WO2023125934A1

WO2023125934A1 - Ai network information transmission method and apparatus, and communication device

Info

Publication number: WO2023125934A1
Application number: PCT/CN2022/143952
Authority: WO
Inventors: 任千尧; 孙鹏
Original assignee: 维沃移动通信有限公司
Priority date: 2021-12-31
Filing date: 2022-12-30
Publication date: 2023-07-06
Also published as: CN116418880A

Abstract

The present application belongs to the technical field of communications. Disclosed are an AI network information transmission method and apparatus, and a communication device. The AI network information transmission method in the embodiments of the present application comprises: a first terminal compressing AI network information, wherein the AI network information comprises at least one of a network structure and a network parameter; and the first terminal sending the compressed AI network information to a second terminal.

Description

AI network information transmission method, device and communication equipment

Cross References to Related Applications

This application claims priority to Chinese Patent Application No. 202111666710.4 filed in China on December 31, 2021, the entire contents of which are hereby incorporated by reference.

technical field

The present application belongs to the technical field of communication, and in particular relates to an AI network information transmission method, device and communication equipment.

Background technique

Artificial Intelligence (AI) is a new technical science that researches and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. more and more widely. At present, people have begun to study the application of AI networks in communication systems. For example, communication data can be transmitted between network-side devices and terminals through AI networks. In a communication system, the entire AI network is usually transmitted together, resulting in a large system overhead.

Contents of the invention

Embodiments of the present application provide an AI network information transmission method, device, and communication equipment, which can solve the problem of relatively large transmission overhead of communication equipment in AI network transmission in the related art.

In the first aspect, an AI network information transmission method is provided, including:

The first end compresses the AI network information, and the AI network information includes at least one of network structure and network parameters;

The first end sends the compressed AI network information to the second end.

In the second aspect, an AI network information transmission method is provided, including:

The second end receives the compressed AI network information sent by the first end, where the AI network information includes at least one of network structure and network parameters.

In a third aspect, an AI network information transmission device is provided, including:

A compression module, configured to compress AI network information, where the AI network information includes at least one of network structure and network parameters;

A sending module, configured to send the compressed AI network information to the second end.

In the fourth aspect, an AI network information transmission device is provided, including:

The receiving module is configured to receive the compressed AI network information sent by the first end, where the AI network information includes at least one of network structure and network parameters.

In a fifth aspect, a communication device is provided, including a processor and a memory, the memory stores programs or instructions that can run on the processor, and when the programs or instructions are executed by the processor, the first The steps of the AI network information transmission method described in the aspect, or the steps of implementing the AI network information transmission method described in the second aspect.

A sixth aspect provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the AI network information transmission method as described in the first aspect are implemented , or implement the steps of the AI network information transmission method as described in the second aspect.

In the seventh aspect, there is provided a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the AI described in the first aspect A network information transmission method, or realize the AI network information transmission method as described in the second aspect.

In an eighth aspect, a computer program/program product is provided, the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement the method described in the first aspect The steps of the AI network information transmission method, or the steps of realizing the AI network information transmission method as described in the second aspect.

In the embodiment of the present application, the first end can send compressed AI network information to the second end, the AI network information includes at least one of the network structure and network parameters, and there is no need to include all The network structure and network parameters of the entire AI network are transmitted together, so that the network structure and network parameters of the AI network can be sent separately, which can effectively reduce the transmission overhead in the communication process.

Description of drawings

FIG. 1 is a block diagram of a wireless communication system to which an embodiment of the present application is applicable;

FIG. 2 is a flow chart of an AI network information transmission method provided by an embodiment of the present application;

Fig. 3 is a flow chart of another AI network information transmission method provided by the embodiment of the present application;

FIG. 4 is a structural diagram of an AI network information transmission device provided by an embodiment of the present application;

Fig. 5 is a structural diagram of another AI network information transmission device provided by the embodiment of the present application;

FIG. 6 is a structural diagram of a communication device provided by an embodiment of the present application;

FIG. 7 is a structural diagram of a terminal provided in an embodiment of the present application;

FIG. 8 is a structural diagram of a network-side device provided by an embodiment of the present application;

FIG. 9 is a structural diagram of another network-side device provided by an embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, but not all of them. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in this application belong to the protection scope of this application.

The terms "first", "second" and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific sequence or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments of the application are capable of operation in sequences other than those illustrated or described herein and that "first" and "second" distinguish objects. It is usually one category, and the number of objects is not limited. For example, there may be one or more first objects. In addition, "and/or" in the description and claims means at least one of the connected objects, and the character "/" generally means that the related objects are an "or" relationship.

It is worth pointing out that the technology described in the embodiment of this application is not limited to the Long Term Evolution (Long Term Evolution, LTE)/LTE-Advanced (LTE-Advanced, LTE-A) system, and can also be used in other wireless communication systems, such as code Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access, OFDMA), Single-carrier Frequency Division Multiple Access (Single-carrier Frequency Division Multiple Access, SC-FDMA) and other systems. The terms "system" and "network" in the embodiments of the present application are often used interchangeably, and the described technologies can be used for the above-mentioned systems and radio technologies as well as other systems and radio technologies. The following description describes the New Radio (NR) system for illustrative purposes, and uses NR terminology in most of the following descriptions, but these techniques can also be applied to applications other than NR system applications, such as the 6th generation (6 ^th Generation, 6G) communication system.

Fig. 1 shows a block diagram of a wireless communication system to which the embodiment of the present application is applicable. The wireless communication system includes a terminal 11 and a network side device 12 . Wherein, the terminal 11 can be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer) or a notebook computer, a personal digital assistant (Personal Digital Assistant, PDA), a palmtop computer, a netbook, a super mobile personal computer (ultra-mobile personal computer, UMPC), mobile Internet device (Mobile Internet Device, MID), augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) equipment, robot, wearable device (Wearable Device) , Vehicle User Equipment (VUE), Pedestrian User Equipment (PUE), smart home (home equipment with wireless communication functions, such as refrigerators, TVs, washing machines or furniture, etc.), game consoles, personal computers (personal computer, PC), teller machine or self-service machine and other terminal side devices, wearable devices include: smart watches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart feet bracelets, smart anklets, etc.), smart wristbands, smart clothing, etc. It should be noted that, the embodiment of the present application does not limit the specific type of the terminal 11 . The network side device 12 may include an access network device or a core network device, where the access network device may also be called a radio access network device, a radio access network (Radio Access Network, RAN), a radio access network function, or a wireless network. access network unit. The access network equipment may include a base station, a wireless local area network (Wireless Local Area Network, WLAN) access point, or a WiFi node, etc., and the base station may be called a node B, an evolved node B (Evolved Node B, eNB), an access point, or a base station. Base Transceiver Station (BTS), radio base station, radio transceiver, Basic Service Set (BSS), Extended Service Set (ESS), Home Node B, Home Evolved Node B, Transmitting Receiving Point (Transmitting Receiving Point, TRP) or some other suitable term in the field, as long as the same technical effect is achieved, the base station is not limited to specific technical terms. It should be noted that, in the embodiment of this application, only The base station in the NR system is taken as an example for introduction, and the specific type of the base station is not limited. The core network equipment may include but not limited to at least one of the following: core network node, core network function, mobility management entity (Mobility Management Entity, MME), access mobility management function (Access and Mobility Management Function, AMF), session management function (Session Management Function, SMF), User Plane Function (UPF), Policy Control Function (Policy Control Function, PCF), Policy and Charging Rules Function (PCRF), edge application service Discovery function (Edge Application Server Discovery Function, EASDF), unified data management (Unified Data Management, UDM), unified data storage (Unified Data Repository, UDR), home subscriber server (Home Subscriber Server, HSS), centralized network configuration ( Centralized network configuration, CNC), network storage function (Network Repository Function, NRF), network exposure function (Network Exposure Function, NEF), local NEF (Local NEF, or L-NEF), binding support function (Binding Support Function, BSF), application function (Application Function, AF), etc. It should be noted that, in the embodiment of the present application, only the core network equipment in the NR system is used as an example for introduction, and the specific type of the core network equipment is not limited.

The following describes in detail the AI network information transmission method provided by the embodiment of the present application through some embodiments and application scenarios with reference to the accompanying drawings.

Please refer to Figure 2, Figure 2 is a flow chart of an AI network information transmission method provided in the embodiment of the present application, as shown in Figure 2, the method includes the following steps:

Step 201, the first end compresses the AI network information, and the AI network information includes at least one of network structure and network parameters;

Step 202, the first end sends the compressed AI network information to the second end.

In the embodiment of the present application, the first end and the second end are communication devices with sending and receiving functions.

Optionally, the first end is one of the network-side device and the terminal, and the second end is the other of the network-side device and the terminal; or, the first end and the second end are different nodes of a terminal; or, the first end and the second end are different nodes of a network side device.

It should be noted that the network side equipment may include access network equipment (for example: base station) and core network equipment. Optionally, the first end may be an access network device, and the second end may be a core network device; or, the first end may be a terminal, and the second end may be a core network device or an access network device; or, the second end may be a core network device or an access network device; The one end and the second end are different nodes of the access network equipment; or, the first end and the second end are different nodes of the core network equipment, etc., and the embodiments of the present application do not list them one by one .

In this embodiment of the present application, the AI network information includes at least one of network structure and network parameters. Optionally, the AI network information may be the network structure and/or network parameters of a certain AI network, or the network structures and/or network parameters of multiple AI networks. In some embodiments, an AI network may also be called an AI neural network, an AI model, or the like. Wherein, the network parameters include weight parameters, hyperparameters and the like of the AI network.

It should be noted that the compressing the AI network information may refer to compressing the AI network information into a file corresponding to the preset model expression method according to the preset model expression method. The so-called model expression method is a kind of data Structure, which describes the AI network structure, network parameters and other information according to certain rules.

Optionally, the AI network information includes network structure and/or network parameters, and the compression of the AI network information by the first end also includes compressing the network structure and/or weight parameters, and compressing the compressed network structure and/or or the weight parameter is sent to the second end. For example, if the AI network information only includes the network structure, the first end only compresses and sends the network structure; or, the AI network information may only include network parameters, then the first end only compresses and sends the network parameters; or , the AI network information includes part of the network structure and part of the network parameters, and the first end compresses and sends the part of the network structure and part of the network parameters. Of course, the specific information content included in the AI network information may also be in other situations, which will not be described in detail here.

In the embodiment of the present application, the AI network information includes at least one of the network structure and network parameters, and there is no need to transmit the entire AI network including all network structures and network parameters together during the communication process, so that the network of the AI network The structure and network parameters can be sent separately, which can effectively reduce the transmission overhead in the communication process.

Optionally, the AI network information may include network structure and network parameters, and the step 201 may include any of the following:

The first end combines and compresses the network structure and the network parameters;

The first end compresses the network structure and the network parameters respectively.

For example, the first end can combine and compress the network structure and network parameters of the AI network into a transmission file based on a preset model expression, or can also compress the network structure and weights based on a preset model expression. The parameters are compressed separately into two separate transfer files.

Optionally, in the case where the first end compresses the network structure and the network parameters respectively, the step 202 may include any of the following:

The first end sends the compressed network structure and compressed network parameters to the second end in combination;

The first end sends the compressed network structure and the compressed network parameters to the second end respectively.

Exemplarily, in the case that the AI network information includes network structure and network parameters, the first end compresses the network structure and network parameters respectively, such as compressing into two independent transmission files, and compresses the two compressed The transmission files are sent to the second end together, or the two compressed transmission files can be sent separately, or only one of the compressed transmission files can be sent, so that the first end can receive the compressed AI network information The transmission method is more flexible.

Optionally, the step 201 may also include any of the following:

The first end converts the AI network information into a corresponding transmission file based on a preset model representation, and compresses the transmission file;

The first end compresses the AI network information based on a preset data format;

The first end obtains the AI network information to be sent and the existing AI network information of the second end, and calculates the AI network information between the AI network information to be sent and the existing AI network information of the second end. Network information difference is compressed;

The first end obtains the AI network information to be sent and the AI network information of the preset AI network, and calculates the AI network information difference between the AI network information to be sent and the AI network information of the preset AI network to compress.

Wherein, the preset model expression method may be an AI network expression method common to both the first end and the second end, such as open neural network exchange (ONNX), TensorFlow, and the like. Among them, TensorFlow is a machine learning framework. Exemplarily, the first end can convert the AI network information into a corresponding transmission file based on ONNX, and then compress the transmission file and send it to the second end. After the second end decompresses the compressed transmission file, it can also Based on ONNX, the decompressed transmission file is converted into its own applicable network structure and network parameters.

It should be noted that the file structure of the AI network information saved under the two different neural network frameworks is different and cannot be read directly. The AI network information is converted into the corresponding transmission file through the preset model expression method and then compressed and transmitted. It also enables two communication devices using different neural network frameworks to realize the reading and application of AI network information.

Optionally, the first end may also compress the AI network information based on a preset data format, such as the protobuf data format used by ONNX, where protobuf is a data exchange format.

Alternatively, the first end may also compress the AI network information difference between the AI network information to be sent and the existing AI network information at the second end, and the first end only needs to send the compressed AI network information difference value to the second end, and then there is no need to compress and send the same AI network information between the AI network information to be sent and the existing AI network information of the second end, which can effectively save the transmission of the first end overhead.

Alternatively, the first end may compress the AI network information difference between the AI network information to be sent and the AI network information of the preset AI network, and the first end only needs to send the compressed AI network information difference to the second end to save the transmission overhead of the first end. Wherein, the preset AI network may be a predetermined protocol or a high-level configuration, such as some fixed AI network templates, or may also be a common AI network template for communication devices. The preset AI network includes initial values of network structure and network parameters.

Optionally, the AI network information difference includes at least one of the following:

specified network parameters;

index of network parameters;

Modified network parameters;

The modified parameter value in the modified network parameter;

The position of the modified parameter value in the modified network parameter;

the location of the modified reference value in the modified network parameters, the reference value being the maximum value in the network parameters;

A non-zero value in the modified network parameter;

the location of non-zero values in the modified network parameters;

Newly added network structure;

Deleted network structures;

Modified network structure.

In the embodiment of the present application, the preset model expression method includes any one of the following: a model expression method stipulated in an agreement, and a user-defined model expression method. Wherein, the self-defined model expression method may refer to a data structure defined by the communication device, which is used to describe the network structure and network parameters of the AI network.

Optionally, the content of the self-defined model expression includes at least one of the following: the network structure of the AI network, the attributes of the network parameters of the AI network, and the values of the network parameters of the AI network. Wherein, the attribute of the network parameter includes information such as the name, identification, and dimension of the network parameter; the value of the parameter of the AI network may be one or more.

Optionally, the representation of the network structure in the preset model representation includes at least one of the following:

The relationship between the network structures of the AI network;

Attributes of the network parameters of the AI network;

The location of non-zero values in the network parameters of the AI network;

The updated numerical position in the network parameters of the AI network.

It should be noted that the association relationship between the network structures may refer to the connection relationship between the input and output of each network structure (or also called nodes), for example, the output of the first node is connected to the input of the second node, and the output of the first node is connected to the input of the second node. The output of the second node is connected to the input of the third node, and so on. The attribute of the network parameter includes information such as a name, an identifier, and a dimension of the network parameter.

In the embodiment of the present application, the first end converts the AI network information into a corresponding transmission file based on a preset model expression method, and compresses the transmission file, including:

The first end converts the AI network information into at least one transmission file based on at least one preset model representation, and one preset model representation corresponds to at least one transmission file;

The first end merges and compresses the at least one transmission file, or the first end compresses the at least one transmission file separately and then merges them.

It should be noted that there may be multiple preset model expressions, and one preset model expression may correspond to at least one transmission file, for example, the network structure and network parameters are respectively converted based on a preset model expression into corresponding transfer files. Of course, it is also possible to convert the network structure and network parameters into a transmission file based on a preset model expression manner.

After the first end converts the AI network information into at least one transmission file based on at least one preset model expression, merge the at least one transmission file together for compression, and send the compressed transmission file, or The at least one transmission file mentioned above is compressed independently, and then combined and sent together after compression, or each compressed transmission file can also be sent separately.

Optionally, the AI network information includes network structure and network parameters, and the first end converts the AI network information into a corresponding transmission file based on a preset model expression, and compresses the transmission file, including:

The first end converts the network structure into a first transmission file based on a preset model representation, converts the network parameters into a second transmission file based on the preset model representation, and converts the first The first transmission file and the second transmission file are respectively compressed.

That is to say, the first end can convert the network structure and network parameters into corresponding transmission files based on the preset model expression, and then obtain two transmission files, and compress the two transmission files respectively, and can separately The two compressed transmission files may be sent, or they may be combined and sent together.

For example, the first end is a base station, and the second end is a terminal. When the terminal initially accesses, the base station may save the network structure of the trained AI network into a transmission file in a corresponding format based on a preset model expression Compress and send to the terminal; if network parameters are required, the base station saves the network parameters based on the preset model expression method as a transmission file for compression and sends to the terminal. Optionally, the transmission file corresponding to the network structure and the transmission file corresponding to the network parameters may be of different file types. In addition, the base station may send the compressed transmission file through the data channel.

Optionally, when the AI network information includes network parameters, the compressed AI network information also includes compressed network parameters, and the first end sends the compressed AI network information to the second end ,include:

The first end sends the compressed network parameters to the second end according to the priority order of the compressed network parameters based on the priority order of the compressed network parameters.

For example, also taking the first end as a base station and the second end as a terminal as an example, when the base station sends compressed network parameters, it may divide the compressed network parameters into N groups according to a preset priority order, Groups with high priority are delivered first, and groups with low priority are delivered later or not delivered when transmission resources are limited.

Optionally, the first end sends the compressed network parameters to the second end according to the priority order based on the priority order of the compressed network parameters, including:

grouping the compressed network parameters by the first end based on the priority order of the compressed network parameters;

In some specific scenarios, such as when the transmission resource is less than the preset threshold, the first end discards the grouped network parameters and sends the remaining network parameters in a preset order, the preset order being the grouped Network parameters are prioritized in order from low to high.

For example, the first end is a base station, and the second end is a terminal. When the base station transmits the compressed network parameters, it may divide the compressed network parameters into N groups according to a preset priority order, and transmit the compressed network parameters in the base station. When the resource is less than the preset threshold, that is, when the transmission resources of the base station are limited, or when there is a burst of high-priority traffic, the base station discards the network parameters in order of priority from low to high, that is, the priority The network parameters of the lowest group are discarded first until the transmission resources of the base station are sufficient to send the network parameters of the remaining groups, which ensures that the network parameters sent by the base station are network parameters with higher priority.

It should be noted that, after receiving the network structure and network parameters sent by the base station, the terminal may use a default value agreed in the protocol or be 0 for the network parameters not received.

In the embodiment of the present application, before the first end compresses the AI network information, the method further includes:

The first end receives first request information sent by the second end, and the first request information is used to request acquisition of target AI network information;

In this case, the first end compresses the AI network information, including:

The first end compresses the target AI network information.

That is to say, the second end can obtain the specified target AI network information based on the first request information, and then the first end compresses the target AI network information based on the first request information and sends it to the second end.

Optionally, before the first end compresses the target AI network information, the method further includes:

The first end judges whether the target AI network information needs to be updated;

When it is determined that the target AI network information needs to be updated, update the target AI network information;

In this case, the first end compresses the target AI network information, including:

The first end compresses the updated target AI network information.

It should be noted that after receiving the first update request sent by the second end, the first end can determine which target AI network information the second end wants to obtain based on the first update request. Before compressing the target AI network information, the first end may determine whether the target AI network information needs to be updated, if necessary, update the target AI network information, and update the target AI network information The network information is compressed and then sent to the second end; if the target AI network information does not need to be updated, the first end may directly compress the target AI network information and then send it to the second end.

For example, the second end sends a weight parameter update request (Request of Weight Updating) to the first end, specifying that one or some specific weight parameters need to be obtained, and the first end determines that an update is required based on the weight parameter update request , to update the specified weight parameters, and send the updated weight parameters to the second end after being compressed based on the preset model representation.

Optionally, the first request information includes at least one of the following:

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.

Wherein, the network effect measurement value may mean that the second end calculates the effect of the AI network according to a certain agreed method, and reports the calculated value, and the first end judges whether AI is required based on the calculated value carried in the first request information. Updates to network information. For example, for channel state information (Channel State Information, CSI) prediction, the terminal (second end) calculates the correlation between the predicted result and the actual measurement result and reports it to the base station (first end). Effect measurement value. The base station judges whether to update the AI network information according to the correlation reported by the terminal and the correlation calculated by its own AI network. If the correlation is poor, it means that the channel changes too fast and does not need to be updated, because the base station side The AI network also performed poorly. If the correlation calculated by the AI network on the base station side is significantly better than the correlation reported by the terminal, it means that the parameters of the AI network of the terminal are outdated and cannot match the current channel, and the base station needs to resend the AI network parameters.

Optionally, the target AI network information includes first target network parameters, and the first end compresses the target AI network information, including:

The first end converts the attributes and parameter values of the first target network parameters into a preset format based on a preset model representation and then compresses them;

Wherein, the attribute of the first target network parameter includes at least one of name, dimension, and length.

Exemplarily, taking the first end as the base station and the second end as the terminal, when the terminal initially accesses, it may select an AI network from the preset AI networks, and send the identifier of the AI network to the terminal, The attributes and parameter values of the network parameters of the AI network can also be converted into a preset format and then compressed, for example, compressed according to a preset model expression, and then sent to the terminal.

Optionally, when the first end is a network side device, the second end is a terminal, the AI network information includes network parameters, and the second end is handed over from the first cell to the second cell, Before the first end compresses the AI network information, the method further includes:

The first end calculates the correlation between the network parameters of the first cell and the network parameters of the second cell, and acquires a second target network parameter, where the second target network parameter includes at least one of the following items: the correlation is less than Network parameters with preset thresholds and the first N network parameters in a preset sequence, where the preset sequence is a sequence in which the correlation of network parameters is arranged in ascending order;

In this case, the first end compresses the AI network information, including:

The first end compresses the second target network parameters;

The first end sends the compressed AI network information to the second end, including:

The first end sends the compressed second target network parameters to the second end.

Exemplarily, when the terminal switches cells, the network side device (such as a base station) may send complete network parameters of an AI network to the terminal, or may also send partial network parameters to the terminal. For example, when a terminal switches from a first cell to a second cell, the second cell may obtain the network parameters corresponding to the first cell from the first cell, calculate the correlation of each network parameter, and the network side device will set the correlation to be less than the preset The threshold network parameters or the first N network parameters with less correlation are compressed and sent to the terminal, and the terminal may only update the received network parameters.

In this embodiment of the application, for the update and transfer of network parameters, the first end and the second end can interact in advance with the attributes of the network parameters to be transferred, including name, dimension, length, etc., and then pass the network parameters to be transferred through the preset The model expression is converted into the corresponding file and then compressed and delivered. Wherein, when the network structure is known, the order of all network parameters can be considered known, and the position of the network parameters that need to be updated in the overall list can be exchanged through bitmaps or combination numbers.

Optionally, when describing the network parameters based on the preset model expression, the first end can compress and transmit the attributes of the network parameters and the parameter values together, and the second end can judge based on the attributes of the received network parameters The length of the network parameters and the corresponding weights.

A network parameter can be a single number or a list of multiple numbers. When updating network parameters, if only part of the values in the list change, the first end can indicate the position of the changed value through the value position. The first end Only the changed value is transmitted, and the second end only updates the received value. Alternatively, when the first end transmits complete network parameters, it can also indicate the position of the non-zero value in the network parameter through the value position, the first end only transmits the non-zero value, and the second end treats the value that has not been received as 0 or a default value. Wherein, the numerical location indication may be an additional indication independent of network parameters, or a radio resource control (Radio Resource Control, RRC) configuration or a medium access control control element (Medium Access Control Control Element, MACCE) configuration.

In order to better understand the technical solutions provided by the embodiments of the present application, several specific examples are given below for illustration.

Embodiment one

The terminal and the base station use the joint AI network for CSI feedback, that is, the terminal converts the channel information into several bits of CSI feedback information through the AI network, and reports it to the base station. The network recovers the channel information.

Since the network of the base station and the terminal needs to be jointly trained, and the channel conditions of different cells are different, new network parameters may also be required. Therefore, when the terminal accesses the network, the base station needs to send the network parameters used by the terminal to the terminal.

The CSI feedback network can be divided into two parts, the terminal coding part and the base station decoding part. Usually, the base station only needs to send the AI network of the terminal coding part to the terminal. The base station can save the network structure and network parameters of the AI network to be sent into a file corresponding to the preset model expression method, such as an ONNX file or a pth file of pytorch, and then pass the entire file through the Packet Data Convergence Protocol (Packet Data Convergence Protocol, PDCP) layer compressed and sent to the terminal. After receiving the file, the terminal loads it into its own AI framework according to the file format to realize inference or continue training using the AI network. Among them, pytorch is a deep learning framework, and pth is a file type in pytorch.

Alternatively, the base station saves the AI network structure as a file corresponding to a preset model expression method, such as a TensorFlow meta file, and then sends the meta file to the terminal after ZIP compression. After the terminal receives the file, it builds the corresponding AI network under its own AI framework. The weight parameter that has not been received defaults to 0 or a fixed initial value. The terminal can use this initial value for training, or it can be used directly This default value is inferred. Among them, meta is a file type in TensorFlow.

Or, according to the pre-defined AI network template (that is, the above-mentioned preset AI network), the base station sends a certain index value (index) to indicate the corresponding AI network template, and then saves the network parameters as a file corresponding to the preset model expression , such as the data file of TensorFlow, or the index and network parameters of the network template can be saved together as a file corresponding to a custom network expression method and sent to the terminal. Among them, data is a file type in TensorFlow. After receiving the corresponding file, the terminal loads the corresponding network structure in its own AI framework according to the index of the pre-defined AI network template, initializes its own network parameters according to the network parameters of the AI network template, and then transfers the network parameters sent by the base station. The parameters are updated to the AI network. If some network parameters are not received, the initial values in the AI template are used.

Embodiment two

In the process of the terminal measuring the downlink channel, the terminal can use the AI network to perform channel measurement, including CSI reference signal (CSI Reference Signal, CSI-RS), demodulation reference signal (Demodulation Reference Signal, DMRS) and other channel estimation and radio resource management (Radio resource management, RRM) measurement, etc. Similarly, due to the channel differences of different cells, when the terminal switches cells, it often needs to switch the used network, because the network of the old cell can no longer support the channel changes in the new cell, and the new cell needs to send a new message to the terminal. network parameters.

The network structure of the AI network used by the terminal in the new cell can generally be considered to be consistent with the AI network structure of the old cell. However, due to changes in channel quality, retraining is required, which wastes time and is difficult to meet real-time requirements. Therefore, the new The cell can send its own trained network parameters with the same network structure as the terminal to the terminal. The terminal continues to train and infer based on these parameters, which can improve real-time efficiency, reduce the number of training times, and quickly converge the network.

The new cell can obtain the network parameters sent to the terminal last time from the old cell, compare its own network parameters with the network parameters sent to the user last time, and calculate the average correlation of weight values, for example:

in,

Indicates the i-th value of the network parameter sent to the terminal last time,

Indicates the i-th value of the network parameter sent to the terminal this time, N is the number of values of this network parameter, and the smaller the calculated C, the more relevant it is. There are many specific correlation calculation formulas. The base station judges which network parameters need to be updated according to the correlation between the two weight parameters, and selects M2 network parameters to be updated from the total M1 network parameters to update.

The base station can first inform the terminal which network parameters need to be updated. Since the terminal knows the network structure, it naturally knows a total of M1 network parameters. The base station and the terminal use the same sorting method for the M1 network parameters. This sorting method corresponds to the preset The base station notifies the terminal of the position of the M2 network parameters that need to be updated in this sort by means of a bitmap (bitmap) or combination number, etc. After receiving the notification from the base station, the terminal determines which M2 network parameters need to be updated , and then receive specific parameter values.

Since it has been determined in advance which network parameters need to be updated, the base station directly compresses the values of the network parameters that need to be updated into a data file in the preset model expression mode, without including weight name, dimension and other information, and directly transmits the values, and the terminal receives the values Afterwards, according to the dimensions of each network parameter that you know, analyze the parameter value of this network parameter at the corresponding position, and update the network parameters in your own AI framework.

Alternatively, the base station can directly compress the name and value of the network parameter to be updated into a file in a preset model expression mode, and the terminal first parses the name of the network parameter, and then obtains the corresponding value according to the name.

Optionally, for each network parameter that needs to be updated, the base station first calculates the corresponding correlation among all K values of the network parameter, and if the correlation is greater than a certain fixed threshold, it is considered that this value does not need to be updated, Then compress the position of all the values that need to be updated into the file corresponding to the preset model expression method through the method of bitmap or combination number and the value that needs to be transmitted. After receiving the compressed file, the terminal will first extract which network parameters It needs to be updated, and then judge which values of this network parameter need to be updated, and only update the values of the corresponding positions.

Specifically, the network parameters that need to be updated, and the numerical position that needs to be updated in each network parameter can be indicated in advance or configured to the terminal, and then the corresponding numerical information can be directly compressed and sent to the terminal.

Embodiment three

When two nodes using the AI network exchange AI network information, it is necessary to compress the AI network information into a file corresponding to a certain preset model expression method. The so-called preset model expression method is a data structure, according to a certain Rules describe the AI network and describe information such as AI network parameters.

For example, the preset model expression methods include TensorFlow, PyTorch, ONNX, etc., which describe the AI network description according to fixed rules, such as describing the network as a combination of several independent nodes, or describing it as a combination of several layers , each layer has an independent function, the specific function is expressed by some weight parameters and activation functions, etc., different frameworks have their own definition of the network, in order to uniformly transmit in the communication system, you can define some communication requirements network description.

The network structure and network parameters of the AI network should be separated to support the transfer of independent network structures and independent network parameters. The network structure can be based on the node, define the basic functions of the node, map all the weights to the input and output, and use the same number to represent the connection through the input and output, so as to describe the entire network through the number of nodes and input and output.

For network parameters, include at least one of the following:

1. The number of parameters is consistent with the number described in the network structure;

2. The dimension of the parameter needs to match the other parameters connected to it;

3. The value of the parameter, that is, the specific parameter content, matches the dimension;

4. In order to reduce the overhead of parameter transfer, the position information of the effective value in the parameter, the parameter that has not changed or is close to 0 may not be passed, and the corresponding position information is required to indicate the position of the effective value.

These contents can be combined into independent files arbitrarily, and the complete network parameters are expressed as the result after the difference, that is, some parameters are not passed, or the complete network is expressed as the result of partial weights, that is, the corresponding compression operation.

Embodiment four

In some scenarios, some functions only require the terminal to use the AI network, and do not require joint training with network-side devices, such as channel prediction and positioning on the terminal side. These AI networks are affected by the channel environment. As the terminal moves, it needs to be constantly Train and update the network. When terminal B enters the area of terminal A, terminal A can pass the trained AI network to terminal B, and terminal B performs training based on terminal A's AI network. Since terminal A and terminal B are in similar areas, the channel The information has a certain correlation, so using the AI network of terminal A to continue training can help terminal B converge faster.

Between any two associated nodes, the training complexity can be reduced by passing the trained AI network, or even directly use the AI network trained by other nodes.

Terminal A and terminal B have the same AI network structure for the same function. The two need to interact mainly with network parameters. Considering the transmission limitation of the sidelink, terminal A can divide the network parameters into N parts, respectively In different time slots (slots), it is transmitted to terminal B. When transmitting each time, information such as network parameter names and dimensions can be uniformly compressed with parameter values or can be directly transmitted after independent configuration. After receiving part of the network parameters of terminal A, terminal B can directly update the network parameters of its corresponding location, and use the updated network to directly infer, or terminal B can wait until all network parameters are received, and update its corresponding network parameters together. Network parameters, which mainly depends on whether some network parameters can work independently. This capability can be notified by terminal A to terminal B, or configured through a common base station. If it can work independently, terminal B will update each time it receives new network parameters. You can update your own network, otherwise you have to wait until all are received and update together.

Terminal B can also specify a network parameter that needs to be updated, and send information such as the name and/or dimension of the network parameter that needs to be updated to terminal A, and can also include information such as the location of the resource that is expected to receive the network parameter. Terminal A according to terminal B According to the order of the network parameters required by terminal A, the corresponding network parameter values are directly compressed into a file corresponding to a preset model expression method and sent to terminal B. After terminal B receives it, it updates the network it needs parameter.

Please refer to FIG. 3. FIG. 3 is another AI network information transmission method provided by the embodiment of the present application. As shown in FIG. 3, the AI network information transmission method includes the following steps:

Step 301. The second end receives the compressed AI network information sent by the first end, where the AI network information includes at least one of network structure and network parameters.

Understandably, the first end compresses the AI network information, and sends the compressed AI network information to the second end. Wherein, the compressing the AI network information may refer to compressing the AI network information into a file corresponding to the preset model representation method according to the preset model representation method, the so-called model representation method is a data structure, Describe the AI network structure, network parameters and other information according to certain rules. For the implementation process of the first terminal compressing the AI network information, reference may be made to the specific description in the method embodiment shown in FIG. 2 above, and details are not repeated here.

Wherein, the AI network information includes network structure and/or network parameters, and the compression of the AI network information by the first end also includes compressing the network structure and/or weight parameters, and compressing the compressed network structure and/or weight parameters Parameters are sent to the second end. For example, if the AI network information only includes the network structure, the first end only compresses and sends the network structure; or, the AI network information may only include network parameters, then the first end only compresses and sends the network parameters; or , the AI network information includes part of the network structure and part of the network parameters, and the first end compresses and sends the part of the network structure and part of the network parameters. Of course, the specific information content included in the AI network information may also be in other situations, which will not be described in detail here.

Optionally, before the step 301, the method further includes:

The second terminal sends first request information to the first terminal, where the first request information is used to request acquisition of target AI network information;

In this case, the step 301 includes:

The second end receives the compressed target AI network information sent by the first end.

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.

It should be noted that the AI network information transmission method provided by the embodiment of the present application is applied to the second end, corresponding to the AI network information transmission method applied to the first end provided in the embodiment of FIG. 2 above. For the specific implementation process of the relevant steps, reference may be made to the description in the above-mentioned method embodiment shown in FIG. 2 , and details are not repeated here to avoid repetition.

In the embodiment of the present application, the second end receives the compressed AI network information sent by the first end, the AI network information includes at least one of the network structure and network parameters, and there is no need to include all network information during the communication process. The structure and network parameters of the entire AI network are compressed and transmitted together, so that the network structure and network parameters of the AI network can be sent separately, which can effectively reduce the transmission overhead in the communication process.

The AI network information transmission method provided in the embodiment of the present application may be executed by an AI network information transmission device. In the embodiment of the present application, the AI network information transmission device provided in the embodiment of the present application is described by taking the AI network information transmission device executing the AI network information transmission method as an example.

Please refer to FIG. 4. FIG. 4 is a structural diagram of an AI network information transmission device provided in an embodiment of the present application. As shown in FIG. 4, the AI network information transmission device 400 includes:

A compression module 401, configured to compress AI network information, where the AI network information includes at least one of network structure and network parameters;

The sending module 402 is configured to send the compressed AI network information to the second end.

Optionally, the AI network information includes network structure and network parameters, and the compression module 401 is configured to perform any of the following:

Combining and compressing the network structure and the network parameters;

Compressing the network structure and the network parameters respectively.

Optionally, when the compression module 401 compresses the network structure and the network parameters respectively, the sending module 402 is configured to perform any of the following:

Combining and sending the compressed network structure and compressed network parameters to the second end;

The compressed network structure and the compressed network parameters are respectively sent to the second end.

Optionally, the compression module 401 is configured to perform any of the following:

Convert the AI network information into a corresponding transmission file based on the preset model expression method, and compress the transmission file;

Compressing the AI network information based on a preset data format;

Acquiring the AI network information to be sent and the existing AI network information of the second end, and performing an AI network information difference between the AI network information to be sent and the existing AI network information of the second end compression;

Acquiring the AI network information to be sent and the AI network information of the preset AI network, and compressing the AI network information difference between the AI network information to be sent and the AI network information of the preset AI network.

specified network parameters;

index of network parameters;

Modified network parameters;

The modified parameter value in the modified network parameter;

The position of the modified parameter value in the modified network parameter;

A non-zero value in the modified network parameter;

the location of non-zero values in the modified network parameters;

Newly added network structure;

Deleted network structure;

Modified network structure.

Optionally, the preset model expression manner includes any one of the following: a protocol-agreed model expression manner, and a user-defined model expression manner.

Optionally, the content of the self-defined model expression includes at least one of the following: the network structure of the AI network, the attributes of the network parameters of the AI network, and the values of the network parameters of the AI network.

The relationship between the network structures of the AI network;

Attributes of the network parameters of the AI network;

The location of non-zero values in the network parameters of the AI network;

The updated numerical position in the network parameters of the AI network.

Optionally, the compression module 401 is also used for:

Converting AI network information into at least one transmission file based on at least one preset model representation, where one preset model representation corresponds to at least one transmission file;

Combining and compressing the at least one transmission file, or compressing and merging the at least one transmission file respectively.

Optionally, the AI network information includes network structure and network parameters, and the compression module 401 is also used for:

Converting the network structure into a first transmission file based on a preset model representation, converting the network parameters into a second transmission file based on the preset model representation, and converting the first transmission file and the The above-mentioned second transmission files are respectively compressed.

Optionally, the compressed AI network information includes compressed network parameters, and the sending module 402 is further configured to:

Based on the priority order of the compressed network parameters, send the compressed network parameters to the second end according to the priority order.

Optionally, the sending module 402 is also configured to:

grouping the compressed network parameters based on the priority order of the compressed network parameters;

When the transmission resource is less than the preset threshold, the grouped network parameters are discarded and the remaining network parameters are sent according to a preset order, the preset order being the order of priority of the grouped network parameters from low to high.

Optionally, the device also includes:

A receiving module, configured to receive first request information sent by the second end, where the first request information is used to request acquisition of target AI network information;

The compression module 401 is further configured to: compress the target AI network information.

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.

Optionally, the device also includes:

A judging module, configured to judge whether the target AI network information needs to be updated;

The compression module 401 is also used for:

Compressing the updated target AI network information.

Optionally, the target AI network information includes first target network parameters, and the compression module 401 is further configured to:

converting the attributes and parameter values of the first target network parameters into a preset format based on a preset model expression manner, and then compressing;

Optionally, the device is one of the network side device and the terminal, and the second end is the other of the network side device and the terminal; or,

The device and the second end are different nodes of a terminal; or,

The device and the second end are different nodes of the network side equipment.

Optionally, the device is a network side device, the second end is a terminal, the AI network information includes network parameters, and when the second end is handed over from a first cell to a second cell, the Before the first end compresses the AI network information, the method further includes:

The first end calculates the correlation between the network parameters of the first cell and the network parameters of the second cell, and obtains a second target network parameter, where the second target network parameter includes at least one of the following items: the correlation is less than Network parameters with preset thresholds, and the first N network parameters in a preset sequence, where the preset sequence is a sequence in which the correlation of network parameters is arranged in ascending order;

The compression module 401 is also used for:

Compressing the second target network parameters;

The sending module 402 is also used for:

Sending the compressed second target network parameters to the second end.

In the embodiment of the present application, the device can send compressed AI network information to the second end, the AI network information includes at least one of the network structure and network parameters, and there is no need to include all network information during the communication process. The structure and network parameters of the entire AI network are transmitted together, so that the network structure and network parameters of the AI network can be sent separately, which can effectively reduce the transmission overhead in the communication process.

The AI network information transmission apparatus 400 in the embodiment of the present application may be an electronic device, such as an electronic device with an operating system, or a component of the electronic device, such as an integrated circuit or a chip. The electronic device may be a terminal, or other devices other than the terminal. Exemplarily, the terminal may include, but not limited to, the types of terminal 11 listed above, and other devices may be servers, Network Attached Storage (NAS), etc., which are not specifically limited in this embodiment of the present application.

The AI network information transmission device 400 provided in the embodiment of the present application can realize each process implemented in the method embodiment shown in FIG. 2 and achieve the same technical effect. To avoid repetition, details are not repeated here.

Please refer to FIG. 5. FIG. 5 is a structural diagram of another AI network information transmission device provided in the embodiment of the present application. As shown in FIG. 5, the AI network information transmission device 500 includes:

The receiving module 501 is configured to receive compressed AI network information sent by the first end, where the AI network information includes at least one of network structure and network parameters.

Optionally, the device also includes:

A sending module, configured to send first request information to the first end, where the first request information is used to request acquisition of target AI network information;

The receiving module 501 is also used for:

Receive the compressed target AI network information sent by the first end.

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.

Optionally, the first end is one of the network side device and the terminal, and the device is the other of the network side device and the terminal; or,

The first end and the device are different nodes of terminals; or,

The first end and the device are different nodes of network side equipment.

In the embodiment of the present application, the device receives the compressed AI network information sent by the first end, the AI network information includes at least one of the network structure and network parameters, and there is no need to include all network information during the communication process. The structure and network parameters of the entire AI network are compressed and transmitted together, so that the network structure and network parameters of the AI network can be sent separately, which can effectively reduce the transmission overhead in the communication process.

The AI network information transmission device 500 provided in the embodiment of the present application can realize each process implemented in the method embodiment shown in FIG. 3 and achieve the same technical effect. To avoid repetition, details are not repeated here.

Optionally, as shown in FIG. 6 , the embodiment of the present application further provides a communication device 600, including a processor 601 and a memory 602, and the memory 602 stores programs or instructions that can run on the processor 601. The When the programs or instructions are executed by the processor 601, the various steps of the embodiment of the AI network information transmission method described above in FIG. 2 or FIG. 3 can be realized, and the same technical effect can be achieved. To avoid repetition, details are not repeated here.

The embodiment of the present application also provides a terminal, and each implementation process and implementation manner of the above-mentioned method embodiment in FIG. 2 or FIG. 3 can be applied to the terminal embodiment, and can achieve the same technical effect. Specifically, FIG. 7 is a schematic diagram of a hardware structure of a terminal implementing an embodiment of the present application.

The terminal 700 includes, but is not limited to: a radio frequency unit 701, a network module 702, an audio output unit 703, an input unit 704, a sensor 705, a display unit 706, a user input unit 707, an interface unit 708, a memory 709, and a processor 710. At least some parts.

Those skilled in the art can understand that the terminal 700 may also include a power supply (such as a battery) for supplying power to various components, and the power supply may be logically connected to the processor 710 through the power management system, so as to manage charging, discharging, and power consumption through the power management system. Management and other functions. The terminal structure shown in FIG. 7 does not constitute a limitation on the terminal, and the terminal may include more or fewer components than shown in the figure, or combine some components, or arrange different components, which will not be repeated here.

It should be understood that, in this embodiment of the present application, the input unit 704 may include a graphics processing unit (Graphics Processing Unit, GPU) 7041 and a microphone 7042, and the graphics processor 7041 is used by the image capture device ( Such as the image data of the still picture or video obtained by the camera) for processing. The display unit 706 may include a display panel 7061, and the display panel 7061 may be configured in the form of a liquid crystal display, an organic light emitting diode, or the like. The user input unit 707 includes at least one of a touch panel 7071 and other input devices 7072 . The touch panel 7071 is also called a touch screen. The touch panel 7071 may include two parts, a touch detection device and a touch controller. Other input devices 7072 may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, switch buttons, etc.), trackballs, mice, and joysticks, which will not be described in detail here.

In the embodiment of the present application, the radio frequency unit 701 may transmit the downlink data from the network side device to the processor 710 for processing after receiving the downlink data; in addition, the radio frequency unit 701 may send uplink data to the network side device. Generally, the radio frequency unit 701 includes, but is not limited to, an antenna, an amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, and the like.

The memory 709 can be used to store software programs or instructions as well as various data. The memory 709 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required by at least one function (such as a sound playing function, image playback function, etc.), etc. Furthermore, memory 709 may include volatile memory or nonvolatile memory, or, memory 709 may include both volatile and nonvolatile memory. Among them, the non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electronically programmable Erase Programmable Read-Only Memory (Electrically EPROM, EEPROM) or Flash. Volatile memory can be random access memory (Random Access Memory, RAM), static random access memory (Static RAM, SRAM), dynamic random access memory (Dynamic RAM, DRAM), synchronous dynamic random access memory (Synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (Double Data Rate SDRAM, DDRSDRAM), enhanced synchronous dynamic random access memory (Enhanced SDRAM, ESDRAM), synchronous connection dynamic random access memory (Synch link DRAM , SLDRAM) and Direct Memory Bus Random Access Memory (Direct Rambus RAM, DRRAM). The memory 709 in the embodiment of the present application includes but is not limited to these and any other suitable types of memory.

The processor 710 may include one or more processing units; optionally, the processor 710 integrates an application processor and a modem processor, wherein the application processor mainly handles operations related to the operating system, user interface, and application programs, etc., Modem processors mainly process wireless communication signals, such as baseband processors. It can be understood that the foregoing modem processor may not be integrated into the processor 710 .

In an implementation manner of the embodiment of this application, the terminal 700 is the first end. Wherein, the processor 710 is configured to: compress the AI network information, where the AI network information includes at least one of network structure and network parameters;

The radio frequency unit 701 is configured to: send the compressed AI network information to the second end.

Optionally, the AI network information includes network structure and network parameters, and the processor 710 is configured to perform any of the following:

Combining and compressing the network structure and the network parameters;

Compressing the network structure and the network parameters respectively.

Optionally, when the processor 710 is configured to compress the network structure and the network parameters respectively, the radio frequency unit 701 is further configured to perform any of the following:

Optionally, the processor 710 is configured to perform any of the following:

Compressing the AI network information based on a preset data format;

specified network parameters;

index of network parameters;

Modified network parameters;

The modified parameter value in the modified network parameter;

The position of the modified parameter value in the modified network parameter;

A non-zero value in the modified network parameter;

the location of non-zero values in the modified network parameters;

Newly added network structure;

Deleted network structure;

Modified network structure.

The relationship between the network structures of the AI network;

Attributes of the network parameters of the AI network;

The location of non-zero values in the network parameters of the AI network;

The updated numerical position in the network parameters of the AI network.

Optionally, the processor 710 is further configured to:

Converting the AI network information into at least one transmission file based on at least one preset model representation, where one preset model representation corresponds to at least one transmission file;

Optionally, the AI network information includes network structure and network parameters, and the processor 710 is further configured to:

Optionally, the compressed AI network information includes compressed network parameters, and the radio frequency unit 701 is further configured to:

Optionally, the radio frequency unit 701 is also used for:

receiving first request information sent by the second end, where the first request information is used to request acquisition of target AI network information;

The processor 710 is also used for:

Compressing the target AI network information.

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.

Optionally, the processor 710 is further configured to:

Judging whether it is necessary to update the target AI network information;

Compressing the updated target AI network information.

Optionally, the target AI network information includes first target network parameters, and the processor 710 is further configured to:

In another implementation manner of the embodiment of this application, the terminal 700 is a second terminal. Wherein, the radio frequency unit 701 is further configured to: receive compressed AI network information sent by the first end, where the AI network information includes at least one of network structure and network parameters.

Optionally, the radio frequency unit 701 is also used for:

Sending first request information to the first end, where the first request information is used to request acquisition of target AI network information;

Receive the compressed target AI network information sent by the first end.

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.

The technical solution provided by this application enables the network structure and network parameters of the AI network to be sent separately, thereby effectively reducing the transmission overhead in the communication process.

The embodiment of the present application also provides a network-side device. The various implementation processes and implementation methods of the above-mentioned method embodiments shown in FIG. 2 and FIG. 3 can be applied to the network-side device embodiment, and can achieve the same technical effect.

Specifically, the embodiment of the present application also provides a network side device. As shown in FIG. 8 , the network side device 800 includes: an antenna 81 , a radio frequency device 82 , a baseband device 83 , a processor 84 and a memory 85 . The antenna 81 is connected to a radio frequency device 82 . In the uplink direction, the radio frequency device 82 receives information through the antenna 81, and sends the received information to the baseband device 83 for processing. In the downlink direction, the baseband device 83 processes the information to be sent and sends it to the radio frequency device 82 , and the radio frequency device 82 processes the received information and sends it out through the antenna 81 .

The method performed by the network side device in the above embodiments may be implemented in the baseband device 83, where the baseband device 83 includes a baseband processor.

The baseband device 83 can include at least one baseband board, for example, a plurality of chips are arranged on the baseband board, as shown in FIG. The program executes the network device operations shown in the above method embodiments.

The network side device may also include a network interface 86, such as a common public radio interface (common public radio interface, CPRI).

Specifically, the network-side device 800 in this embodiment of the present disclosure further includes: instructions or programs stored in the memory 85 and operable on the processor 84, and the processor 84 calls the instructions or programs in the memory 85 to execute FIG. 4 or FIG. 5 The methods executed by each module shown in the figure achieve the same technical effect, so in order to avoid repetition, they are not repeated here.

Specifically, the embodiment of the present application also provides another network side device. As shown in FIG. 9 , the network side device 900 includes: a processor 901 , a network interface 902 and a memory 903 . Wherein, the network interface 902 is, for example, a common public radio interface (common public radio interface, CPRI).

Specifically, the network-side device 900 in this embodiment of the present disclosure further includes: instructions or programs stored in the memory 903 and executable on the processor 901, and the processor 901 calls the instructions or programs in the memory 903 to execute FIG. 4 or FIG. 5 The methods executed by each module shown in the figure achieve the same technical effect, so in order to avoid repetition, they are not repeated here.

The embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by the processor, each process of the method embodiment described above in FIG. 2 or FIG. 3 is implemented. , and can achieve the same technical effect, in order to avoid repetition, it will not be repeated here.

Wherein, the processor is the processor in the terminal described in the foregoing embodiments. The readable storage medium includes a computer-readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk, and the like.

The embodiment of the present application further provides a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the above-mentioned Figure 2 or Figure 3. Each process of the method embodiment described above can achieve the same technical effect, so in order to avoid repetition, details are not repeated here.

It should be understood that the chip mentioned in the embodiment of the present application may also be called a system-on-chip, a system-on-chip, a system-on-a-chip, or a system-on-a-chip.

The embodiment of the present application further provides a computer program/program product, the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement the above-mentioned Figure 2 or Figure 3 The various processes of the method embodiments can achieve the same technical effect, and are not repeated here to avoid repetition.

The embodiment of the present application also provides a communication system, including: a terminal and a network-side device, the terminal can be used to perform the steps of the method described in Figure 2 above, and the network-side device can be used to perform the method described in Figure 3 above or, the terminal may be used to perform the steps of the method described in FIG. 3 above, and the network side device may be used to perform the steps of the method described in FIG. 2 above.

It should be noted that, in this document, the term "comprising", "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, It also includes other elements not expressly listed, or elements inherent in the process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising that element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved. Functions are performed, for example, the described methods may be performed in an order different from that described, and various steps may also be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus a necessary general-purpose hardware platform, and of course also by hardware, but in many cases the former is better implementation. Based on this understanding, the essence of the technical solution of this application or the part that contributes to related technologies can be embodied in the form of computer software products, which are stored in a storage medium (such as ROM/RAM, disk, CD) contains several instructions to enable a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in various embodiments of the present application.

The embodiments of the present application have been described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementations. The above-mentioned specific implementations are only illustrative and not restrictive. Those of ordinary skill in the art will Under the inspiration of this application, without departing from the purpose of this application and the scope of protection of the claims, many forms can also be made, all of which belong to the protection of this application.

Claims

An artificial intelligence AI network information transmission method, comprising:

The first end compresses the AI network information, and the AI network information includes at least one of network structure and network parameters;

The first end sends the compressed AI network information to the second end.
The method according to claim 1, wherein the AI network information includes network structure and network parameters, and the first end compresses the AI network information, including any of the following:

The first end combines and compresses the network structure and the network parameters;

The first end compresses the network structure and the network parameters respectively.
The method according to claim 2, wherein, when the first end compresses the network structure and the network parameters respectively, the first end sends the compressed AI to the second end Network information, including any of the following:

The first end sends the compressed network structure and compressed network parameters to the second end in combination;

The first end sends the compressed network structure and the compressed network parameters to the second end respectively.
The method according to claim 1, wherein the first end compresses the AI network information, including any of the following:

The first end converts the AI network information into a corresponding transmission file based on a preset model representation, and compresses the transmission file;

The first end compresses the AI network information based on a preset data format;

The first end obtains the AI network information to be sent and the existing AI network information of the second end, and calculates the AI network information between the AI network information to be sent and the existing AI network information of the second end. Network information difference is compressed;

The first end obtains the AI network information to be sent and the AI network information of the preset AI network, and calculates the AI network information difference between the AI network information to be sent and the AI network information of the preset AI network to compress.
The method according to claim 4, wherein the AI network information difference includes at least one of the following:

specified network parameters;

index of network parameters;

Modified network parameters;

The modified parameter value in the modified network parameter;

The position of the modified parameter value in the modified network parameter;

the location of the modified reference value in the modified network parameters, the reference value being the maximum value in the network parameters;

A non-zero value in the modified network parameter;

the location of non-zero values in the modified network parameters;

Newly added network structure;

Deleted network structure;

Modified network structure.
The method according to claim 4, wherein, the preset model expression method includes any one of the following: a protocol-agreed model expression method, and a self-defined model expression method.
The method according to claim 6, wherein the content of the self-defined model expression includes at least one of the following: the network structure of the AI network, the attributes of the network parameters of the AI network, and the values of the network parameters of the AI network.
The method according to claim 4, wherein the expression of the network structure in the preset model expression includes at least one of the following:

The relationship between the network structures of the AI network;

Attributes of the network parameters of the AI network;

The location of non-zero values in the network parameters of the AI network;

The updated numerical position in the network parameters of the AI network.
The method according to claim 4, wherein the first end converts the AI network information into a corresponding transmission file based on a preset model expression, and compresses the transmission file, including:

The first end converts the AI network information into at least one transmission file based on at least one preset model representation, and one preset model representation corresponds to at least one transmission file;

The first end merges and compresses the at least one transmission file, or the first end compresses the at least one transmission file separately and then merges them.
The method according to claim 4, wherein the AI network information includes network structure and network parameters, and the first end converts the AI network information into a corresponding transmission file based on a preset model representation, and the transmission Files are compressed, including:

The first end converts the network structure into a first transmission file based on a preset model representation, converts the network parameters into a second transmission file based on the preset model representation, and converts the first The first transmission file and the second transmission file are respectively compressed.
The method according to any one of claims 1-10, wherein the compressed AI network information includes compressed network parameters, and the first terminal sends the compressed AI network information to the second terminal ,include:

The first end sends the compressed network parameters to the second end according to the priority order of the compressed network parameters based on the priority order of the compressed network parameters.
The method according to claim 11, wherein the first end sends the compressed network parameters to the second end according to the priority order based on the priority order of the compressed network parameters, comprising:

grouping the compressed network parameters by the first end based on the priority order of the compressed network parameters;

When the transmission resource is less than the preset threshold, the first end discards the grouped network parameters according to a preset order and sends the remaining network parameters, and the preset order is that the priority of the grouped network parameters starts from Lowest to highest order.
The method according to any one of claims 1-10, wherein, before the first end compresses the AI network information, the method further includes:

The first end receives first request information sent by the second end, and the first request information is used to request acquisition of target AI network information;

The first end compresses the AI network information, including:

The first end compresses the target AI network information.
The method according to claim 13, wherein the first request information includes at least one of the following:

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.
The method according to claim 13, wherein, before the first end compresses the target AI network information, the method further comprises:

The first end judges whether the target AI network information needs to be updated;

When it is determined that the target AI network information needs to be updated, update the target AI network information;

The first end compresses the target AI network information, including:

The first end compresses the updated target AI network information.
The method according to claim 13, wherein the target AI network information includes first target network parameters, and the first end compresses the target AI network information, including:

The first end converts the attributes and parameter values of the first target network parameters into a preset format based on a preset model representation and then compresses them;

Wherein, the attribute of the first target network parameter includes at least one of name, dimension, and length.
The method according to any one of claims 1-10, wherein the first end is one of a network-side device and a terminal, and the second end is the other of a network-side device and a terminal; or,

The first end and the second end are different nodes of a terminal; or,

The first end and the second end are different nodes of the network side equipment.
The method according to any one of claims 1-10, wherein the first end is a network-side device, the second end is a terminal, the AI network information includes network parameters, and at the second end In the case of switching from the first cell to the second cell, before the first end compresses the AI network information, the method further includes:

The first end calculates the correlation between the network parameters of the first cell and the network parameters of the second cell, and acquires a second target network parameter, where the second target network parameter includes at least one of the following items: the correlation is less than Network parameters with preset thresholds and the first N network parameters in a preset sequence, where the preset sequence is a sequence in which the correlation of network parameters is arranged in ascending order;

The first end compresses the AI network information, including:

The first end compresses the second target network parameters;

The first end sends the compressed AI network information to the second end, including:

The first end sends the compressed second target network parameters to the second end.
An AI network information transmission method, comprising:

The second end receives the compressed AI network information sent by the first end, where the AI network information includes at least one of network structure and network parameters.
The method according to claim 19, wherein, before the second end receives the compressed AI network information sent by the first end, the method further comprises:

The second terminal sends first request information to the first terminal, where the first request information is used to request acquisition of target AI network information;

The second end receives the compressed AI network information sent by the first end, including:

The second end receives the compressed target AI network information sent by the first end.
The method according to claim 20, wherein the first request information includes at least one of the following:

the name of the requested network parameter;

Identification of the requested network parameters;

network structure update request;

Network parameter update request;

Network Effect Measures for AI Networks.
The method according to claim 19, wherein the first end is one of the network-side device and the terminal, and the second end is the other of the network-side device and the terminal; or,

The first end and the second end are different nodes of a terminal; or,

The first end and the second end are different nodes of the network side equipment.
An AI network information transmission device, comprising:

A compression module, configured to compress AI network information, where the AI network information includes at least one of network structure and network parameters;

A sending module, configured to send the compressed AI network information to the second end.
An AI network information transmission device, comprising:

The receiving module is configured to receive the compressed AI network information sent by the first end, where the AI network information includes at least one of network structure and network parameters.
A communication device, comprising a processor and a memory, the memory stores programs or instructions that can run on the processor, and when the programs or instructions are executed by the processor, any one of claims 1-18 is implemented The steps of the AI network information transmission method described in claim 1, or the steps of the AI network information transmission method described in any one of claims 19-22.
A readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the AI network information transmission method according to any one of claims 1-18 are realized , or realize the steps of the AI network information transmission method described in any one of claims 19-22.