Model transmission method and device, electronic device, and storage medium

Abstract

This disclosure relates to the field of communication technology, and in particular to a model transmission method, apparatus, electronic device, and storage medium. The specific implementation scheme is as follows: A target model is transmitted to one or more core nodes in the core layer of a communication network; a first supply-demand field strength of the target model is obtained in the core nodes; multiple second supply-demand field strengths are obtained based on the demand for the target model from the aggregation node and its corresponding multiple branch nodes; the comprehensive supply-demand field strength of the aggregation node is obtained based on the second supply-demand field strengths; the difference between the first supply-demand field strength and the comprehensive supply-demand field strength is calculated to see if it reaches a preset transmission threshold; and when the difference reaches the preset transmission threshold, the core node transmits the target model to the aggregation node, and the aggregation node continues to transmit the target model to the branch nodes. This achieves the sharing of the target model in the intelligent and simplified network, while transmitting the model according to the needs of each network node, avoiding waste of network resources and reducing model transmission latency.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to model transmission methods, apparatus, electronic devices and storage media. Background Technology

[0002] In the future Internet of Everything, network nodes are trending towards intelligence. This intelligence leads to a rapid expansion of the information space, even resulting in a dimensional disaster. This exacerbates the difficulty of representing the information carrying space, making it difficult for traditional network service capabilities to match the high-dimensional information space. The excessive amount of data transmitted in communication makes information service systems unable to continuously meet people's complex, diverse, and intelligent information transmission needs. However, using artificial intelligence models to encode, transmit, and decode business information can significantly reduce the amount of data transmitted in communication services, greatly improving information transmission efficiency. These models are relatively stable and possess reusability and transmissibility. The transmission and reuse of these models will help enhance network intelligence while reducing overhead and resource waste, forming a highly intelligent and simplified network.

[0003] The core of the intelligent and simplified network lies in the transmission model. Because artificial intelligence models have numerous parameters and iterate rapidly, when a new model (a completely new model or an updated version) appears in the intelligent and simplified network, it needs to be shared within the network, transmitting the new model to network nodes that require it. However, since not all nodes in the network need this new model, transmitting it to every node would inevitably waste network resources (e.g., transmission incurs communication costs, and storage consumes memory resources). Therefore, how to proactively transmit the new model to network nodes that require it is a problem that urgently needs to be solved. Summary of the Invention

[0004] This disclosure provides a method, apparatus, device, and storage medium for actively transmitting models in a smart and simplified network, enabling the sharing of new models in the smart and simplified network while maintaining high transmission accuracy. The model is transmitted according to the needs of each network node, avoiding waste of network resources.

[0005] According to one aspect of this disclosure, a model transfer method is provided, comprising:

[0006] Transmit the target model to one or more core nodes in the core layer of the communication network;

[0007] Obtain the first supply and demand field strength of the target model in the core node;

[0008] Multiple second supply and demand field intensities are obtained based on the demand of the convergence node and the corresponding multiple branch nodes of the access layer of the communication network for the target model.

[0009] The comprehensive supply and demand field strength of the convergence node is obtained based on the strengths of the multiple second supply and demand fields;

[0010] Calculate the difference between the first supply and demand field strength and the comprehensive supply and demand field strength, and determine whether the difference reaches a preset transmission threshold.

[0011] In response to the difference reaching the preset transmission threshold, the target model is transmitted from the core node to the aggregation node. After receiving the target model, the aggregation node continues to transmit the target model to the branch node.

[0012] In response to the difference not reaching the preset transmission threshold, the core node stops transmitting the target model to the aggregation node.

[0013] Optionally, obtaining the first supply and demand field strength includes: when the target model is a brand new model, calculating the first supply and demand field strength based on the model index of the target model;

[0014] When the target model is an updated model, the first supply and demand field strength is calculated based on the model indicators of the target model and the demand of the core node for the historical version of the target model.

[0015] Optionally, when the target model is an update model, the preset transmission threshold is obtained according to the following formulas:

[0016] |E α |=ε|E x -E ix_min |

[0017] Among them, E α E represents the preset transmission threshold; x E represents the intensity of the first supply and demand field; ix_min This represents the minimum value in the second supply and demand field strength; ε is a percentage.

[0018] Optionally, the model transfer method further includes: when the target model is an update model, calculating the transfer probability of the target model according to the following formula:

[0019]

[0020] Where η represents the transmission probability; E x E represents the intensity of the first supply and demand field; ix E represents the overall supply and demand field strength of the convergence node; ix_min This represents the minimum value among the second supply and demand field strengths.

[0021] Optionally, the model transfer method further includes: when the target model is a completely new model, obtaining the preset transfer threshold according to the following formulas:

[0022] |E α |=ε|E x |

[0023] Among them, E α E represents the preset transmission threshold; x ε represents the intensity of the first supply and demand field; ε is a percentage.

[0024] Optionally, the model transfer method further includes: when the target model is a completely new model, calculating the transfer probability of the target model according to the following formula:

[0025]

[0026] Where η represents the transmission probability; E x E represents the intensity of the first supply and demand field; ix This represents the overall supply and demand field strength of the convergence node.

[0027] Optionally, the comprehensive supply and demand field strength of the convergence node obtained based on the plurality of second supply and demand field strengths needs to satisfy the following formula:

[0028]

[0029] At the same time, the value of δ needs to satisfy the following formula:

[0030]

[0031] in, This represents the overall supply and demand field strength of the convergence node; Each of the x branch nodes represents the intensity of the second supply and demand field; E represents the intensity of the second supply and demand field before the aggregation node; α This represents the preset transmission threshold.

[0032] Optionally, the model metrics include model accuracy and / or the amount of data required for model training and / or the number of model parameters.

[0033] According to another aspect of this disclosure, a model transmission device is provided, comprising:

[0034] The transmission module is used to transmit the target model to one or more core nodes in the core layer of the communication network;

[0035] The first acquisition module is used to acquire the first supply and demand field strength of the target model in the core node;

[0036] The second acquisition module is used to acquire multiple second supply and demand field strengths based on the demand of the access layer aggregation node and the corresponding multiple branch nodes of the communication network for the target model.

[0037] The statistics module is used to obtain the comprehensive supply and demand field strength of the convergence node based on the strength of the multiple second supply and demand fields;

[0038] The calculation module is used to calculate the difference between the first supply and demand field strength and the comprehensive supply and demand field strength, and to determine whether the difference reaches a preset transmission threshold.

[0039] The execution module, in response to the difference reaching the preset transmission threshold, transmits the target model to the aggregation node through the core node, and after the aggregation node receives the target model, it continues to transmit the target model to the branch node through the aggregation node;

[0040] In response to the difference not reaching the preset transmission threshold, the execution module controls the core node to stop transmitting the target model to the aggregation node.

[0041] This disclosure also provides an electronic device, including:

[0042] At least one processor; and

[0043] A memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the model transfer method described in any of the above technical solutions.

[0045] This disclosure also provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to perform the model transfer method according to any one of the above embodiments.

[0046] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the model transfer method according to any one of the above embodiments.

[0047] Based on the above-mentioned transmission model method, apparatus, device, and storage medium, this disclosure enables the sharing of new models in a smart and simplified network. At the same time, it transmits models according to the needs of each network node, avoiding waste of network resources, reducing model transmission latency, and better protecting user privacy and security.

[0048] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0049] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0050] Figure 1 This is a flowchart illustrating the steps of the model transfer method in an embodiment of this disclosure;

[0051] Figure 2 This is a flowchart of the model transfer method in the embodiments of this disclosure;

[0052] Figure 3 This is a schematic block diagram of the model transmission device in the embodiments of this disclosure. Detailed Implementation

[0053] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0054] In the intelligent and simplified network, business information is primarily disseminated through artificial intelligence (AI) models. By using AI models to compress the first business information to be transmitted into second business information related to the AI ​​model, the data communication volume in the network is significantly reduced, with compression efficiency far exceeding traditional compression algorithms. Specifically, the sending device uses a pre-configured first model to extract the first business information and obtain the second business information to be transmitted; the sending device then transmits the second business information to the receiving device. The receiving device receives the second business information and uses a pre-configured second model to recover the second business information to obtain third business information. The third business information recovered by the second model has slight quality differences compared to the original first business information, but the content is identical, providing a virtually indistinguishable user experience. Before the sending device transmits the second business information to the receiving device, an update module determines whether the receiving device needs to update the second model, and if so, transmits a pre-configured third model to the receiving device. The receiving device then uses the third model to update the second model. By processing business information through a pre-trained AI model, the data transmission volume in communication services can be significantly reduced, greatly improving information transmission efficiency. These models are relatively stable and possess reusability and propagation capabilities. Model propagation and reuse will help enhance network intelligence while reducing overhead and resource waste. The model can be divided into several model slices according to different segmentation rules. These model slices can also be transmitted between different network nodes, and the model slices can be assembled into a model. Model slices can be stored distributed across multiple network nodes. When a network node discovers that it is missing or needs to update a certain model or model slice, it can request it from surrounding nodes that may have that slice.

[0055] The transmission of the business information and the model both occur within a communication network, based on network protocols. The network nodes traversed along the paths for transmitting the business information and the model include intelligent routers. The functions of intelligent routers include, but are not limited to, business information transmission, model transmission, absorbing model self-updates, and security protection. The transmission function of intelligent routers involves transmitting business information or models from a source node to a destination node, with multiple paths existing between the source and destination nodes. The model transmission function of intelligent routers can transmit model slices, improving the model transmission rate by rationally arranging model slices to travel along multiple paths and transmitting model slices via multiple paths.

[0056] This disclosure provides a model transfer method, such as Figure 1 As shown, it includes:

[0057] Step S101: Transmit the target model to one or more core nodes of the core layer of the communication network and obtain the first supply and demand field strength of the target model in the core nodes.

[0058] Step S102: Obtain multiple second supply and demand field strengths based on the demand of the aggregation node of the access layer of the communication network and the corresponding multiple branch nodes for the target model.

[0059] Step S103: Obtain the comprehensive supply and demand field strength of the convergence node based on the strength of multiple second supply and demand fields;

[0060] Step S104: Calculate the difference between the first supply and demand field strength and the comprehensive supply and demand field strength, and determine whether the difference reaches a preset transmission threshold.

[0061] In step S105, in response to the difference reaching the preset transmission threshold, the target model is transmitted from the core node to the aggregation node. After receiving the target model, the aggregation node continues to transmit the target model to the branch node.

[0062] In step S106, in response to the difference not reaching the preset transmission threshold, the core node stops transmitting the target model to the aggregation node until the difference between the first supply and demand field strength and the comprehensive supply and demand field strength reaches the preset transmission threshold. Only then will the core node actively transmit the target model to the aggregation node.

[0063] Specifically, since communication networks contain tens of thousands of network nodes—for example, the core layer can have up to 1,000 large nodes and the access layer can have up to 200,000 small nodes—if a new model is transmitted to each node individually when it appears in the intelligent simplified network, the transmission latency would be extremely high. Furthermore, not every node needs the model, and transmitting and storing the model both consume network resources, leading to resource waste. This embodiment proposes the concept of model supply and demand strength. Since each node has a different demand for the target model, supply and demand strength refers to the demand of each node for the target model. For example... Figure 2 As shown, the propagation process of the model can include the following two stages: ① When a new model (which can be a completely new model or an updated version of the model) appears in the intelligent simplified network, the new model is first propagated among the core nodes (such as large servers) of the network core layer, that is, it is gradually transmitted from the first core node that receives the new model to all other core nodes; ② After the new model is deployed or updated in the core nodes, the core nodes then transmit the new model to the non-core nodes of the access layer in the region. The non-core nodes include the aggregation node and the branch nodes connected to the aggregation node. The core nodes do not directly transmit the new model to the branch nodes, but transmit it to the aggregation node. The aggregation node then transmits the new model to its own branch nodes.

[0064] Through the above technical solution, in this embodiment, the deployment of a completely new model or the update of a new model version must first be performed on the core node in the core layer. After the core node receives the target model, it updates the supply and demand field strength of the node to serve as the supply node for the next stage of model transmission. This disclosure introduces the concept of supply and demand field strength. For ease of description, the supply and demand field strength of the x-th version of the target model at the supply node is defined as the first supply and demand field strength E. x To highlight the supply characteristics of the model, the first supply and demand field strength E is set. x The direction is positive. The supply and demand field strength E of the target model... x The magnitude of the supply-demand field intensity E is related to the model metrics of the model to be transmitted (including but not limited to model accuracy, the amount of data required for model training, and the number of model parameters) and the demand of the node for the historical version of the target model. Generally, it is considered that the higher the model accuracy, the larger the amount of data required for model training, and the more model parameters there are, the greater the demand for the historical version of the model, thus increasing the supply-demand field intensity E. x The larger it is. Furthermore, in order to more accurately solve the model indicators and to analyze the demand and supply-demand field strength E of historical model versions... x The non-linear relationship between them can be designed by taking various model indicators and the demand for historical model versions as inputs, and E x The modulus value is used as the output to build a neural network for training.

[0065] After the core node deploys the target model, it acts as a supply node. Nodes outside the core layer act as demand nodes (including aggregation nodes and their branch nodes), and can receive the target model transmitted from the core node. Demand nodes may not have the ability to actively generate entirely new models or update model versions; they can only receive models transmitted from supply nodes. We define the supply and demand field strength of the target model on the demand nodes as E. ix Let represent the x-th generation version of the model deployed at node i, to highlight E ix Based on the demand characteristics, a second supply and demand market strength E is set. ix Direction and the strength of the first supply and demand field E x The direction is opposite, that is, E ix In the negative direction, as the supply node transmits the target model to the demand node, the supply and demand field strength gradually decreases. It should be noted that the supply and demand field strength E at the demand node... ix The magnitude of E depends only on the demand for the target model at the demand node, and is unrelated to model metrics. The higher the demand, the stronger the supply-demand field E at the demand node. ix The larger the value, the greater the second supply and demand field strength E becomes when the demand at a certain demand node for the target model is 0. ixThe size is equal to 0. Because there are aggregation nodes and branch nodes in the access layer, the core node does not directly transmit the target model to the branch nodes, but first transmits it to the aggregation node, which then allocates it according to the demand of each branch node. Therefore, after obtaining the supply and demand field strength E of each demand node... ix Next, it is necessary to collect the demand of its branch nodes in the aggregation node, and to collect a total demand based on its own demand and the demand of its branch nodes, namely the comprehensive supply and demand field strength. The core node then transmits the target model to the aggregation node based on the comprehensive supply and demand field strength.

[0066] Furthermore, during the process of propagating the target model from supply nodes to demand nodes, since the conditions for inter-node transmission must reach a preset transmission threshold before model probability propagation can proceed, demand nodes that have not reached the preset transmission threshold must first propagate their demand for the target model back to higher-level convergence nodes, until the supply and demand field strength E of the core node is reached. x Supply and demand field strength E at the convergence node ix Once the difference reaches a preset transmission threshold, the core layer supply node can transmit the model probabilistically to the access layer demand node. After the demand node receives and updates the model, it updates the supply-demand field strength E of the demand node based on the transmitted model metrics and the node's demand for the previous version of the model (if the target model is a completely new model, i.e., there is no historical version, this can be ignored, and only the model metrics need to be considered). ix This serves as the supply node for the next stage of model transmission (the aggregation node transmits the target model to its branch nodes) until the transmission is completed to all access layer nodes that meet the transmission conditions.

[0067] As an optional implementation, in step S101, the method of obtaining the first supply and demand field strength of the target model in the core node may include: when the target model is a brand new model, calculating the first supply and demand field strength based on the model index of the target model; when the target model is an updated model, calculating the first supply and demand field strength based on the model index of the target model and the demand of the core node for the historical version of the target model. An updated model refers to a target model that is a new version model, and it also has a historical version model. When the target model is an updated model, calculating its supply and demand field strength also needs to consider the demand of the core node for its previous version model.

[0068] As an optional implementation, when the target model is an update model, the preset transmission threshold is obtained according to the following formulas:

[0069] |E α |=ε|E x -E ix_min |

[0070] Among them, E α Indicates the preset transmission threshold; Ex Indicates the strength of the primary supply and demand market; E ix_min This represents the minimum value in the second supply and demand field strength; ε is a percentage, which can be determined according to the actual application scenario, for example, it can be set to 20%. Setting a reasonable preset transmission threshold can, on the one hand, prevent the core node from having a large transmission workload and excessive load, thus reducing network communication overhead; on the other hand, it can ensure that models that need the target model can obtain it in a timely manner, realizing model sharing.

[0071] Furthermore, the supply and demand intensity E of the core node x Supply and demand field strength E at the convergence node ix Once the difference reaches a preset transmission threshold, the target model can be probabilistically transmitted between the core node and the aggregation node (i.e., the conditions for transmission are met, transmission is possible but not guaranteed). When the target model is a completely new model, the transmission probability of the target model is calculated according to the following formula:

[0072]

[0073] Where η represents the transmission probability; E x Indicates the strength of the primary supply and demand market; E ix E represents the overall supply and demand intensity of the convergence node; ix_min This represents the minimum value in the second supply and demand field strength.

[0074] As an optional implementation, when the target model is a completely new model, the preset transmission threshold is obtained according to the following formulas:

[0075] |E α |=ε|E x |

[0076] Among them, E α Indicates the preset transmission threshold; E x This represents the initial supply and demand strength; ε is a percentage, which can be set to 40% or other values ​​depending on the actual application scenario. Because some demand nodes have insufficient demand for the target model (such as individual user nodes), the supply and demand strength of these nodes is relatively low. Therefore, multiple branch nodes need to propagate their demand back to the upper-level aggregation node to merge the supply and demand strengths. Once the supply and demand strength of the aggregation node has sufficiently increased to meet a preset transmission threshold, the backpropagation of demand from branch nodes to the aggregation node stops, completing the merging of the aggregation node's supply and demand strengths and forming a comprehensive supply and demand strength.

[0077] Furthermore, when the target model is a completely new model, the transmission probability of the target model is calculated according to the following formula:

[0078]

[0079] Where η represents the transmission probability; E x Indicates the strength of the primary supply and demand market; E ix This indicates the overall supply and demand intensity of the convergence node.

[0080] As an optional implementation, the combined supply and demand field strength of the convergence node, obtained from multiple second supply and demand field strengths, needs to satisfy the following formula:

[0081]

[0082] At the same time, the value of δ needs to satisfy the following formula:

[0083]

[0084] in, This indicates the overall supply and demand intensity at the convergence node; These represent the second supply and demand field strengths of x branch nodes, respectively; E represents the intensity of the second supply and demand field before the convergence node; α This indicates the preset transmission threshold.

[0085] For example, such as Figure 2 As shown, taking node B1 as an example, it consists of branch nodes C1, C2, and C3. x Demand is propagated back to convergence node B1, and the supply and demand strength is merged with that of the convergence node to form a comprehensive supply and demand strength. It needs to be greater than or equal to the preset transmission threshold E α Only then can transmission be possible, that is, satisfying the requirement. The value of δ should be chosen such that it satisfies the overall supply and demand field strength. Greater than nodes B1, C1, C2, and C x The supply and demand field strength at any node is less than that at nodes B1, C1, C2, and C3. x The sum of the supply and demand field strengths enables the target model to be transmitted and to meet the needs of each access layer node as much as possible.

[0086] This disclosure also provides a model transmission device, such as Figure 3 As shown, it includes:

[0087] Transmission module 301 is used to transmit the target model to one or more core nodes of the core layer of the communication network;

[0088] The first acquisition module 302 is used to acquire the first supply and demand field strength of the target model in the core node;

[0089] The second acquisition module 303 is used to acquire multiple second supply and demand field strengths based on the demand of the convergence node of the access layer of the communication network and the corresponding multiple branch nodes for the target model.

[0090] The statistics module 304 is used to obtain the comprehensive supply and demand field strength of the convergence node based on the strength of multiple second supply and demand fields;

[0091] The calculation module 305 is used to calculate the difference between the first supply and demand field strength and the comprehensive supply and demand field strength, and to determine whether the difference reaches a preset transmission threshold.

[0092] Execution module 306, in response to the difference reaching the preset transmission threshold, transmits the target model to the aggregation node through the core node, and after the aggregation node receives the target model, it continues to transmit the target model to the branch node through the aggregation node.

[0093] When the difference does not reach the preset transmission threshold, the execution module 306 controls the core node to stop transmitting the target model to the aggregation node.

[0094] Specifically, in the intelligent simplified network, to fully describe the model propagation process between nodes, this embodiment proposes the concept of model supply and demand field strength. Since each node has a different demand for the target model, supply and demand field strength refers to the demand of each node for the target model. For example... Figure 2 As shown, the propagation process of the model can include the following two stages: ① When a new model (which can be a completely new model or an updated version of the model) appears in the intelligent simplified network, the new model is first propagated among the core nodes (such as large servers) of the network core layer, that is, it is gradually transmitted from the first core node that receives the new model to all other core nodes; ② After the new model is deployed or updated in the core nodes, the core nodes then transmit the new model to the non-core nodes of the access layer in the region. The non-core nodes include the aggregation node and the branch nodes connected to the aggregation node. The core nodes do not directly transmit the new model to the branch nodes, but transmit it to the aggregation node. The aggregation node then transmits the new model to its own branch nodes.

[0095] Through the above technical solution, in this embodiment, the deployment of a completely new model or the update of a new model version must first be performed on the core node in the core layer. After the core node receives the target model, it updates the supply and demand field strength of the node to serve as the supply node for the next stage of model transmission. This disclosure introduces the concept of supply and demand field strength. For ease of description, the supply and demand field strength of the x-th version of the target model at the supply node is defined as the first supply and demand field strength E. x To highlight the supply characteristics of the model, the first supply and demand field strength E is set. x The direction is positive. The supply and demand field strength E of the target model... xThe size of the supply-demand field intensity E is related to the model metrics of the model to be transmitted and the demand of the node for the historical version of the target model. Model metrics include, but are not limited to, model accuracy, the amount of data required for model training, and the number of model parameters. Generally, higher model accuracy, a larger amount of data required for model training, and a greater number of model parameters all indicate a greater demand for the historical version of the model. x The larger it is. Furthermore, in order to more accurately solve the model indicators and to analyze the demand and supply-demand field strength E of historical model versions... x The non-linear relationship between them can be designed by taking various model indicators and the demand for historical model versions as inputs, and E x The modulus value is used as the output to build a neural network for training.

[0096] After the core node deploys the target model, it acts as a supply node. Nodes outside the core layer act as demand nodes (including aggregation nodes and their branch nodes), and can receive the target model transmitted from the core node. Demand nodes may not have the ability to actively generate entirely new models or update model versions; they can only receive models transmitted from supply nodes. We define the supply and demand field strength of the target model on the demand nodes as E. ix Let represent the x-th generation version of the model deployed at node i, to highlight E ix Based on the demand characteristics, a second supply and demand market strength E is set. ix Direction and the strength of the first supply and demand field E x The direction is opposite, that is, E ix In the negative direction, as the supply node transmits the target model to the demand node, the supply and demand field strength gradually decreases. It should be noted that the supply and demand field strength E at the demand node... ix The magnitude of E depends only on the demand for the target model at the demand node, and is unrelated to model metrics. The higher the demand, the stronger the supply-demand field E at the demand node. ix The larger the value, the greater the second supply and demand field strength E becomes when the demand at a certain demand node for the target model is 0. ix The size is equal to 0. Because there are aggregation nodes and branch nodes in the access layer, the core node does not directly transmit the target model to the branch nodes, but first transmits it to the aggregation node, which then allocates it according to the demand of each branch node. Therefore, after obtaining the supply and demand field strength E of each demand node... ix Afterwards, the statistics module 304 also needs to collect the demand of its branch nodes in the aggregation node, and calculate a total demand based on its own demand and the demand of its branch nodes, namely the comprehensive supply and demand field strength. The core node transmits the target model to the aggregation node based on the comprehensive supply and demand field strength.

[0097] Furthermore, during the process of propagating the target model from supply nodes to demand nodes, since the conditions for inter-node transmission must reach a preset transmission threshold before model probability propagation can proceed, demand nodes that have not reached the preset transmission threshold must first propagate their demand for the target model back to higher-level convergence nodes, until the supply and demand field strength E of the core node is reached. x Supply and demand field strength E at the convergence node ix Once the difference reaches a preset transmission threshold, the core layer supply node can transmit the model probabilistically to the access layer demand node. After the demand node receives and updates the model, it updates the supply-demand field strength E of the demand node based on the transmitted model metrics and the node's demand for the previous version of the model (if the target model is a completely new model, i.e., there is no historical version, this can be ignored, and only the model metrics need to be considered). ix This serves as the supply node for the next stage of model transmission (the aggregation node transmits the target model to its branch nodes) until the transmission is completed to all access layer nodes that meet the transmission conditions.

[0098] As an optional implementation, in step S101, the method of obtaining the first supply and demand field strength of the target model in the core node may include: when the target model is a brand new model, calculating the first supply and demand field strength based on the model index of the target model; when the target model is an updated model, calculating the first supply and demand field strength based on the model index of the target model and the demand of the core node for the historical version of the target model. An updated model refers to a target model that is a new version model, and it also has a historical version model. When the target model is an updated model, calculating its supply and demand field strength also needs to consider the demand of the core node for its previous version model.

[0099] As an optional implementation, when the target model is an updated model, the calculation module 305 obtains the preset transmission threshold according to the following formulas:

[0100] |E α |=ε|E x -E ix_min |

[0101] Among them, E α Indicates the preset transmission threshold; E x Indicates the strength of the primary supply and demand market; E ix_min This represents the minimum value in the second supply and demand field strength; ε is a percentage, which can be determined according to the actual application scenario, for example, it can be set to 20%. Setting a reasonable preset transmission threshold can, on the one hand, prevent the core node from having a large transmission workload and excessive load, thus reducing network communication overhead; on the other hand, it can ensure that models that need the target model can obtain it in a timely manner, realizing model sharing.

[0102] Furthermore, the supply and demand intensity E of the core nodex Supply and demand field strength E at the convergence node ix Once the difference reaches a preset transmission threshold, the target model can be transmitted probabilistically between the core node and the aggregation node (i.e., the conditions for transmission are met, transmission is possible but not guaranteed). When the target model is a completely new model, the calculation module 305 calculates the transmission probability of the target model according to the following formula:

[0103]

[0104] Where η represents the transmission probability; E x Indicates the strength of the primary supply and demand market; E ix E represents the overall supply and demand intensity of the convergence node; ix_min This represents the minimum value in the second supply and demand field strength.

[0105] As an optional implementation, when the target model is a completely new model, the calculation module 305 obtains the preset transmission threshold according to the following formulas:

[0106] |E α |=ε|E x |

[0107] Among them, E α Indicates the preset transmission threshold; E x This represents the initial supply and demand strength; ε is a percentage, which can be set to 40% or other values ​​depending on the actual application scenario. Because some demand nodes have insufficient demand for the target model (such as individual user nodes), the supply and demand strength of these nodes is relatively low. Therefore, multiple branch nodes need to propagate their demand back to the upper-level aggregation node to merge the supply and demand strengths. Once the supply and demand strength of the aggregation node has sufficiently increased to meet a preset transmission threshold, the backpropagation of demand from branch nodes to the aggregation node stops, completing the merging of the aggregation node's supply and demand strengths and forming a comprehensive supply and demand strength.

[0108] Furthermore, when the target model is a completely new model, the calculation module 305 calculates the transmission probability of the target model according to the following formula:

[0109]

[0110] Where η represents the transmission probability; E x Indicates the strength of the primary supply and demand market; E ix This indicates the overall supply and demand intensity of the convergence node.

[0111] As an optional implementation, the combined supply and demand field strength of the convergence node, obtained from multiple second supply and demand field strengths, needs to satisfy the following formula:

[0112]

[0113] At the same time, the value of δ needs to satisfy the following formula:

[0114]

[0115] in, This indicates the overall supply and demand intensity at the convergence node; E represents the second supply and demand field strengths of x branch nodes respectively; B1 E represents the intensity of the second supply and demand field before the convergence node; α This indicates the preset transmission threshold.

[0116] For example, such as Figure 2 As shown, taking node B1 as an example, it consists of branch nodes C1, C2, and C3. x Demand is propagated back to convergence node B1, and the supply and demand strength is merged with that of the convergence node to form a comprehensive supply and demand strength. It needs to be greater than or equal to the preset transmission threshold E α Only then can transmission be possible, that is, satisfying the requirement. The value of δ should be chosen such that it satisfies the overall supply and demand field strength. Greater than nodes B1, C1, C2, and C x The supply and demand field strength at any node is less than that at nodes B1, C1, C2, and C3. x The sum of the supply and demand field strengths enables the target model to be transmitted and to meet the needs of each access layer node as much as possible.

[0117] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0118] Specifically, electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0119] The device includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). The RAM can also store various programs and data required for device operation. The computing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0120] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0121] The computing unit can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of computing units include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit performs the various methods and processes described above, such as the model transfer method in the above embodiments. For example, in some embodiments, the model transfer method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the computing unit, one or more steps of the model transfer method described above may be performed. Alternatively, in other embodiments, the computing unit may be configured to perform the model transfer method by any other suitable means (e.g., by means of firmware).

[0122] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0123] The program code used to implement the model transfer method of this disclosure can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0124] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0125] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0126] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0127] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0128] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0129] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A model transfer method, characterized in that, include: Transmit the target model to one or more core nodes in the core layer of the communication network; The first supply and demand field strength of the target model in the core node is obtained. The supply and demand field strength is used to characterize the demand or supply of the target model for each node. Multiple second supply and demand field intensities are obtained based on the demand of the aggregation node and the corresponding multiple branch nodes of the access layer of the communication network for the target model. The comprehensive supply and demand field strength of the convergence node is obtained based on the strengths of the multiple second supply and demand fields; Calculate the difference between the first supply and demand field strength and the comprehensive supply and demand field strength, and determine whether the difference reaches a preset transmission threshold. In response to the difference reaching the preset transmission threshold, the target model is transmitted from the core node to the aggregation node. After receiving the target model, the aggregation node continues to transmit the target model to the branch node. In response to the difference not reaching the preset transmission threshold, the core node stops transmitting the target model to the aggregation node.

2. The model transmission method according to claim 1, characterized in that, Obtaining the first supply and demand field strength includes: when the target model is a brand new model, calculating the first supply and demand field strength based on the model index of the target model; When the target model is an updated model, the first supply and demand field strength is calculated based on the model indicators of the target model and the demand of the core node for the historical version of the target model.

3. The model transmission method according to claim 1, characterized in that, When the target model is an update model, the preset transmission threshold is obtained according to the following formulas: in, This represents the preset transmission threshold; This indicates the strength of the first supply and demand field; This represents the minimum value among the second supply and demand field strengths; It is a percentage.

4. The model transfer method according to claim 1, characterized in that, Also includes: When the target model is an updated model, the transmission probability of the target model is calculated according to the following formula: in, Indicates the transmission probability; This indicates the strength of the first supply and demand field; This represents the overall supply and demand field strength of the convergence node; This represents the minimum value among the second supply and demand field strengths.

5. The model transmission method according to claim 1, characterized in that, Also includes: When the target model is a completely new model, the preset transmission threshold is obtained according to the following formulas: in, This represents the preset transmission threshold; This indicates the strength of the first supply and demand field; It is a percentage.

6. The model transfer method according to claim 1, characterized in that, Also includes: When the target model is a completely new model, the transmission probability of the target model is calculated according to the following formula: in, Indicates the transmission probability; This indicates the strength of the first supply and demand field; This represents the overall supply and demand field strength of the convergence node.

7. The model transfer method according to claim 1, characterized in that, The overall supply and demand strength of the convergence node obtained from the multiple second supply and demand strengths needs to satisfy the following formula: ， at the same time, The value of needs to satisfy the following formula: in, This represents the overall supply and demand field strength of the convergence node; Each of the x branch nodes represents the intensity of the second supply and demand field; This indicates the intensity of the second supply and demand field before the aggregation node is integrated; This indicates the preset transmission threshold.

8. The model transmission method according to claim 2, characterized in that, The model metrics include model accuracy and / or the amount of data required for model training and / or the number of model parameters.

9. A model transmission device, characterized in that, include: The transmission module is used to transmit the target model to one or more core nodes in the core layer of the communication network; The first acquisition module is used to acquire the first supply and demand field strength of the target model in the core node. The supply and demand field strength is used to characterize the demand or supply of the target model for each node. The second acquisition module is used to acquire multiple second supply and demand field strengths based on the demand of the access layer aggregation node and the corresponding multiple branch nodes of the communication network for the target model. The statistics module is used to obtain the comprehensive supply and demand field strength of the convergence node based on the strength of the multiple second supply and demand fields; The calculation module is used to calculate the difference between the first supply and demand field strength and the comprehensive supply and demand field strength, and to determine whether the difference reaches a preset transmission threshold. The execution module, in response to the difference reaching the preset transmission threshold, transmits the target model to the aggregation node through the core node, and after the aggregation node receives the target model, it continues to transmit the target model to the branch node through the aggregation node; In response to the difference not reaching the preset transmission threshold, the execution module controls the core node to stop transmitting the target model to the aggregation node.

10. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the model transfer method according to any one of claims 1-8.

11. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to execute the model transfer method according to any one of claims 1-8.

12. A computer program product comprising a computer program that, when executed by a processor, implements the model transfer method according to any one of claims 1-8.

Images