up as separate network elements. Optionally, the near real-time RIC and non-real-time RIC can also be part of other devices. For example, the near real-time RIC can be set in the RAN node (e.g., in CU, DU), while the non-real-time RIC can be set in the OAM, cloud server, core network device, or other network device.

[0071] Figure 3A is a schematic diagram of a communication system applicable to the communication method of this application embodiment. As shown in Figure 3A, the communication system 100 may include at least one network device, such as network device 110 shown in Figure 3A; the communication system 100 may also include at least one terminal device, such as terminal device 120 and terminal device 130 shown in Figure 3A. Network device 110 and terminal devices (such as terminal device 120 and terminal device 130) can communicate via a wireless link. The communication devices in this communication system, for example, network device 110 and terminal device 120, can communicate via multi-antenna technology.

[0072] Figure 3B is a schematic diagram of another communication system applicable to the communication method of this application embodiment. Compared with the communication system 100 shown in Figure 3A, the communication system 200 shown in Figure 3B further includes an AI network element 140. The AI ​​network element 140 is used to perform AI-related operations, such as building training datasets or training AI models.

[0073] In one possible implementation, network device 110 can send data related to the training of the AI ​​model to AI network element 140, which then constructs a training dataset and trains the AI ​​model. For example, the data related to the training of the AI ​​model may include data reported by the terminal device. AI network element 140 can send the results of operations related to the AI ​​model to network device 110, which then forwards them to the terminal device. For example, the results of operations related to the AI ​​model may include at least one of the following: a trained AI model, model evaluation results, or test results. Exemplarily, a portion of the trained AI model may be deployed on network device 110, and another portion on the terminal device. Alternatively, the trained AI model may be deployed on network device 110. Or, the trained AI model may be deployed on the terminal device.

[0074] It should be understood that Figure 3B is only used as an example of the AI ​​network element 140 being directly connected to the network device 110. In other scenarios, the AI ​​network element 140 can also be connected to a terminal device. Alternatively, the AI ​​network element 140 can be connected to both the network device 110 and a terminal device simultaneously. Alternatively, the AI ​​network element 140 can also be connected to the network device 110 through a third-party network element. This application embodiment does not limit the connection relationship between the AI ​​network element and other network elements.

[0075] AI element 140 can also be set as a module in network devices and / or terminal devices, for example, in network device 110 or terminal device shown in Figure 3A.

[0076] It should be noted that Figures 3A and 3B are simplified schematic diagrams for ease of understanding. For example, the communication system may also include other devices, such as wireless relay devices and / or wireless backhaul devices, which are not shown in Figures 3A and 3B. In practical applications, the communication system may include multiple network devices or multiple terminal devices. This application does not limit the number of network devices and terminal devices included in the communication system.

[0077] To facilitate understanding of the solutions in the embodiments of this application, the terms that may be involved in the embodiments of this application are explained below.

[0078] I. Artificial Intelligence (AI)

[0079] Figure 4 illustrates the AI ​​application framework in Network Node (NR). The data source stores data inputs from network devices, CUs, DUs, terminal devices, or other nodes. It serves as a database for AI model training and data analysis inference; for example, the Data Collection module in Figure 4 represents the data source. The Model Training module analyzes the training data provided by the Data Collection to produce the optimal AI model. The Model Inference module uses the AI ​​model and, based on the inference data provided by the Data Collection, provides reasonable AI-based predictions (i.e., prediction data) for network operation, or guides the network to make policy adjustments. These policy adjustments are planned uniformly by the Actor entities and sent to multiple network entities for execution. Furthermore, after the network applies the relevant policies, its specific performance is again input into the data source for storage.

[0080] An AI model is an algorithm or computer program that enables AI functionality. It represents the mapping relationship between the model's input and output. Types of AI models include neural networks, linear regression models, decision tree models, support vector machines (SVMs), Bayesian networks, Q-learning models, and other machine learning (ML) models.

[0081] II. Deep Neural Networks (DNN)

[0082] Deep Neural Networks (DNNs) are a specific implementation of machine learning. According to the general approximation theorem, neural networks can theoretically approximate any continuous function, thus enabling them to learn arbitrary mappings. Traditional communication systems rely on extensive expert knowledge to design communication modules, while DNN-based deep learning communication systems can automatically discover hidden pattern structures from large datasets, establish mapping relationships between data, and achieve performance superior to traditional modeling methods.

[0083] The idea behind DNNs originates from the neuronal structure of the brain. Each neuron performs a weighted summation of its input values, and the result is passed through a non-linear function to generate the output, as shown in Figure 5. Specifically, assume the input to the neuron is x = [x0,...,x...]. n The weights corresponding to the inputs are w = [w0,...,w...]. nThe bias of the weighted summation is b. The form of the nonlinear function can be diverse. For example, the function f is the maximum value function of max{0,x}. The effect of the neuron's execution can be... DNNs typically have a multi-layered structure, with each layer containing multiple neurons. The input layer processes the received values ​​through neurons and then passes them to the hidden layers. Similarly, the hidden layers then pass the calculation results to the final output layer, producing the final output of the DNN, as shown in Figure 6.

[0084] DNNs typically have at least one hidden layer, which directly impacts their ability to extract information and fit functions. Increasing the number of hidden layers or widening each layer can improve the DNN's function fitting ability. The weights in each neuron constitute the parameters of the DNN network model. These parameters are optimized through training, enabling the DNN network to extract data features and express mapping relationships. DNNs generally use supervised or unsupervised learning strategies to optimize their model parameters.

[0085] III. Graph Convolutional Network (GCN)

[0086] GCN refers to a general term for algorithms that use neural networks to learn graph-structured data, extract and discover features and patterns in graph-structured data, and meet the needs of graph learning tasks such as clustering, classification, prediction, segmentation, and generation. The input data structure of the GNN model is a graph. In the abstract definition of mathematics, a graph is a collection of nodes connected to each other by edges. In computation, a graph can be represented by a node feature matrix and an adjacency matrix.

[0087] The node feature matrix characterizes the initial features of each node in the graph, while the adjacency matrix characterizes the graph's topology, i.e., the connections between the nodes. The adjacency matrix is ​​determined based on the correlation between the nodes. For example, if the correlation between two nodes is greater than a preset threshold, they are considered connected, neighbors, and their corresponding elements in the adjacency matrix are set to 1. If the correlation between two nodes is less than the preset threshold, they are considered disconnected, and their corresponding elements in the adjacency matrix are set to 0.

[0088] IV. Model Training

[0089] Model training refers to training model parameters using optimization algorithms based on sample data to minimize the loss function. During AI model training, to ensure the model's output is as close to the true value as possible, the weight vector of each layer is updated by comparing the current output value with the true value and considering the difference. For example, if the output value is higher than the true value, the model parameters are adjusted to lower the output value. This process is repeated until the AI ​​model's output value is closer to the true value.

[0090] A loss function is an important equation used to compare the difference between the output value and the true value. The output value of the loss function is also called the loss value. A higher loss value indicates a greater difference. The training objective of an AI model is to minimize the loss value, making it below a threshold, or to make the loss function value meet the target requirements. For example, if the AI ​​model is a neural network, adjusting the model parameters of the neural network includes adjusting at least one of the following parameters: the number of layers in the neural network, its width, the weights of the neurons, or the parameters in the activation function of the neurons.

[0091] Model training includes initial model training, model update, and model retraining. Initial model training can be understood as training an initialized model with a dataset to produce a usable model parameter set. Model update can be understood as adjusting all or some of the parameters of a trained model. Model retraining can be understood as adding new sample data to the training dataset to obtain a new training dataset, and then retraining the model based on this new training dataset.

[0092] V. AI-based use cases

[0093] AI-based use cases can also be understood as AI functions. In communication scenarios, there are several typical AI use cases, including energy saving, load balancing, mobility optimization, channel state information-reference signal (CSI-RS) feedback enhancement, beam management enhancement, and positioning accuracy enhancement.

[0094] This application uses positioning enhancement as an example to briefly explain its principle. Positioning enhancement is mainly used to improve positioning accuracy. It obtains channel measurement results by measuring reference signals, which can also be understood as channel responses. Several channel responses are used as input to an AI model to obtain the final location of the terminal device. Optionally, the reference signal includes a channel sounding reference signal (SRS) or a downlink positioning reference signal (PRS). This AI positioning model can be deployed on the location management function (LMF) side, or the terminal device or network device can use the AI ​​model to extract features of the channel responses. The LMF uses the extracted features as input to the AI ​​model to obtain the location of the terminal device.

[0095] Taking uplink positioning as an example, the terminal device sends an SRS to the network device. The network device measures the SRS to obtain the channel response. Figure 7A shows a schematic diagram of AI-based uplink positioning. In Figure 7A, the AI ​​model is deployed on the LMF side. The network device sends the channel response to the LMF. The LMF uses the channel responses of multiple network devices as input to the AI ​​model and outputs the location information of the terminal device. Assuming that the number of antennas of the network device is 16 and the number of subcarriers is 4096, then each network device needs to send 65536 (16×4096) complex information to the LMF. Figure 7B shows another schematic diagram of AI-based uplink positioning. In Figure 7B, the AI ​​model is deployed on both the network device and the LMF. Each network device uses the channel response as input, extracts the channel features of the channel response through the AI ​​model on the network device side, and then sends it to the LMF. The LMF uses the received channel features as input and uses the AI ​​model on the LMF side to obtain the location of the terminal device. The dimension of the channel features is determined by the output dimension of the AI ​​model on the network device side. For example, the network device side can extract features of dimension

[0128] from the channel response of dimension [16,4096] and send them to the LMF.

[0096] Downlink positioning is similar to uplink positioning, except that the network device sends a PRS (Pulse Responsibility Scale) to the terminal device. The terminal device measures the PRS to obtain the channel response and sends the channel response, or features extracted from the channel response by the AI ​​model, to the LMF (Low-Level Function). The channel response-based features mentioned in this application can be the channel response itself or features extracted from the channel response by the AI ​​model.

[0097] To improve the performance of the model, this application proposes a data processing method. This method can be applied to the same information system described above.

[0098] It should be understood that in this application, the indication includes direct indication (also known as explicit indication) and implicit indication. Direct indication information A refers to information A being included; implicit indication information A refers to information A being indicated through the correspondence between information A and information B, and through direct indication information B. The correspondence between information A and information B can be predefined, pre-stored, pre-burned, or pre-configured.

[0099] It should be understood that in this application, information C is used to determine information D, including both situations where information D is determined solely based on information C and situations where it is determined based on information C and other information. Furthermore, information C can also be used to determine information D indirectly, for example, where information D is determined based on information E, and information E is determined based on information C.

[0100] Furthermore, in the embodiments of this application, "network element A sends information A to network element B" can be understood as network element B being the destination of information A or an intermediate network element in the transmission path between the destination and network element B, which may include sending information directly or indirectly to network element B. "Network element B receives information A from network element A" can be understood as network element A being the source of information A or an intermediate network element in the transmission path between the source and network element A, which may include receiving information directly or indirectly from network element A. Information may undergo necessary processing between the source and destination, such as format changes, but the destination can understand the valid information from the source. Similar expressions in this application can be understood in a similar way and will not be elaborated further here.

[0101] As shown in Figure 8, the data processing method includes steps 801 to 802, with step 801 being optional. The execution entity corresponding to the method shown in Figure 8 is an AI node or an AI node chip. An AI node refers to a device that has deployed an AI model and possesses AI functionality. For example, an AI node can be a network device or a terminal device.

[0102] For example, the network device can be one or more devices in the core network device, access network node (RAN node), or OAM as shown in Figure 1. For instance, the near real-time RIC is set in the RAN node (e.g., in CU, DU), while the non-real-time RIC is set in the OAM, in the cloud server, in the core network device, or in other network devices. The RIC can obtain the node input feature matrix from the RAN node (e.g., CU, CU-CP, CU-UP, DU, and / or RU), where the multiple elements in the node input feature matrix are the input features of multiple nodes.

[0103] For example, near real-time RIC and non-real-time RIC can also be set up separately as a network element, and the network device can be a near real-time RIC or a non-real-time RIC.

[0104] 801. Obtain the node input feature matrix, which contains multiple elements representing the input features of multiple nodes.

[0105] For example, an AI device may obtain the node input feature matrix by: reading the node input feature matrix from the AI ​​device's local storage space; receiving the node input feature matrix sent from another device; or processing other matrices to obtain the node input feature matrix, such as processing the initial node feature matrix. This application does not limit how the AI ​​device obtains the node input feature matrix.

[0106] 802. Input the node input feature matrix into the first model to obtain the node output feature matrix. The node output feature matrix includes multiple elements, which are the output features of multiple nodes. The parameters in the first model include the correlation between two nodes among the multiple nodes. The correlation between two nodes among the multiple nodes is determined by comparing the correlation between the two nodes and the magnitude of a first threshold. The first threshold is the element corresponding to the two nodes in the first threshold vector. The first threshold vector is obtained based on the first function used by the first model. The first function is differentiable.

[0107] In this embodiment of the application, the correlation between two nodes among the plurality of nodes is determined by comparing the correlation between the two nodes with the magnitude of a first threshold. It can be further understood that: when the correlation between two nodes among the plurality of nodes is greater than the first threshold, the two nodes are correlated; when the correlation between two nodes among the plurality of nodes is less than or equal to the first threshold, the two nodes are not correlated.

[0108] Alternatively, the two nodes described in this embodiment can also be understood as a pair of nodes. For example, the description can be presented by replacing two nodes with a pair of nodes: the parameters in the first model include the correlation of multiple pairs of nodes. Each pair of nodes in these multiple pairs includes two nodes, and both nodes in each pair belong to the multiple pairs of nodes. The first pair of nodes is one of these multiple pairs of nodes. When the correlation of the first pair of nodes is greater than a first threshold, the two nodes in the first pair are correlated; when the correlation of the first pair of nodes is less than or equal to the first threshold, the two nodes in the first pair are not correlated. The first threshold is the element corresponding to the first pair of nodes in the first threshold vector.

[0109] Correlation is used to represent the degree of association between two nodes. A higher correlation indicates a stronger association between the two nodes, while a lower correlation indicates a weaker association. Optionally, the correlation between the two nodes can also be determined based on distance, channel results, or response.

[0110] For example, when the correlation between two nodes is determined based on the distance between them, the correlation is inversely proportional to the distance between the two nodes. Understandably, the greater the distance between two nodes, the lower their correlation; conversely, the smaller the distance between two nodes, the greater their correlation.

[0111] For example, when the correlation between two nodes is determined based on a first parameter between the channel measurement results or channel responses of the two nodes, the correlation between the two nodes is directly proportional to the first parameter between the channel measurement results or channel responses of the two nodes. It can be understood that the larger the first parameter between the channel measurement results or channel responses of the two nodes, the greater the correlation between the two nodes; the smaller the first parameter between the channel measurement results or channel responses of the two nodes, the smaller the correlation between the two nodes. The first parameter includes at least one of the following: time-domain correlation, frequency-domain correlation, spatial correlation, statistical correlation, signal strength, time delay and phase difference, or signal bandwidth and center frequency. The channel measurement result refers to the result obtained by measuring a reference signal, which can also be understood as the channel response. Optionally, the reference signal is a PRS or SRS.

[0112] The correlation between the two nodes described above can also be expressed as the correlation of their channel characteristics. The correlation of channel characteristics between two nodes typically refers to the similarity or relationship between two or more wireless signals. In the field of wireless communication, signal correlation can be used to describe the correlation of signals in the time domain, frequency domain, or space. The following explanation, based on signal correlation, addresses at least one of the first parameters described above, namely, time-domain correlation, frequency-domain correlation, spatial correlation, statistical correlation, signal strength, time delay and phase difference, or signal bandwidth and center frequency:

[0113] 1. Time-domain correlation refers to the similarity between two signals in the time domain. It can be described by the autocorrelation function or cross-correlation function of the time-domain responses of the two signals. Combining the above description of the correlation between the first parameter and the two nodes, when the first parameter includes time-domain correlation, it is understandable that the greater the similarity between the signals of the two nodes in the time domain (i.e., the greater the time-domain correlation), the greater the correlation between the two nodes; conversely, the smaller the similarity between the signals of the two nodes in the time domain (i.e., the smaller the time-domain correlation), the smaller the correlation between the two nodes.

[0114] 2. Frequency correlation refers to the similarity between two signals in the frequency domain. Frequency correlation can be described by the autocorrelation function or cross-correlation function of the frequency domain responses of the two signals. Combining the above description of the correlation between the first parameter and the two nodes, when the first parameter includes frequency domain correlation, it can be understood that the greater the similarity between the signals of the two nodes in the frequency domain (i.e., the greater the frequency domain correlation), the greater the correlation between the two nodes; conversely, the smaller the similarity between the signals of the two nodes in the frequency domain (i.e., the smaller the frequency domain correlation), the smaller the correlation between the two nodes.

[0115] 3. Spatial correlation refers to the spatial correlation between two signals in wireless communication. It can also be described as the similarity of signal strength or pattern between the signals of two nodes at different geographical locations. Combining the above description of the first parameter and the correlation between two nodes, when the first parameter includes spatial correlation, it is understandable that the greater the similarity of signal strength or pattern between the signals of two nodes at different geographical locations—that is, the greater the spatial correlation—the greater the correlation between the two nodes; conversely, the smaller the similarity of signal strength or pattern between the signals of two nodes at different geographical locations—that is, the smaller the spatial correlation—the smaller the correlation between the two nodes.

[0116] 4. Statistical correlation refers to the correlation between two signals described by statistical methods, such as using covariance or correlation coefficient to reflect statistical correlation. Combining the above description of the correlation between the first parameter and the two nodes, when the first parameter includes statistical correlation, it is understandable that the larger the covariance or correlation coefficient between the signals of the two nodes, i.e., the greater the statistical correlation, the greater the correlation between the two nodes; conversely, the smaller the covariance or correlation coefficient between the signals of the two nodes, i.e., the smaller the statistical correlation, the smaller the correlation between the two nodes.

[0117] 5. Signal strength refers to the intensity or power level of a signal. Understandably, the correlation between two signals can be reflected by the similarity between their signal strength or power levels. Combining the above description of the correlation between the first parameter and two nodes, when the first parameter includes the signal strength or power levels of the two nodes, the greater the similarity between the signal strength or power levels of the two nodes, the greater the correlation between the two signals, and correspondingly, the greater the correlation between the two nodes; conversely, the smaller the similarity between the signal strength or power levels of the two nodes, the smaller the correlation between the two signals, and correspondingly, the smaller the correlation between the two nodes.

[0118] 6. Time delay and phase difference refer to the time delay and phase difference of a signal arriving at the receiver in a multipath propagation environment. Understandably, the correlation between two signals can be reflected by the time delay and phase difference of their arrival at the receiver in a multipath propagation environment. Combining the above description of the correlation between the first parameter and the two nodes, when the first parameter includes the time delay and phase difference of the signals from the two nodes, the greater the similarity of the time delays or the greater the similarity of the phase differences, the greater the correlation between the two signals, and correspondingly, the greater the correlation between the two nodes. Conversely, the smaller the similarity of the time delays or the smaller the similarity of the phase differences, the smaller the correlation between the two signals, and correspondingly, the smaller the correlation between the two nodes.

[0119] 7. The bandwidth and center frequency of a signal refer to the bandwidth occupied by the signal when it is transmitted and the center frequency of the signal. Understandably, the correlation between two signals can be reflected by their bandwidth and center frequency. Combining the above description of the correlation between the first parameter and two nodes, when the first parameter includes the bandwidth and center frequency of the signals of the two nodes, the greater the similarity of the bandwidths of the signals of the two nodes, or the closer the distance between the center frequencies of the signals of the two nodes, the greater the correlation between the two signals, and correspondingly, the greater the correlation between the two nodes. Conversely, the smaller the similarity of the bandwidths of the signals of the two nodes, or the greater the distance between the center frequencies of the signals of the two nodes, the smaller the correlation between the two signals, and correspondingly, the smaller the correlation between the two nodes.

[0120] The signals of the two nodes described above can be understood as signals transmitted by the two nodes respectively. For example, the signals of node 1 and node 2 can be understood as signal 1 transmitted by node 1 and signal 2 transmitted by node 2.

[0121] Optionally, before training the first model, the first model can initialize the correlation between two nodes among the preset plurality of nodes, and then adjust it through model training to obtain a more accurate correlation. This application embodiment does not limit how the correlation between two nodes among the plurality of nodes included in the first model is determined.

[0122] The first threshold vector includes multiple elements, each representing a threshold required to determine the correlation between its corresponding two nodes. The values ​​of different elements in the first threshold vector can be the same or different. For example, the first threshold vector includes threshold 1 and threshold 2, where threshold 1 corresponds to nodes 1 and 2, and threshold 2 corresponds to nodes 3 and 4. Therefore, the correlation between nodes 1 and 2 is determined by comparing their correlation with threshold 1. If the correlation between nodes 1 and 2 is greater than threshold 1, then nodes 1 and 2 are correlated; if the correlation is less than or equal to threshold 1, then nodes 1 and 2 are not correlated. Similarly, the correlation between nodes 3 and 4 is determined by comparing their correlation with threshold 2. If the correlation is greater than threshold 2, then nodes 3 and 4 are correlated; if the correlation is less than or equal to threshold 2, then nodes 3 and 4 are not correlated.

[0123] Furthermore, the correlation between two nodes among these multiple nodes can be represented by constructing a node adjacency matrix, in which the elements are determined based on the correlation between their corresponding two nodes. For example, when two nodes are correlated, the value of the element corresponding to those two nodes in the node adjacency matrix is ​​1; when two nodes are not correlated, the value of the element corresponding to those two nodes in the node adjacency matrix is ​​0. An element in this node adjacency matrix can be represented by the following formula (1):

[0124] Among them, A FP_ij This represents the elements corresponding to nodes i and j in the adjacency matrix, where i and j are both positive integers less than the total number of nodes N. ij t represents the correlation between node i and node j. ij This represents the value of the element corresponding to node i and node j in the first threshold vector.

[0125] Understandably, the node adjacency matrix and the node input feature matrix can be combined to represent a graph. The AI ​​device can determine the node output feature matrix based on this node adjacency matrix and the node input feature matrix through a first model.

[0126] Furthermore, during the process of the AI ​​device inputting the node input feature matrix into the first model to obtain the final node output feature matrix, the node input feature matrix can be filtered based on the node adjacency matrix to obtain a sparse feature matrix. Then, the node input feature matrix is ​​obtained based on this sparse feature matrix and the first model. In one possible implementation, this sparse matrix can be represented by the following formula (2):

[0127] in, Let X represent the sparse characteristic matrix. FP Let A represent the node input feature matrix. FP Represents the node adjacency matrix. This represents the element-wise multiplication of two matrices. Understandably, in the node adjacency matrix, the element corresponding to two unrelated nodes has a value of 0, while the element corresponding to two related nodes has a value of 1. This method filters out the input features corresponding to unrelated nodes from the node input feature matrix.

[0128] From the above formula (1), it can be seen that the functional expression of the elements in the node adjacency matrix is ​​a step function. Differentiability means that a function has a continuous rate of change at a certain point, as shown in Figure 9, the step function δ(x) = 1. (-∞，0] (x) is discontinuous at 0, therefore the step function is not differentiable.

[0129] Since the step function is non-differentiable, the function expressions corresponding to the elements in the sparse feature matrix obtained by multiplying the node adjacency matrix and the node input feature matrix are also non-differentiable. However, in this embodiment, the first threshold vector in the first model needs to be updated using gradient optimization, which requires the functions in the first model to be differentiable. During the training of the first threshold vector using the first model, a first function can be used to optimize the node adjacency matrix. This first function is differentiable, thus making the function expressions of the elements in the node adjacency matrix differentiable. Therefore, the AI ​​device can use this first function to train the first threshold vector using the first model.

[0130] It should also be noted that the differentiability of the first function includes two cases: the first function is completely differentiable, and the first function is partially differentiable and partially non-differentiable. Complete differentiability of the first function means that the first function is differentiable over the entire domain, that is, it is differentiable for any value of the input parameter. Partial differentiability and partial non-differentiability means that the first function is differentiable when the input parameter takes on a first range of values, but not differentiable when the input parameter takes on a second range of values.

[0131] The following shows several possible expressions for the first function:

[0132] Example 1: The first function satisfies the following formula (3):

[0133] in, Let be the output parameter of the first function, and x be the input parameter of the first function. The ReLU function is a linear rectified function, satisfying ReLU(-tanh(γx))=max(0,-tanh(γx)). The max function is the maximum value function. The tanh function is the hyperbolic tangent function, satisfying... γ is a preset coefficient.

[0134] Example 2: The first function satisfies the following formula (4):

[0135] in, Let x be the output parameter of the first function, and let x be the input parameter of the first function. The ReLU function is a linear rectified function, satisfying ReLU(-sigmoid(γx)+b)=max(0,-sigmoid(γx)+b). The max function is the maximum value function, and the sigmoid function satisfies… a, b, and γ are preset coefficients. Optionally, a can be 2 and b can be 0.5.

[0136] Example 3: The first function satisfies the following formula (5):

[0137] in, Let x be the output parameter of the first function, and let x be the input parameter of the first function. The ReLU function is a linear rectified function, and the ReLU function satisfies ReLU(-softsign(γx))=max(0,-softsign(γx)). The max function is the maximum value function, and the softsign function satisfies... γ is a preset coefficient.

[0138] In all three examples above, the first function is completely differentiable. Figure 10 shows the graphs of the first functions corresponding to Examples 1, 2, and 3. The first functions for Examples 1, 2, and 3 can all effectively truncate values ​​greater than 0, but the difference lies in the slope corresponding to input parameters less than 0. The expression of the first function in Example 1 is closer to the effect of a step function than the expressions in Examples 2 and 3.

[0139] It should also be noted that the first function may have other forms, and the embodiments of this application do not limit the form of the first function.

[0140] Based on the data processing method described in the embodiments of this application, since the first function is differentiable, the first threshold vector can be trained using the first function through the first model. The first threshold vector used to determine the association between nodes is a trainable parameter in the first model, and each element in the first threshold vector represents the threshold required to determine the association between its corresponding specific two nodes. The first threshold vector can also be updated and optimized through a backpropagation process. Therefore, the first threshold vector makes the process of determining whether there is a correlation between nodes more intelligent and accurate, which is beneficial to improving model performance.

[0141] In one possible implementation, the process of the first device performing step 802 can be a training process for the first model, or a process of using the first model for model inference. The model training mentioned in this embodiment includes initial model training, model updating, and model retraining. The following detailed description uses step 802 as an example of a training process for the first model:

[0142] After inputting the feature matrices into the input nodes of the first model to obtain the node output feature matrices, the AI ​​device can compare the node output feature matrices with the node true feature matrices. Following the direction in which the node output feature matrices converge towards the node true feature matrices, the model parameters of the first model are updated. Alternatively, it can be understood that the AI ​​device calculates the difference between the node output feature matrices and the node true feature matrices using a loss function to obtain a loss value, and updates the model parameters of the first model in the direction of reducing this loss value. The model parameters in the first model include a first threshold vector, and the elements in the node true feature matrices represent the true features of the multiple nodes.

[0143] Taking the first model as a two-layer GCN as an example, the AI ​​device makes the node output feature matrix converge to the node's true feature matrix according to the following formula (6):

[0144] in, This is represented as the truth feature matrix of a node, meaning the truth feature matrix of a node is an N-dimensional matrix. FP A matrix of dimension ×2, N FP Indicates the number of nodes. This represents the node adjacency matrix, meaning the node adjacency matrix is ​​an N-ary matrix. FP ×N FP A matrix of dimension σ. σ is a non-linear activation function. This represents the node's input feature matrix, meaning the node's input feature matrix is ​​an N-ary matrix. FP ×D F A matrix of dimension D FThis represents the dimension of the input features, and W1 and W2 are the training parameters in the first model. (The rest of the text appears to be a garbled string of characters and doesn't translate directly.) The node output feature matrix can be calculated. Then, according to the node-based output feature matrix Y FP To the node true characteristic matrix The direction of convergence is determined by adjusting the model parameters W1 and W2 of the first model so that the node output feature matrix converges to the node true feature matrix.

[0145] In one possible implementation, when the correlation between two nodes is determined based on the distance between them, the value of each element in the first threshold vector is greater than or equal to 0, and less than or equal to the maximum distance between any two nodes. The maximum distance between any two nodes can also be understood as the maximum distance at which the two nodes can receive signals from each other. When the distance between two nodes exceeds this maximum distance, the two nodes will be unable to transmit signals. Since the first threshold vector is a trainable parameter vector, to avoid the values ​​of the elements in the first threshold vector obtained after the first model training not satisfying the condition [0, l], max Requirements within, l max Let be the maximum distance between any two nodes. Therefore, a threshold scaling operation can be applied during training. For example, the AI ​​device can obtain a second threshold vector through training with the first function, and then scale the second threshold vector to obtain a first threshold vector, where the first threshold vector and the second threshold vector satisfy the following formula (7):

[0146] in, Let t be the first threshold vector. v For the second threshold vector, l max The sigmoid function satisfies the maximum distance between any two nodes.

[0147] In another possible implementation, when the correlation between two nodes is determined based on a first parameter between the channel measurement results or channel responses of the two nodes, the first parameter includes at least one of the following: time-domain correlation, frequency-domain correlation, spatial correlation, statistical correlation, signal strength, time delay and phase difference, or signal bandwidth and center frequency. The value of each element in the first threshold vector is greater than or equal to 0 and less than or equal to 1. Since the first threshold vector is a trainable parameter vector, in order to avoid the value of the elements in the first threshold vector obtained after the first model training not satisfying the requirement of being in the interval [0,1], a threshold scaling operation can be applied during the training process. For example, the AI ​​device can obtain the second threshold vector by training the first function, and then scale the second threshold vector to obtain the first threshold vector, wherein the first threshold vector and the second threshold vector satisfy the following formula (8):

[0148] in, Let t be the first threshold vector. v For the second threshold vector, the sigmoid function satisfies

[0149] In one possible implementation, the first model is used for node localization, and the node's output features are used to indicate the node's location information. Optionally, the node's input features may refer to features such as the node's channel characteristics, the node's channel response, and the distance between the node and other neighboring nodes. This application does not limit the specific type of the node's input features.

[0150] It should also be noted that the first model can also be applied to models in other scenarios besides positioning. This application embodiment does not limit the application scenario of the first model.

[0151] The apparatus provided in the embodiments of this application will be described below.

[0152] This application divides the device into functional modules according to the above method embodiments. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. It should be noted that the module division in this application is illustrative and only represents one logical functional division; other division methods may be used in actual implementation. The device of the embodiment of this application will be described in detail below with reference to Figures 11 to 13.

[0153] Figure 11 is a schematic diagram of a device provided in an embodiment of this application. As shown in Figure 11, the device includes a processing module 1101 and a transceiver module 1102. The transceiver module 1102 can implement corresponding communication functions, and the processing module 1101 is used to implement corresponding processing functions. For example, the transceiver module 1102 can also be referred to as an interface, communication interface, communication module, or input / output interface, etc.

[0154] In some embodiments of this application, the device can be used to perform the actions performed by the AI ​​node in the above method embodiments. In this case, the AI ​​node can be the AI ​​node itself or a chip or functional module configurable within the AI ​​node. The transceiver module 1102 is used to perform transceiver-related operations or input / output-related operations of the AI ​​node in the above method embodiments, and the processing module 1101 is used to perform processing-related operations of the AI ​​node in the above method embodiments.

[0155] For example, processing module 1101 can be used to input the node input feature matrix into a first model to obtain a node output feature matrix. The node input feature matrix includes multiple elements representing the input features of multiple nodes, and the node output feature matrix includes multiple elements representing the output features of multiple nodes. The parameters in the first model include the correlation between two nodes among the multiple nodes. When the correlation between two nodes among the multiple nodes is greater than a first threshold, the two nodes are correlated. When the correlation between two nodes among the multiple nodes is less than or equal to the first threshold, the two nodes are not correlated. The first threshold is the element corresponding to two nodes in the first threshold vector, and the first threshold vector is obtained based on a first function used by the first model, and the first function is differentiable.

[0156] For specific explanations of terms or steps in the above embodiments, please refer to the descriptions in the above method embodiments, which will not be detailed here.

[0157] The specific descriptions of the transceiver module and processing module shown in the above embodiments are merely examples. For the specific functions or execution steps of the transceiver module and processing module, please refer to the above method embodiments, which will not be described in detail here.

[0158] It is understandable that the module division in the above-mentioned device is merely a logical functional division. Each function can correspond to a functional module, or two or more functions can be integrated into one functional module. In actual implementation, all or some modules can be integrated into one physical entity, or they can be distributed across different physical entities. Furthermore, the above-mentioned functional modules can be implemented in hardware, software, or a combination of both.

[0159] In one 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 application-specific integrated circuits (ASICs), or one or more central processing units (CPUs), one or more microcontroller units (MCUs), one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0160] The apparatus of the embodiments of this application has been described above. The possible product forms of the apparatus are described below. Any product possessing the functions of the apparatus described in FIG11 above falls within the protection scope of the embodiments of this application. The following description is merely illustrative and does not limit the product form of the apparatus of the embodiments of this application to this.

[0161] In one possible implementation, in the device shown in FIG11, the processing module 1101 can be one or more processing circuits, and the transceiver module 1102 can be a transceiver circuit, or the transceiver module 1102 can also be a transmitting module and a receiving module. The transmitting module can be a transmitting circuit, and the receiving module can be a receiving circuit, which are integrated into a single device, such as a transceiver circuit. In the embodiments of this application, the processing circuit and the transceiver circuit can be coupled, etc., and the connection method of the processing circuit and the transceiver circuit is not limited in the embodiments of this application. In the process of performing the above method, the process of sending information in the above method can be the process of the processing circuit outputting the above information. When outputting the above information, the processing circuit outputs the above information to the transceiver circuit so that the transceiver circuit can transmit (or output). After the above information is output by the processing circuit, it may need to undergo other processing before reaching the transceiver circuit. Similarly, the process of receiving information in the above method can be the process of the processing circuit receiving the input above information. When the processing circuit receives the input information, the transceiver circuit receives the above information and inputs it into the processing circuit. Furthermore, after the transceiver circuit receives the aforementioned information, the information may need to undergo further processing before being input into the processing circuit.

[0162] Figure 12 is a schematic diagram of the structure of an apparatus provided in an embodiment of this application. As shown in Figure 12, the apparatus 120 includes one or more processing circuits 1220 and transceiver circuits 1210.

[0163] In some embodiments of this application, the apparatus can be used to perform the steps, methods, or functions performed by the above-described apparatus. For example, the processing circuit 1220 can be used to perform the functions or steps implemented by the processing module 1101 shown in FIG. 11, and the transceiver circuit 1210 can be used to perform the functions or steps implemented by the transceiver module 1102 shown in FIG. 11. Detailed descriptions of the processing circuit 1220 and the transceiver circuit 1210 can be found in FIG. 11 or the method embodiments shown above, and will not be elaborated further here.

[0164] For example, the processing circuitry may be one or more processors, or all or part of the circuitry within one or more processors. The transceiver circuitry may be a transceiver, an input / output circuit, or an interface circuit, etc.

[0165] For example, in various implementations of the apparatus shown in FIG12, the transceiver circuitry may include a receiver for performing a receiving function (or operation) and a transmitter for performing a transmitting function (or operation). The transceiver circuitry is also used to communicate with other devices / appliances via a transmission medium.

[0166] Optionally, the device 120 may further include one or more memories 1230 for storing program instructions and / or data. The memory 1230 is coupled to the processing circuitry 1220. The coupling in this embodiment is 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 processing circuitry 1220 may operate in conjunction with the memory 1230. The processing circuitry 1220 may execute the program instructions stored in the memory 1230. Optionally, at least one of the above-mentioned memories may be included in the processing circuitry.

[0167] This application embodiment does not limit the specific connection medium between the transceiver circuit 1210, processing circuit 1220, and memory 1230. In Figure 12, the memory 1230, processing circuit 1220, and transceiver circuit 1210 are connected via a bus 1240, which is represented by a thick line in Figure 12. The connection methods between other components are only illustrative and are not intended to be limiting. The bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used in Figure 12, but this does not mean that there is only one bus or one type of bus.

[0168] In the embodiments of this application, the processing circuit may be a general-purpose processing circuit, a digital signal processing circuit, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc., and can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processing circuit may be a microprocessor circuit or any conventional processing circuit, etc. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processing circuit, or being executed by a combination of hardware and software modules in the processing circuit, etc.

[0169] In this application embodiment, the memory may include, but is not limited to, non-volatile memory such as hard disk drive (HDD) or solid-state drive (SSD), random access memory (RAM), erasable programmable read-only memory (EPROM), read-only memory (ROM), or compact disc read-only memory (CD-ROM), etc. Memory is any storage medium capable of carrying or storing program code having instruction or data structure forms, and capable of being read and / or written by a computer (such as the device shown in this application), but is not limited to this. The memory in this application embodiment may also be a circuit or any other device capable of implementing storage functions, used to store program instructions and / or data.

[0170] For example, the processing circuit 1220 is mainly used to process communication protocols and communication data, control the entire device, execute software programs, and process the data of the software programs. The memory 1230 is mainly used to store software programs and data. The transceiver circuit 1210 may include a control circuit and an antenna. The control circuit is mainly used for converting baseband signals to radio frequency signals and processing radio frequency signals. The antenna is mainly used for transmitting and receiving radio frequency signals in the form of electromagnetic waves. Input / output devices, such as touch screens, displays, and keyboards, are mainly used to receive user input data and output data to the user.

[0171] When the device is powered on, the processing circuit 1220 can read the software program in the memory 1230, interpret and execute the instructions of the software program, and process the data of the software program. When data needs to be transmitted wirelessly, the processing circuit 1220 performs baseband processing on the data to be transmitted and outputs the baseband signal to the radio frequency (RF) circuit. The RF circuit performs RF processing on the baseband signal and transmits the RF signal outward in the form of electromagnetic waves through the antenna. When data is sent to the device, the RF circuit receives the RF signal through the antenna, converts the RF signal into a baseband signal, and outputs the baseband signal to the processing circuit 1220. The processing circuit 1220 converts the baseband signal into data and processes the data.

[0172] In another implementation, the radio frequency circuit and antenna can be set up independently of the processing circuit that performs baseband processing. For example, in a distributed scenario, the radio frequency circuit and antenna can be arranged remotely, independent of the device.

[0173] The apparatus shown in this application embodiment may have more components than those in Figure 12, and this application embodiment does not limit this. The methods performed by the processing circuit and transceiver circuit shown above are merely examples, and the specific steps performed by the processing circuit and transceiver circuit can be referred to the methods described above.

[0174] In another possible implementation, in the device shown in FIG11, the processing module 1101 can be one or more logic circuits, and the transceiver module 1102 can be an input / output interface, or a communication interface, or an interface circuit, or an interface, etc. Alternatively, the transceiver module 1102 can also be a transmitting module and a receiving module, where the transmitting module can be an output interface and the receiving module can be an input interface, and the transmitting module and the receiving module are integrated into one module, such as an input / output interface.

[0175] Figure 13 is a schematic diagram of a device provided in an embodiment of this application. As shown in Figure 13, the device includes a logic circuit 1301 and an interface circuit 1302. That is, the processing module 1101 can be implemented using the logic circuit 1301, and the transceiver module 1102 can be implemented using the interface circuit 1302. The logic circuit 1301 can be a chip, a processing circuit, an integrated circuit, or a system-on-a-chip (SoC) chip, etc., and the interface circuit 1302 can be a communication interface, an input / output interface, pins, etc. For example, Figure 13 illustrates the device as a chip, which includes the logic circuit 1301 and the interface circuit 1302.

[0176] In this embodiment, the logic circuit and the interface can also be coupled to each other. The specific connection method of the logic circuit and the interface is not limited in this embodiment. For example, the logic circuit 1301 can be used to execute the functions or steps implemented by the processing module 1101 shown in FIG. 11, and the interface circuit 1302 can be used to execute the functions or steps implemented by the transceiver module 1102 shown in FIG. 11. For a detailed description of the logic circuit 1301 and the interface circuit 1302, please refer to FIG. 11 or the method embodiment shown above, which will not be detailed here.

[0177] The apparatus shown in the embodiments of this application can be implemented in hardware or software, and the embodiments of this application do not limit this.

[0178] This application also provides a communication system, which includes an apparatus that can be used to perform the methods in any of the foregoing embodiments.

[0179] In addition, this application also provides a computer program for implementing the operations and / or processes performed by various devices in the method provided in this application.

[0180] This application also provides a computer-readable storage medium storing computer code that, when executed on a computer, causes the computer to perform the operations and / or processes performed by the various devices in the methods provided in this application.

[0181] This application also provides a computer program product comprising computer code or a computer program that, when run on a computer, causes the operations and / or processes performed by various entities in the method provided in this application to be executed.

[0182] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, devices, or modules, or they may be electrical, mechanical, or other forms of connection.

[0183] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the technical effects of the solutions provided in the embodiments of this application.

[0184] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0185] If the integrated module is implemented as a software functional module and sold or used as an independent product, it 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 readable storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned readable 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.

[0186] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A data processing method, characterized in that, The method includes: The node input feature matrix is ​​input into the first model to obtain the node output feature matrix. The node input feature matrix includes multiple elements that are the input features of multiple nodes, and the node output feature matrix includes multiple elements that are the output features of the multiple nodes. The parameters in the first model include the correlation between two nodes among multiple nodes; when the correlation between two nodes among multiple nodes is greater than a first threshold, the two nodes are correlated; when the correlation between two nodes among multiple nodes is less than or equal to the first threshold, the two nodes are not correlated; the first threshold is the element corresponding to the two nodes in the first threshold vector, and the first threshold vector is obtained based on the first function used by the first model, and the first function is differentiable.

2. The method according to claim 1, characterized in that, The first function satisfies: Among them, the γ is the output parameter of the first function, x is the input parameter of the first function, the ReLU function is a linear rectified function, the tanh function is a hyperbolic tangent function, and γ is a preset coefficient.

3. The method according to claim 1, characterized in that, The first function satisfies: Among them, the Let be the output parameter of the first function, x be the input parameter of the first function, the ReLU function be a linear rectified function, and a, b, and γ be preset coefficients. The sigmoid function satisfies...

4. The method according to claim 1, characterized in that, The first function satisfies: Among them, the Let x be the output parameter of the first function, x be the input parameter of the first function, the ReLU function be a linear rectified function, γ be a preset coefficient, and the softsign function satisfy...

5. The method according to any one of claims 1 to 4, characterized in that, The value of each element in the first threshold vector is greater than or equal to 0, and less than or equal to the maximum distance between any two nodes.

6. The method according to any one of claims 1 to 4, characterized in that, The value of each element in the first threshold vector is greater than or equal to 0 and less than or equal to 1.

7. The method according to any one of claims 1 to 6, characterized in that, The first model is used for node localization, and the output features are used to indicate the location information of the node.

8. An apparatus, characterized in that, The apparatus includes modules or units for performing the method according to any one of claims 1 to 7.

9. An apparatus, characterized in that, The apparatus includes a processor for executing a computer program or instructions stored in a memory to implement the method as described in any one of claims 1-7.

10. A chip, characterized in that, The chip includes a processor configured to enable the chip to implement the method as described in any one of claims 1-7.

11. A computer-readable storage medium, characterized in that, The storage medium stores a computer program or instructions, and when the computer program or instructions are executed by the communication device, the method as described in any one of claims 1 to 7 is performed.

12. A computer program product, characterized in that, When the computer program product is executed, the method described in any one of claims 1 to 7 is performed.

Images