Communication method, storage device and electronic device

Abstract

The embodiment of the invention provides a communication method, a storage medium and an electronic device, and the communication method comprises the steps: receiving a reasoning request from first management equipment through second management equipment; and the second management device performs AI / ML reasoning according to the reasoning request. According to the method and the device, the problems of lack of a clear reasoning request and lack of AI / ML reasoning result evaluation and feedback in related technologies are solved, and the effects of performing AI / ML reasoning according to the clear reasoning request, performing feedback adjustment on the ML entity according to the evaluation of the AI / ML reasoning result and improving the communication performance are achieved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of communications, and in particular, to a communication method, a storage device, and an electronic apparatus. Background Art

[0002] SA5 is studying and finalizing other aspects of the Artificial Intelligence / Machine Learning (AI / ML) management specifications for Rel-18, with a particular focus on enabling and managing AI / ML capabilities in various areas of the 5G system, including management and orchestration, the 5G Core (5GC), and the Next Generation Radio Access Network (NG-RAN). Management and orchestration can be defined as Management Data Analytical (MDA) in TS 28.104. The 5GC can be defined as the Network Data Analytics Function (NWDAF) in TS 23.288. NG-RAN can be defined as Radio Access Network (RAN) intelligence in TS 38.300 and TS 38.401.

[0003] The AI / ML management functions defined in SA5 Rel-18 include the management and operation of ML training, ML testing, AI / ML simulation, ML entity deployment, and AI / ML reasoning. The AI / ML reasoning stage is an indispensable part of the entire AI / ML management function, but currently 3GPP TS28.105 does not define AI / ML reasoning requests. AI / ML reasoning requests need to support both reasoning requests based on ML entities and reasoning requests based on AI / ML reasoning functions, so the present invention will propose a solution for this. In addition, in the AI / ML management process in the related art, the trained model can be simulated by AI / ML to test the reasoning performance of the model, but the effect of the AI / ML reasoning results after application and the impact on the object also need to be collected to evaluate the actual performance of the reasoning results.

[0004] In summary, related technologies lack clear reasoning requests and lack evaluation and feedback of AI / ML reasoning results. Summary of the Invention

[0005] Embodiments of the present invention provide a communication method, a storage device, and an electronic apparatus to at least address the problems in related technologies of lacking clear inference requests and lacking evaluation and feedback of AI / ML inference results.

[0006] According to one embodiment of the present invention, a communication method is provided for AI / ML reasoning, including: a second management device receiving an reasoning request from a first management device; and the second management device performing the AI / ML reasoning according to the reasoning request.

[0007] According to another embodiment of the present invention, a communication method is provided for AI / ML reasoning, including: a first management device sends an reasoning request to a second management device to instruct the second management device to perform the AI / ML reasoning according to the reasoning request.

[0008] According to yet another embodiment of the present invention, a computer-readable storage medium is provided, in which a computer program is stored. The computer program is configured to execute the steps of any one of the above method embodiments when run.

[0009] According to another embodiment of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any one of the above method embodiments.

[0010] This invention provides a communication method for AI / ML inference, in which a second management device receives an inference request from a first management device; the second management device then performs AI / ML inference based on the inference request. This method addresses the problems of the related art, which lack clear inference requests and lack evaluation and feedback of AI / ML inference results. It achieves the goal of performing AI / ML inference based on clear inference requests and providing feedback and adjustments to ML entities based on the evaluation of the AI / ML inference results, thereby improving communication performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a hardware structure block diagram of a computer terminal for a communication method according to an embodiment of the present invention;

[0012] Figure 2 is a schematic diagram of a network architecture for operating a communication method according to an embodiment of the present invention;

[0013] Figure 3 This is the process of the communication method of the embodiment of the present invention Figure 1 ;

[0014] Figure 4 This is the process of the communication method of the embodiment of the present invention Figure 2 ;

[0015] Figure 5 This is the process of the communication method of the embodiment of the present invention Figure 3 ;

[0016] Figure 6This is the process of the communication method of the embodiment of the present invention Figure 4 ;

[0017] Figure 7 This is the process of the communication method of the embodiment of the present invention Figure 5 ;

[0018] Figure 8 This is the process of the communication method of the embodiment of the present invention Figure 6 ;

[0019] Figure 9 This is the process of the communication method of the embodiment of the present invention Figure 7 ;

[0020] Figure 10 This is the process of the communication method of the embodiment of the present invention Figure 8 ;

[0021] Figure 11 This is a schematic diagram of the AI / ML operational workflow in related technologies;

[0022] Figure 12 is a flow chart of a communication method according to a first embodiment of the present invention;

[0023] Figure 13 is a flow chart of a communication method according to a second embodiment of the present invention;

[0024] Figure 14 This is a flow chart of a communication method according to a third embodiment of the present invention. DETAILED DESCRIPTION

[0025] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings and in combination with embodiments.

[0026] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0027] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a computer terminal as an example, Figure 1 FIG. 1 is a hardware structure diagram of a computer terminal of a communication method according to an embodiment of the present invention. Figure 1 As shown, the computer terminal may include one or more ( Figure 1 Only one is shown) a processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices) and a memory 104 for storing data. The computer terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer terminal. For example, the computer terminal may also include Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0028] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the communication method in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implementing the above-mentioned method. The memory 104 may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0029] The transmission device 106 is used to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by a communications provider of a computer terminal. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0030] The embodiment of the present application can be run on Figure 2 In the network architecture shown, Figure 2 Schematic diagram of the network architecture for running the communication method according to an embodiment of the present invention. Figure 2As shown in Figure 1, the network architecture is a Service Based Management Architecture (SBMA), which includes a Business Support System (BSS), a Cross Domain Management Function (CDMnF), a Domain Management Function (Domain MnF), and a Network Element (NE). The Cross Domain Management Function is used to manage one or more Domain Management Functions. The Domain Management Function can be used to manage one or more Network Elements. The following is a brief introduction to each unit.

[0031] like Figure 2 As shown, the business support system is oriented towards communication services (CS), providing functions and management services such as billing, settlement, accounting, customer service, sales, network monitoring, communication service lifecycle management, and business intent translation. The business support system can be an operator's operating system or an operating system for a vertical industry.

[0032] like Figure 2As shown, the cross-domain management function unit is also called the network management function unit (Network Management Function, NMF), which can be a network management entity such as the network management system (Network Management System, NMS), network management service producer (MnS Producer), network management service consumer (MnS Consumer), network function management service consumer (Network Function Management Service Consumer, NFMS_C). Among them, the cross-domain management function unit provides one or more of the following management functions or management services: network life cycle management, network deployment, network fault management, network performance management, network configuration management, network assurance, network optimization functions and translation of the network intent of the service producer (Intent from Communication Service Provider, Intent-CSP), etc. The network referred to in the above management functions or management services may include one or more network elements or sub-networks, or it may be a network slice. That is to say, the network management function unit can be a network slice management function unit (Network Slice Management Function, NSMF), or a cross-domain management data analysis function unit (Management Data Analytical Function, MDAF), or a cross-domain self-organizing network function (Self-Organization Network function, SON Function) or a cross-domain intent management function unit (Intent-Driven Management Service, Intent Driven MnS).

[0033] like Figure 2As shown in the figure, the domain management function unit is also called the sub-network management function unit (Network Subnet Management Function, NSMF) or the network element management function unit. It can be a wireless automation engine (MBB Automation Engine, MAE), an element management system (Element Management System, EMS), a network function management service provider (Network Function Management Service Provider, NFMS_P), a network slice sub-network management function unit (Network Slice Subnet Management Function, NSSMF), a domain management data analysis function unit (Domain Management Data Analytical Function, Domain MDAF), a domain self-organization network function (Self-Organization network function, SON Function), a domain intent management function unit, an MnS Producer, an MnS Consumer and other network element management entities. Domain management function units can be categorized as follows: by network type, they can be divided into: Radio Access Network (RAN) Domain Management Function (RAN domain MnF), Core Network Domain Management Function (CN domain MnF), Transport Network Domain Management Function (TN domain MnF), etc. It should be noted that a domain management function unit can also be a domain network management system that can manage one or more of the access network, core network, or transport network. By administrative region, they can be divided into: a domain management function unit for a specific region, such as the domain management function unit of City A, the domain management function unit of City B, etc. A domain management function unit provides one or more of the following functions or management services: lifecycle management of subnetworks or network elements, deployment of subnetworks or network elements, fault management of subnetworks or network elements, performance management of subnetworks or network elements, assurance of subnetworks or network elements, optimization of subnetworks or network elements, and translation of intents from network operators (Intent-NOPs) for subnetworks or network elements. The subnetwork here includes one or more network elements.A subnetwork can also include subnetworks, that is, one or more subnetworks form a larger subnetwork. The subnetwork here can also be a network slice subnetwork.

[0034] like Figure 2 As shown, a network element is an entity that provides network services. The network element may include a core network element, a radio access network element, or a transport network element. In one embodiment, the core network element may include, but is not limited to, an Access and Mobility Management Function (AMF) entity, a Session Management Function (SMF) entity, a Policy Control Function (PCF) entity, a Network Data Analysis Function (NWDAF) entity, a Network Repository Function (NRF), a gateway, and the like. The radio access network network element may include but is not limited to: various base stations (such as the next generation base station (Generation Node B, gNB), evolved base station (Evolved Node B, eNB), Central Unit Control Panel (CUCP), Central Unit (CU), Distributed Unit (DU), Centralized User Plane Unit (CUUP), etc. In this application, the network function NF is also referred to as the network element NE. The network element can provide one or more of the following management functions or management services: network element lifecycle management, network element deployment, network element fault management, network element performance management, network element assurance, network element optimization function, and network element intent translation.

[0035] The cross-domain management functional unit can be used for model training of AI / ML models and model reasoning of AI / ML models; the domain management functional unit can be used for model training of AI / ML models and model reasoning of AI / ML models; the network element is an entity that provides network services, including core network network elements and access network network elements, which can provide at least one of model training of AI / ML models and model reasoning of AI / ML models.

[0036] The following is an explanation of some terms in the embodiments of the present invention.

[0037] ML Entity: A manageable artifact of an ML model. An ML entity may contain metadata related to the model. This metadata may include, for example, the applicable runtime context of the ML model.

[0038] ML models: Mathematical algorithms that can be “trained” using data and human input as examples to replicate the decisions humans would make when fed the same information. ML models are proprietary and not subject to standardization.

[0039] ML model training: The process performed by the ML training function to obtain training data, run that data through the ML model, infer the associated loss, and adjust the parameterization of the ML model based on the calculated loss.

[0040] ML initial training: ML model training to generate the initial version of the ML entity.

[0041] ML retraining: The process of retraining a previously trained ML model. In this process, the new version of the trained ML entity supports the same type of inference as the previous version of the ML entity. That is, the data types of inference input and inference output remain the same between the two versions of the ML entity, but the parameter values ​​of the retrained model may be different.

[0042] ML federated training: ML training is performed on a set of ML models that are trained and targeted for inference.

[0043] ML Training: Refers to the end-to-end process that enables the ML training function to perform initial training or retraining of the ML model (as described above). ML training may include interactions with other parties to collect and format the data required for ML model training.

[0044] ML training function: A logical function with the ability to train ML models.

[0045] AI / ML inference: Refers to the process of running a set of input data through a trained ML entity to generate a set of output data (e.g., predictions).

[0046] AI / ML reasoning function: logical functions that use ML models for reasoning.

[0047] In this embodiment, a communication method is provided for AI / ML reasoning. Figure 3 This is the process of the communication method of the embodiment of the present invention Figure 1 ,like Figure 3 As shown, the process includes the following steps:

[0048] Step S302: The second management device receives an inference request from the first management device.

[0049] In the embodiment of the present invention, the first management device is used to correspond to the AI / ML reasoning consumer in AI / ML reasoning, and the second management device is used to correspond to the AI / ML reasoning producer in AI / ML reasoning. In the embodiment of the present invention, the management service consumers are ML training consumers and AI / ML reasoning consumers, and the management service producers are ML training producers and AI / ML reasoning producers. The above-mentioned management service consumers or management service producers can be Figure 2 Cross-domain management functional unit, domain management functional unit, network element, etc.

[0050] In an exemplary embodiment, the inference request includes at least one of the following: an inference category; inference task data; inference task requirements; inference task strategy; and an ML entity used in the inference request.

[0051] In an embodiment of the present invention, the inference task data indicates the inference data used for this inference task, which may be the address of the data used for this inference task, or the data category and performance indicators that need to be collected, such as network performance management (PM), key performance indicators (KPI), etc. The inference task requirements indicate the requirements of this inference task, which may be specific performance requirements (a specific indicator such as accuracy needs to be less than a fixed value), energy consumption requirements (the energy consumption of this training needs to be less than a fixed value), time requirements (inference task execution time), and iteration requirements (number of iterations). The inference task strategy includes one or more of the following: an inference strategy based on energy consumption (minimum energy consumption); an inference strategy based on time (shortest inference time); a training strategy based on inference performance (optimal performance); and a training strategy based on cost (minimum cost). Among them, when the inference task strategy indicates the strategy for collaborative inference of multiple ML entities, it may include the order of inference of ML entities.

[0052] In an embodiment of the present invention, the ML entity used in the above-mentioned reasoning request may be one ML entity, or multiple ML entities, or a collaborative reasoning set of ML entities.

[0053] Step S304: The second management device performs AI / ML reasoning according to the reasoning request.

[0054] In an exemplary embodiment, before the second management device performs AI / ML reasoning according to the reasoning request, the method further includes: the second management device creates an inference instance according to the reasoning request; and the second management device sends an inference instance creation success message to the first management device. Figure 4 This is the process of the communication method of the embodiment of the present invention Figure 2 ,like Figure 4 As shown, the process includes the following steps:

[0055] Step S402: The second management device receives an inference request from the first management device.

[0056] Step S404: The second management device creates an inference instance according to the inference request.

[0057] Step S406: The second management device sends a message indicating that the inference instance is created successfully to the first management device.

[0058] Step S408: The second management device performs AI / ML reasoning according to the reasoning request.

[0059] In an exemplary embodiment, the second management device performs AI / ML reasoning according to the reasoning request, including: the second management device determines the reasoning scheme according to the reasoning request; the second management device sends the reasoning scheme to the first management device or automatically executes the reasoning scheme.

[0060] In one embodiment, when the second management device sends an inference plan to the first management device, the second management device needs to receive feedback confirmation information from the first management device before executing the inference plan.

[0061] In an embodiment of the present invention, the second management device determines an inference scheme based on the inference category and inference requirements in the inference request, wherein the inference scheme includes selecting one or more appropriate ML entities or ML entity sets, i.e., ML entity collaborative inference sets, from the ML entity repository.

[0062] In an exemplary embodiment, the inference scheme includes at least one of the following: ML entities used by the inference task; inference task strategy; collaborative inference strategy; inference task execution time; inference task execution node; inference task resource; and inference task environment.

[0063] In one embodiment, the collaborative reasoning strategy can indicate how to perform reasoning when multiple models participate in the reasoning process, for example, the order of reasoning of each model, and the input-output relationship of the model; for example, the output of model A serves as the input of model B and model C, and the outputs of model B and model C serve as the input of model D at the same time, and so on.

[0064] In an exemplary embodiment, before the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: when the reasoning request does not include an ML entity, the second management device selects an ML entity or multiple ML entities or a collaborative reasoning set of ML entities in the ML entity repository according to the reasoning request.

[0065] In an exemplary embodiment, before the second management device selects an ML entity, multiple ML entities, or a collaborative reasoning set of ML entities in the ML entity repository according to the reasoning request, the method further includes: the second management device requests ML entity capability information from the ML entity repository; and the second management device receives the ML entity capability information from the ML entity repository.

[0066] In the embodiment of the present invention, the inference request may include an inference category but not an ML entity. The inference request may also include an inference category and an ML entity. Figure 5 This is the process of the communication method of the embodiment of the present invention Figure 3 ,like Figure 5 As shown, the process includes the following steps:

[0067] Step S502: The second management device receives an inference request from the first management device.

[0068] In this embodiment of the present invention, after the AI / ML inference producer queries the ML entity repository for ML entity capability information, if the ML entity repository contains an ML entity that meets the specified inference category and inference requirements, the AI / ML inference producer can proceed with the subsequent inference process. If the ML entity repository does not contain an ML entity that meets the specified inference category and inference requirements, the AI / ML inference producer must initiate ML training, namely, ML entity training, and obtain the ML entity before proceeding with the subsequent inference process.

[0069] Step S504: When the inference request does not include the ML entity, the second management device requests the ML entity capability information from the ML entity repository.

[0070] In an exemplary embodiment, the ML entity capability information includes at least one of the following: ML entity number; ML entity version; reasoning category; ML entity training context; ML entity reasoning context; ML entity performance; ML entity energy consumption information; ML entity complexity information; ML entity interoperability information; reasoning input; reasoning output; ML entity collaborative reasoning set association.

[0071] In this embodiment of the present invention, the ML entity inference context is used to indicate the ML entity inference history or to evaluate inference performance. ML entity energy consumption information is used to indicate the energy consumption required to run the ML entity or to indicate the energy consumption during ML entity inference. ML entity complexity information is used to indicate the complexity of the ML entity, and ML entity interoperability information is used to indicate the information required for ML entities to interoperate. Inference input is the data used for inference. Inference output is the output result of model inference.

[0072] Step S506: The second management device receives ML entity capability information from the ML entity warehouse.

[0073] Step S508 : The second management device selects an ML entity or multiple ML entities or a collaborative reasoning set of ML entities in the ML entity repository according to the reasoning request.

[0074] Step S510: The second management device performs AI / ML reasoning according to the reasoning request.

[0075] In an embodiment of the present invention, the second management device needs to start reasoning, i.e., AI / ML reasoning, based on network resource conditions.

[0076] In an exemplary embodiment, before the second management device performs the AI / ML inference based on the inference request, the method further includes: if neither the inference request nor the ML entity repository includes an ML entity, the second management device performs ML entity training based on the inference category to obtain the ML entity. In this embodiment of the present invention, the ML entity repository not including the ML entity means that the ML entity repository does not have any ML entities that are suitable for the current inference category and inference task requirements. Figure 6 This is the process of the communication method of the embodiment of the present invention Figure 4 ,like Figure 6 As shown, the process includes the following steps:

[0077] Step S602: The second management device receives an inference request from the first management device.

[0078] Step S604: When the inference request does not include the ML entity, the second management device requests the ML entity capability information from the ML entity repository.

[0079] Step S606: The second management device receives ML entity capability information from the ML entity warehouse.

[0080] Step S608 : When neither the inference request nor the ML entity repository includes the ML entity, the second management device performs ML entity training according to the inference category to obtain the ML entity.

[0081] In the embodiment of the present invention, the ML entity obtained through training, i.e., the model, will be used for this reasoning task. The ML entity may be one ML entity, multiple ML entities, or a set of ML entities for collaborative reasoning.

[0082] Step S610: The second management device performs AI / ML reasoning according to the reasoning request.

[0083] In an embodiment of the present invention, if there is no ML entity suitable for the reasoning category and reasoning requirement in the ML entity warehouse, and ML entity training is not started, the AI / ML reasoning producer reports to the AI / ML reasoning consumer that there is no suitable ML entity.

[0084] In an exemplary embodiment, after the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: the second management device sending an AI / ML reasoning report to the first management device.

[0085] Figure 7 This is the process of the communication method of the embodiment of the present invention Figure 5 ,like Figure 7 As shown, the process includes the following steps:

[0086] Step S702: The second management device receives an inference request from the first management device.

[0087] Step S704: The second management device performs AI / ML reasoning according to the reasoning request.

[0088] Step S706: The second management device sends an AI / ML inference report to the first management device.

[0089] In an exemplary embodiment, the AI / ML reasoning report includes at least one of the following: AI / ML reasoning results; AI / ML reasoning demand report; AI / ML reasoning context; objects affected by executing AI / ML reasoning results; and reasoning data usage.

[0090] In an embodiment of the present invention, an AI / ML inference requirement report may be used to indicate the fulfillment of the requirements indicated in the inference requirements of an inference request. The type of objects affected by the execution of the AI / ML inference result includes at least one of the following: a base station; a core network element; a user cell; a network slice; or a user identity.

[0091] In this embodiment of the present invention, if the inference requirement includes energy consumption requirements or the inference strategy is based on energy consumption, the inference requirement report must include the energy consumption of this inference. Similarly, if the inference requirement includes performance requirements, the inference requirement report must include the performance indicators of this inference. Inference data usage indicates whether data from non-inference requests is used. Inference context includes background information about this inference request, such as time, conditions, etc., as well as the reason for using inference data indicated in non-inference requests and the data address.

[0092] In this embodiment of the present invention, the inference data usage can be used to indicate whether data in non-inference requests is used. The inference context is used to indicate the background information of the inference request, such as the time and conditions, as well as the reason for the use of the inference data indicated in the non-inference request and the data address.

[0093] In an embodiment of the present invention, after the AI / ML inference producer, i.e., the second management device, sends an AI / ML inference report to the AI / ML inference consumer, i.e., the first management device, it indicates that the inference has ended. It is necessary to judge whether the inference performance meets the requirements based on the inference report. If the requirements are met, the inference process ends here.

[0094] In an exemplary embodiment, after the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: when the reasoning performance of the AI / ML reasoning does not meet the reasoning requirements, the second management device re-trains the ML entity and re-performs the AI / ML reasoning based on the obtained new ML entity.

[0095] The embodiment of the present invention also provides a communication method for AI / ML reasoning. Figure 8 This is the process of the communication method of the embodiment of the present invention Figure 6 ,like Figure 8 As shown, the process includes the following steps:

[0096] Step S802: The first management device sends an inference request to the second management device to instruct the second management device to perform AI / ML inference according to the inference request.

[0097] In an exemplary embodiment, after the first management device sends an inference request to the second management device, the method further includes: the first management device receives an AI / ML inference report from the second management device, wherein the AI / ML inference report includes an object affected by executing an inference result in the AI / ML inference report. Figure 9 This is the process of the communication method of the embodiment of the present invention Figure 7 ,like Figure 9 As shown, the process includes the following steps:

[0098] In step S902 , the first management device sends an inference request to the second management device to instruct the second management device to perform AI / ML inference according to the inference request.

[0099] Step S904: The first management device receives an AI / ML reasoning report from the second management device, wherein the AI / ML reasoning report includes objects affected by executing a reasoning result in the AI / ML reasoning report.

[0100] In an exemplary embodiment, after the first management device receives the AI / ML reasoning report from the second management device, the method further includes: the first management device requests to execute the reasoning result and subscribes to entity status information after the execution of the reasoning result from the object affected by the execution of the reasoning result; the first management device evaluates the execution performance of the reasoning result based on the entity status information; and the first management device updates the ML entity reasoning context of the ML entity used for AI / ML reasoning based on the evaluation result. Figure 10 This is the process of the communication method of the embodiment of the present invention Figure 8 ,like Figure 10 As shown, the process includes the following steps:

[0101] Step S1002: The first management device sends an inference request to the second management device to instruct the second management device to perform AI / ML inference according to the inference request.

[0102] Step S1004: The first management device receives an AI / ML reasoning report from the second management device, wherein the AI / ML reasoning report includes objects affected by executing a reasoning result in the AI / ML reasoning report.

[0103] Step S1006 : The first management device requests to execute the inference result, and subscribes to entity state information after the execution of the inference result from the objects affected by the execution of the inference result.

[0104] In one embodiment, the above-mentioned entity status information may be entity performance information, for example, including PM, KPI, Quality of Experience (QOE), Minimization of Drive Test Report (MDT Report), configuration information, power, cell coverage, fault information, alarm information, and the like.

[0105] Step S1008: The first management device evaluates the execution performance of the inference result according to the entity state information.

[0106] In one embodiment, the evaluation may be performed in the form of scoring, or in the form of grading, or in the form of feasibility or not, which is not limited here.

[0107] Step S1010: The first management device updates the ML entity reasoning context of the ML entity used for AI / ML reasoning according to the evaluation result.

[0108] Through the above steps, a communication method for AI / ML inference is provided, in which a second management device receives an inference request from a first management device; the second management device then performs AI / ML inference based on the inference request. This method addresses the problems of the related art, which lack clear inference requests and lack evaluation and feedback of AI / ML inference results. It achieves the effect of performing AI / ML inference based on clear inference requests and providing feedback and adjustments to ML entities based on the evaluation of the AI / ML inference results, thereby improving communication performance.

[0109] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0110] In this embodiment, a communication device is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments. Details already described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0111] The communication device provided in an embodiment of the present invention is used for AI / ML reasoning and can be provided on a second management device. The communication device includes a receiving module for receiving a reasoning request from a first management device and an execution module for performing AI / ML reasoning based on the reasoning request.

[0112] The communication device provided in an embodiment of the present invention is used for AI / ML reasoning and can be set in a first management device, including a sending module for sending an inference request to a second management device to instruct the second management device to perform AI / ML reasoning according to the inference request.

[0113] It should be noted that each of the above modules can be implemented through software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or, the above modules are located in different processors in any combination. In the actual implementation process, the module naming method and module function division of the above communication device are only examples and are not restrictive. In the actual implementation process, the module naming method and module function division can be set according to actual conditions, as long as the communication method in the above embodiment can be implemented. Similarly, the communication device is not limited to being set on the first management device or the second management device, and can be adjusted according to actual conditions.

[0114] An embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored. The computer program is configured to execute the steps of any one of the above method embodiments when running.

[0115] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk, or an optical disk.

[0116] An embodiment of the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0117] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0118] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail here.

[0119] Obviously, those skilled in the art will appreciate that the various modules or steps of the present invention described above can be implemented using a general-purpose computing device, can be centralized on a single computing device, or can be distributed across a network of multiple computing devices. They can be implemented using program code executable by the computing device, and thus, can be stored in a storage device and executed by the computing device. In some cases, the steps shown or described herein can be performed in a different order than that shown, or can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0120] In order to enable those skilled in the art to better understand the technical solution of the present invention, it is described below in conjunction with scenario embodiments.

[0121] Example 1

[0122] In Example 1, the AI / ML operation workflow in the related technology is first introduced. Figure 11 This is a schematic diagram of the AI / ML operational workflow in related technologies, such as Figure 11 As shown in Figure 1, the AI / ML operation workflow mainly includes four stages: training stage, simulation stage, deployment stage, and inference stage.

[0123] Training phase: includes ML training and ML testing. ML training involves the training of one or a group of ML models, including initial training and retraining. It also includes the verification of ML entities to evaluate the performance of ML entities when executed on training data and verification data. If the verification results do not meet expectations (for example, the variance is unacceptable), the ML model associated with the entity needs to be retrained. ML model training is the initial phase of the workflow. ML testing tests the verified ML entity to evaluate the performance of the trained ML model when executed on test data. If the test results meet expectations, the ML entity can enter the next phase, otherwise the ML model associated with the entity may need to be retrained.

[0124] Simulation phase: ML simulation runs ML entities in a simulation environment for inference. The goal is to evaluate the inference performance of ML entities in the simulation environment before applying them to the target network or system. The simulation phase is optional and can be skipped in the AI / ML operational workflow.

[0125] Deployment phase: ML entity loading, the process of making trained ML entities available to target AI / ML inference functions (also known as atomic operation serialization). The deployment phase may not be required in some cases, such as when the training function and inference function are co-located.

[0126] Inference phase: AI / ML inference, which uses the trained ML entities to perform inference through AI / ML inference capabilities.

[0127] In an embodiment of the present invention, the parameters involved in the above-mentioned AI / ML operation workflow include model complexity parameters and model performance parameters.

[0128] Model complexity parameters include the number of model parameters and the amount of computation required. The number of model parameters refers to the number of learnable parameters in a machine learning model. These parameters are typically weights and biases, which represent the model's features and relationships. The number of model parameters directly affects the model's complexity and capacity, and is generally closely related to the model's expressiveness and performance. A larger number of parameters may mean a more complex model, but it may also increase the risk of overfitting.

[0129] Model computational complexity refers to the computing resources required by a model when performing inference or training. Model computational complexity is typically measured using metrics such as floating point operations (FLOPs), multiply-accumulate operations (MACs), and addition operations (M Adds). FLOPs, MACs, and M Adds are metrics used to quantify the computational complexity of an algorithm, particularly in the fields of machine learning and signal processing. Here's what each term means:

[0130] FLOPs represents the number of floating-point operations (such as addition, subtraction, multiplication, and division) a computing device performs in one second. It measures the raw computing power of a system. FLOPs can also be used to measure the computational complexity of an algorithm, indicating the number of floating-point operations required to perform a specific task.

[0131] MACs specifically refer to the number of multiply-accumulate operations performed by an algorithm or hardware component. In many machine learning models, such as convolutional neural networks (CNNs), the majority of computational work comes from MAC operations, which involve multiplying two numbers together and then adding the result to an accumulator.

[0132] M Adds is similar to MACs, but counts the number of multiplication and addition operations separately. Therefore, for each MAC operation, two operations are calculated (one multiplication and one addition, respectively). Compared to MACs, M Adds provides a more detailed breakdown of computational complexity. These metrics are important for understanding and comparing the efficiency of different algorithms, architectures, and hardware platforms, and help optimize algorithms in resource-constrained environments (such as mobile devices or edge computing devices) to improve speed and resource consumption efficiency.

[0133] Model performance parameters include: accuracy, precision, recall, F1 score, mean squared error (MSE), mean absolute error (MAE), and root mean square error (RMSE).

[0134] In the first embodiment, the communication method flow for AI / ML reasoning is described based on the case where the reasoning category is specified in the reasoning request but no ML entity is included. As described above, in this embodiment of the present invention, the AI / ML reasoning consumer corresponds to the first management device in the above embodiment, and the AI / ML reasoning producer corresponds to the second management device in the above embodiment. Figure 12 is a flow chart of a communication method according to a first embodiment of the present invention, Figure 12 As shown, the following steps are included:

[0135] In step S1201 , the AI / ML inference consumer sends an inference request to the AI / ML inference producer, requesting the creation of an AI / ML inference instance.

[0136] Among them, the inference request includes inference category, inference task data, and inference task requirements. Among them, the inference category (Inference Type) indicates the ML entity inference type, which can be MDA Type, Analytic ID, etc. Inference data indicates the inference data used for this inference task, and is not limited to the address of the inference data. Inference requirements indicate the requirements of this inference task, which can be specific performance requirements, energy consumption requirements or inference strategies, among which performance requirements indicate the model performance parameters that need to be reported. Energy consumption requirements indicate the energy consumption of this inference task, such as less than 10kWh. Inference task strategy indicates the solution scheduling strategy for this inference task, such as energy consumption-based strategy, optimal performance-based strategy, shortest inference time-based strategy, etc.

[0137] In step S1202 , the AI / ML inference producer creates an AI / ML inference instance based on the received inference request.

[0138] In step S1203 , the AI / ML inference producer notifies the AI / ML inference consumer that the AI / ML inference instance is created successfully.

[0139] In step S1204 , the AI / ML inference producer queries the ML entity repository for ML entity capability information.

[0140] In step S1205 , the ML entity repository feeds back ML entity capability information to the AI / ML inference producer.

[0141] The ML entity capability information includes the content shown in Table 1.

[0142] Table 1 Example table of ML entity capability information

[0143]

[0144] In this embodiment of the present invention, after the AI / ML inference producer queries the ML entity repository for ML entity capability information, if the ML entity repository contains an ML entity that meets the specified inference category and inference requirements, the AI / ML inference producer can proceed with the subsequent inference process. If the ML entity repository does not contain an ML entity that meets the specified inference category and inference requirements, the AI / ML inference producer must initiate ML training, namely, ML entity training, and obtain the ML entity before proceeding with the subsequent inference process.

[0145] First, for the case where there are ML entities in the ML entity warehouse that are suitable for the reasoning category and reasoning requirements:

[0146] In step S1206 , the AI / ML inference producer determines an inference solution based on the inference category and inference requirements in the inference request.

[0147] The inference scheme includes selecting one or more appropriate ML entities or a set of ML entities from the ML entity repository.

[0148] In step S1207 , the AI / ML inference producer informs the AI / ML inference consumer of the inference plan, including one or more ML entities or a set of ML entities selected for this inference, i.e., the ML entity collaborative inference set.

[0149] In step S1208, the AI / ML inference producer starts inference based on network resource conditions.

[0150] In step S1209 , the AI / ML inference producer sends an AI / ML inference report to the AI / ML inference consumer.

[0151] Among them, the AI / ML reasoning report includes reasoning results, reasoning requirements report, and reasoning context. Among them, the reasoning result is determined by the reasoning type. The reasoning requirements report is used to indicate the implementation of the requirements indicated in the reasoning requirements of the reasoning request, such as performance requirements report, energy consumption requirements report, etc. When the reasoning requirements include energy consumption requirements or the reasoning strategy of the reasoning requirements is an energy consumption-based strategy, the reasoning requirements report must include the energy consumption of this reasoning; similarly, when the reasoning requirements include performance requirements, the reasoning requirements report must include the performance indicators of this reasoning. Inference data usage, indicating whether data in non-reasoning requests is used. Reasoning context, background information of this reasoning request, such as time, conditions, etc., the reason why the inference data indicated in the non-reasoning request is used, data address, etc.

[0152] Afterwards, if there is no ML entity in the ML entity repository that is suitable for the inference category and inference requirement:

[0153] In step S1210 , the AI / ML inference producer automatically triggers ML entity training for the inference category, and the trained model will be used for this inference task.

[0154] In step S1211 , the AI / ML inference producer informs the AI / ML inference consumer of one or more ML entities or a set of ML entities selected for this inference.

[0155] In step S1212 , the AI / ML inference producer restarts the AI / ML inference based on the trained ML entity.

[0156] In step S1213, the AI / ML inference producer sends an AI / ML inference report to the AI / ML inference consumer. The inference report includes the inference results, the inference requirement report, and the inference context. The inference context indicates that the ML entity used this time is obtained from ML training rather than an existing model.

[0157] In an embodiment of the present invention, if there is no ML entity suitable for the reasoning category and reasoning requirement in the ML entity warehouse, and ML entity training is not started, in step S1214, the AI / ML reasoning producer reports to the AI / ML reasoning consumer that there is no suitable ML entity.

[0158] Example 2

[0159] In the second embodiment, based on the case where the inference category is specified in the inference request, including the case where the ML entity is included, the process of the communication method for AI / ML inference is introduced. Figure 13 is a flow chart of a communication method according to a second embodiment of the present invention, Figure 13 As shown, the following steps are included:

[0160] In step S1301 , the AI / ML inference consumer sends an inference request to the AI / ML inference producer, requesting the creation of an AI / ML inference instance.

[0161] The inference request includes the inference category, the ML entity used, the inference data, and the inference requirement. The ML entity used indicates the ML entity used in this inference request, which can be one or more ML entities or an ML entity coordination group (multiple ML entities used for joint inference).

[0162] In step S1302 , the AI / ML inference producer creates an AI / ML inference instance based on the received AI / ML inference request.

[0163] Step S1303 : The AI / ML inference producer notifies the AI / ML inference consumer that the AI / ML inference instance is created successfully.

[0164] Step S1304: The AI / ML inference producer starts inference.

[0165] This also includes the AI / ML inference producer determining the inference plan based on the inference requirements in the inference request, and the inference plan includes the ML entity used in the request.

[0166] In step S1305 , the AI / ML inference producer sends an AI / ML inference report to the AI / ML inference consumer.

[0167] Among them, the AI / ML reasoning report includes reasoning results, reasoning requirements report, and reasoning context.

[0168] In an embodiment of the present invention, after the AI / ML inference producer sends the AI / ML inference report to the AI / ML inference consumer in step S1305, it indicates that the inference has ended. It is necessary to determine whether the inference performance meets the requirements based on the inference report. If the requirements are met, the inference process ends here.

[0169] If the inference performance does not meet the requirements, the following steps are also included:

[0170] In step S1306 , the AI / ML inference producer automatically requests the ML entity to retrain.

[0171] In step S1307 , the AI / ML inference producer restarts the AI / ML inference based on the retrained ML entity.

[0172] In step S1308, the AI / ML inference producer sends an AI / ML inference report to the AI / ML inference consumer. The AI / ML inference report includes the inference result, inference requirement report, and inference context. The inference context indicates that the ML entity used this time is obtained through retraining.

[0173] Example 3

[0174] In the AI / ML management process of related technologies, the trained model can be simulated through AI / ML to test the model's reasoning performance, but the effect of the AI / ML reasoning results after application and the impact on the object (i.e., the managed entity) also need to be collected to evaluate the actual performance of the reasoning results.

[0175] Therefore, in the third embodiment, the process of evaluating and providing feedback on the inference results is introduced.

[0176] Figure 14 Flowchart of the communication method of the third embodiment of the present invention. Figure 14 As shown, the following steps are included:

[0177] Step S1401: The AI / ML inference consumer sends an inference request to the AI / ML inference producer.

[0178] In step S1402 , the AI / ML inference producer sends an inference report to the AI / ML inference consumer.

[0179] The reasoning report includes objects affected by the execution of the reasoning result.

[0180] Step S1403 : The AI / ML inference consumer requests to execute the inference result.

[0181] In step S1404, the AI / ML inference consumer subscribes to the affected objects to report the status after the inference results are executed, including but not limited to PM, KPI, etc.

[0182] In an embodiment of the present invention, the types of objects among the objects affected by the execution of AI / ML reasoning results include at least one of the following: base station; core network element; user cell; network slice; user identity.

[0183] Step S1405: The affected object responds to the request and provides feedback status information.

[0184] In step S1406, the AI / ML inference consumer evaluates the performance of the AIML inference result execution based on the received feedback and scores the model performance.

[0185] In the embodiment of the present invention, the evaluation may be performed in a scoring manner, a grading manner, or a feasibility manner, which is not limited here.

[0186] In step S1407, the AI / ML inference consumer updates the involved ML entities based on the evaluation results, for example, updating the model performance score of the ML entity.

[0187] In summary, embodiments of the present invention provide a communication method for AI / ML inference. For management service consumers, an AI / ML inference request indicates the ML entity to be used or the inference category. AI / ML inference requirements are included in the AI / ML inference request. Training requirements include at least one of the following: AI / ML inference strategy, priority information, and energy consumption requirements. The execution performance of the inference results is evaluated and scored, and the relevant information is synchronized to the ML entity repository. For management service producers, when an ML entity is specified in the AI / ML inference request, the AI / ML inference producer orchestrates and generates an AI / ML inference plan based on the specified ML entity. When only the inference category is specified in the AI / ML inference request, the AI / ML inference producer queries the ML entity repository for ML entity capability information to determine a suitable ML entity. If no suitable ML entity exists for the current inference task, the AI / ML inference producer can request ML training to obtain an ML entity suitable for the current inference task. If the performance of the existing ML entity fails to meet the inference requirements, the AI / ML inference producer can request ML retraining to obtain an ML entity suitable for the current inference task. The AI / ML inference producer sends an AIML inference report to the consumer, including the inference result, inference requirement report, inference context, and objects affected by the inference result. For ML entities, in this embodiment of the present invention, a model performance score is added to each trained ML entity to indicate the performance evaluation result after the inference result is executed.

[0188] The communication method provided by the embodiments of the present invention utilizes an AI / ML inference producer to receive inference requests from an AI / ML inference consumer; the AI / ML inference producer then performs AI / ML inference based on the inference request. This method addresses the problems of related technologies, such as the lack of clear inference requests and the lack of AI / ML inference result evaluation and feedback. It achieves the goal of performing AI / ML inference based on clear inference requests and providing feedback adjustments to ML entities based on the evaluation of the AI / ML inference results, thereby improving communication performance.

[0189] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A communication method, characterized in that: For AI / ML inference, including: The second management device receives the inference request from the first management device; The second management device performs the AI / ML reasoning according to the reasoning request.

2. The method according to claim 1, characterized in that in, The inference request includes at least one of the following: Reasoning category; reasoning task data; reasoning task requirements; reasoning task strategy; machine learning (ML) entity used by the reasoning request.

3. The method according to claim 1, characterized in that Before the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: The second management device creates an inference instance according to the inference request; The second management device sends a message indicating that the inference instance was created successfully to the first management device.

4. The method according to claim 1, wherein The second management device performs the AI / ML reasoning according to the reasoning request, including: The second management device determines an inference scheme according to the inference request; The second management device sends the inference solution to the first management device or automatically executes the inference solution.

5. The method according to claim 4, characterized in that in, The reasoning scheme includes at least one of the following: ML entities used by reasoning tasks; reasoning task strategies; collaborative reasoning strategies; reasoning task execution time; reasoning task execution nodes; reasoning task resources; and reasoning task environment.

6. The method according to claim 2, characterized in that Before the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: In a case where the inference request does not include the ML entity, the second management device selects an ML entity or multiple ML entities or a collaborative inference set of ML entities in the ML entity repository according to the inference request.

7. The method according to claim 6, characterized in that Before the second management device selects an ML entity, a plurality of ML entities, or a collaborative reasoning set of ML entities in the ML entity repository, the method further includes: The second management device requests ML entity capability information from the ML entity repository; The second management device receives the ML entity capability information from the ML entity repository.

8. The method according to claim 7, characterized in that in, The ML entity capability information includes at least one of the following: ML entity number; ML entity version; the reasoning category; ML entity training context; ML entity reasoning context; ML entity performance; ML entity energy consumption information; ML entity complexity information; ML entity interoperability information; reasoning input; reasoning output; ML entity collaborative reasoning set association.

9. The method according to claim 2, characterized in that Before the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: In a case where neither the inference request nor the ML entity repository includes the ML entity, the second management device performs ML entity training according to the inference category to obtain the ML entity.

10. The method according to claim 1, characterized in that After the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: The second management device sends an AI / ML inference report to the first management device.

11. The method according to claim 10, characterized in that in, The AI / ML reasoning report includes at least one of the following: AI / ML inference results; AI / ML inference demand report; AI / ML inference context; objects affected by executing AI / ML inference results; inference data usage.

12. The method according to claim 1, characterized in that After the second management device performs the AI / ML reasoning according to the reasoning request, the method further includes: When the reasoning performance of the AI / ML reasoning does not meet the reasoning requirements, the second management device re-trains the ML entity and re-performs the AI / ML reasoning based on the acquired new ML entity or multiple ML entities or ML entity collaborative reasoning set.

13. A communication method, characterized in that: For AI / ML inference, including: The first management device sends an inference request to the second management device to instruct the second management device to perform the AI / ML inference according to the inference request.

14. The method according to claim 13, characterized in that After the first management device sends the inference request to the second management device, the method further includes: The first management device receives an AI / ML reasoning report from the second management device, wherein the AI / ML reasoning report includes an object affected by executing a reasoning result in the AI / ML reasoning report.

15. The method according to claim 14, characterized in that After the first management device receives the AI / ML inference report from the second management device, the method further includes: The first management device requests execution of the inference result, and subscribes to entity state information after execution of the inference result from an object affected by execution of the inference result; The first management device evaluates the execution performance of the inference result according to the entity state information; The first management device updates the ML entity reasoning context of the ML entity used by the AI / ML reasoning according to the evaluation result.

16. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program implements the method described in any one of claims 1 to 15 when executed by a processor.

17. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 15 is implemented.

Images