A residual propagation method and residual propagation apparatus for a network model

Abstract

The present disclosure provides a residual propagation method and residual propagation device of a network model, and relates to the technical field of communication. The specific implementation scheme is: applied to a network, the network includes a routing path, the routing path includes a source node, an intermediate node and a sink node; the source node stores all model slices in the preset sink node demand, and the sink node demand is the demand of the sink node for the model slices; the method comprises: traversing along the routing path from the sink node to the source node direction, and the intermediate node and / or the source node sends the model slices in the sink node demand to the sink node. In the present disclosure, the intermediate node and / or the source node transmit the required part according to the sink node demand to the sink node, thereby forming model residual propagation, and further reducing the time delay of the sink node obtaining the required model slices, reducing the data redundancy in the end-to-end communication process, and improving the resource utilization of the overall network.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, and in particular to a residual propagation method and apparatus for a network model. 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 communications means that information service systems cannot continuously meet people's complex, diverse, and intelligent information transmission needs. However, using artificial intelligence models to encode, propagate, 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 propagation capabilities. The propagation and reuse of these models will help enhance network intelligence while reducing overhead and resource waste, forming a highly intelligent and simplified network.

[0003] Given the highly intelligent nodes, minimalist network structure, integration of virtual and physical elements, and digital twins in the Intelligent Simplified Network, what is propagated within it will no longer be just traditional content data, but rather relatively stable, computationally generated models. The network possesses storage capabilities; models reside within the network, potentially on the end-user's side or in the cloud. Each node can absorb numerous models from the network to achieve self-evolution, a method similar to knowledge distillation. The essence of model propagation is federated learning, which requires corresponding protocols for support and control. Therefore, the current technical challenge lies in how to transmit models during communication. Summary of the Invention

[0004] This disclosure provides a residual propagation method and residual propagation device for a network model.

[0005] According to a first aspect of this disclosure, a residual propagation method for a network model is provided, wherein the method is applied to a network, the network including at least one routing path, all routing paths including a source node and a destination node, and intermediate nodes disposed between the source node and the destination node;

[0006] Among them, the source node stores all the model slices in the preset sink node requirements, and the sink node requirements are the sink node's requirements for model slices;

[0007] Residual propagation methods include:

[0008] Get the route path;

[0009] Traverse along the routing path from the sink node to the source node, with intermediate nodes and / or the source node sending the model slices required by the sink node to the sink node.

[0010] According to a second aspect of this disclosure, a residual propagation apparatus for a network model is provided, wherein it is applied in a network, the network including at least one routing path, all routing paths including a source node and a destination node, and an intermediate node disposed between the source node and the destination node;

[0011] Among them, the source node stores all the model slices in the preset sink node requirements, and the sink node requirements are the sink node's requirements for model slices;

[0012] The residual propagation device includes:

[0013] The path acquisition unit is used to obtain the routing path;

[0014] The processing unit is used to traverse along the routing path from the sink node to the source node, and intermediate nodes and / or source nodes send the model slices required by the sink node to the sink node.

[0015] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0016] At least one processor; and

[0017] A memory that communicates with at least one processor; wherein,

[0018] The memory stores instructions that can be executed by at least one processor, such that at least one processor can perform the methods described above.

[0019] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause a computer to perform the methods described above.

[0020] The beneficial effects of the technical solution provided in this disclosure are:

[0021] The solution provided in this embodiment of the present disclosure performs a reverse traversal along the routing path from the sink node to the source node. For the model slices that need to be transmitted, the intermediate nodes and / or source nodes traversed transmit the required parts to the sink node according to the requirements of the sink node, thereby forming model residual propagation. This not only reduces the latency for the sink node to obtain the required model slices, but also reduces data redundancy in the end-to-end communication process and improves the overall network resource utilization.

[0022] 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

[0023] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0024] Figure 1 This is a flowchart illustrating the residual propagation method for a network model provided in Embodiment 1 of this disclosure;

[0025] Figure 2 This is a flowchart illustrating step S102 of the residual propagation method for the network model provided in Embodiment 1 of this disclosure.

[0026] Figure 3 This is the routing path diagram of Embodiment 1 of this disclosure without model residual propagation;

[0027] Figure 4-8 This is a flowchart of the model residual propagation in Embodiment 1 of this disclosure;

[0028] Figure 9 This is the routing path diagram after the residual propagation of the model according to Embodiment 1 of this disclosure has ended;

[0029] Figure 10 This is a schematic diagram of the residual propagation device according to Embodiment 2 of this disclosure;

[0030] Figure 11 This is a block diagram of an electronic device according to an embodiment of the present disclosure. Detailed Implementation

[0031] 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.

[0032] In the intelligent and simplified network, business information is primarily disseminated through artificial intelligence (AI) models. By using AI models to compress the initial business information into a second business information related to the AI ​​model, the data communication volume in the network is significantly reduced, with compression efficiency surpassing traditional compression algorithms. Specifically, the sending device uses a pre-configured first model to extract the initial 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 the pre-configured second model to reconstruct it into a third business information. The third business information reconstructed 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. If an update is required, the module transmits the pre-configured third model to the receiving device, which 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 help enhance network intelligence while reducing overhead and resource waste. A 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 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 nearby nodes that may have that slice.

[0033] The transmission of service information and models occurs at the network layer of the communication network, based on network layer protocols. The network nodes traversed along the path of transmitting service information and models include intelligent routers. The functions of intelligent routers include, but are not limited to, service information transmission, model transmission, model self-updating, and security protection. The transmission function of intelligent routers involves transmitting service information or models from source nodes to destination nodes, with multiple paths existing between them. The model transmission function of intelligent routers can transmit model slices, improving transmission speed by strategically arranging model slices to travel along multiple paths.

[0034] Example 1

[0035] Figure 1This disclosure illustrates a residual propagation method for a network model, which is applied to a network including at least one routing path. All routing paths include a source node and a destination node, as well as intermediate nodes disposed between the source node and the destination node.

[0036] Among them, the source node stores all the model slices in the preset sink node requirements, and the sink node requirements are the sink node's requirements for model slices;

[0037] like Figure 1 As shown, it includes:

[0038] Step S101: Obtain the routing path;

[0039] Step S102: Traverse along the routing path from the sink node to the source node. Intermediate nodes and / or source nodes send the model slices required by the sink node to the sink node.

[0040] This disclosure performs a reverse traversal along the routing path from the sink node to the source node. For model slices that need to be transmitted, the intermediate nodes and / or source nodes traversed transmit the required parts to the sink node according to the sink node's requirements, thereby forming model residual propagation. This not only reduces the latency for the sink node to obtain the required model slices, but also reduces data redundancy in the end-to-end communication process and improves the overall network resource utilization.

[0041] Specifically, when the routing path is determined, the source node, intermediate node and sink node on the routing path are also determined. At this time, the transmission nodes can be traversed in reverse order from the sink node to the source node along the routing path. The transmission nodes include intermediate nodes and source nodes. Only the model slices required by the sink node need to be transmitted, thus forming the model residual transmission.

[0042] For example, the model slices required by the sink node include the first model slice and the second model slice. When the first intermediate node under the sink node has both the second and third model slices, it is only necessary to transmit the second model slice to the sink node. Then, continue traversing to the second intermediate node. At this time, the second intermediate node only has the second model slice. Since the sink node already has the second model slice and the second intermediate node is the last intermediate node, skip the second intermediate node and traverse to the source node. The source node has all the model slices required by the sink node. At this time, the source node sends the first model slice to the sink node.

[0043] Specifically, intermediate nodes store model slices, but the model slices in intermediate nodes may include model slices already existing in the destination nodes, and the model slices in intermediate nodes may also include model slices required by the destination nodes.

[0044] For example, if the sink node already has the third model slice and the fourth model slice, then the sink node needs to store the first model slice and the second model slice.

[0045] At this point, the first intermediate node in the routing path has a second model slice and a third model slice. That is, the first intermediate node at this point has the model slice that the sink node already has—the third model slice, and the model slice that the sink node needs—the second model slice.

[0046] At this point, the second intermediate node in the routing path only contains the third model slice, meaning that the second intermediate node at this time only contains the model slice that the destination node already has—the third model slice.

[0047] This disclosure provides a possible implementation, wherein determining the sink node requirements includes the following steps:

[0048] Based on the types and quantities of model slices in the source node and the type and quantity of model slices in the sink node, determine the types and quantities of model slices that the sink node is still missing.

[0049] Specifically, since the source node stores all the model slices required by the sink node, it is possible to search for model slices that the sink node is currently missing in the source node. These missing model slices represent the sink node's requirements for model slices.

[0050] Specifically, model slices are AI models trained by various intelligent nodes in the intelligent network, and can include various types depending on the actual situation and node requirements.

[0051] For example, model slices can be used for editing models and generating animated models, and include, but are not limited to, classification models, segmentation models, graph neural network models, etc.

[0052] This disclosure provides a possible implementation in which the source node also stores all existing model slices in the destination node.

[0053] Specifically, at this point, the source node not only stores all the model slices required by the sink node, but also stores all the model slices already present in the sink node.

[0054] For example, the source nodes at this time include: 2 classification model slices, 2 segmentation model slices and 3 graph neural network model slices;

[0055] At this point, the sink node includes: 1 classification model slice and 1 segmentation model slice;

[0056] Therefore, the sink node's requirements for model slices at this time are: 1 classification model slice, 1 segmentation model slice, and 3 graph neural network model slices.

[0057] This disclosure provides a possible implementation, wherein determining the sink node requirements includes the following steps:

[0058] Assign a unique number to each model slice;

[0059] The missing model slices in the sink node are determined based on the model slice numbers in the source node and the sink node.

[0060] For example, each model slice is assigned a unique number. Suppose there are ten types of model slices, labeled A1-A10.

[0061] When the source node stores not only all the model slices required by the sink node, but also all the model slices already existing in the sink node;

[0062] The source nodes at this time include: model slice A1, model slice A2, model slice A3, model slice A4, model slice A5, model slice A6, model slice A7, model slice A8, model slice A9 and model slice A10;

[0063] The sink nodes at this time include: model slice A1, model slice A2, model slice A3, model slice A4, model slice A5, model slice A6 and model slice A7;

[0064] Therefore, the sink node's requirements for model slices at this time are: model slice A8, model slice A9, and model slice A10.

[0065] This disclosure provides a possible implementation, such as... Figure 2 As shown, step S102 specifically includes the following steps:

[0066] Step S1021: Traverse along the routing path from the sink node to the source node;

[0067] Step S1022: Determine whether the intermediate node has at least one model slice in the sink node requirement;

[0068] If it exists, the intermediate node sends the model slices in the requirements of the existing sink node to the sink node, and then executes step S1023;

[0069] If it does not exist, continue traversing along the routing path towards the source node and return to step S1022 until the requirements of the destination node are met or the requirements of the destination node cannot be met after traversing all intermediate nodes. When the requirements of the destination node for model slices cannot be met after traversing all intermediate nodes, the source node sends the model slices required by the destination node to the destination node.

[0070] Step S1023: Determine whether the requirements for the destination node are met;

[0071] If the condition is met, end the iteration.

[0072] If not satisfied, update the sink node requirements and continue traversing along the routing path towards the source node, returning to step S1022, until the sink node requirements are satisfied or the sink node requirements cannot be satisfied even after traversing all intermediate nodes. When the sink node's requirements for model slices cannot be satisfied even after traversing all intermediate nodes, the source node sends the model slices in the sink node's requirements to the sink node.

[0073] Specifically, suppose a node B is a source node and another node C is a destination node. The two nodes form a defined routing path. The routing path also includes an intermediate node set in at least one of the source nodes B and C. The routing path is defined as a routing path. There are model slices of different numbers and types at each transmission node along the routing path. The transmission node includes the intermediate node and the source node B.

[0074] When the destination node C wants to obtain some missing model slices, it does not directly send a request to the source node B for transmission. Instead, it queries each transmission node in the routing path one by one, starting from the previous hop transmission node of the destination node C, in reverse order. The specific steps are as follows:

[0075] Determine if an intermediate node contains at least one of the required sink node slices in the model slice:

[0076] If the intermediate node does not have the model slice required by the destination node, skip the intermediate node and continue searching to the previous hop transmission node.

[0077] If an intermediate node contains a model slice required by a destination node, then the transmitting node sends the model slice it contains, which is required by the destination node, to destination node C, and determines whether the model slice in the intermediate node satisfies the requirements of the destination node.

[0078] If the intermediate node does not meet the requirements of the destination node, the requirements of the destination node are updated, and then the process is traced back to the previous hop transmission node to find the model slice in the updated requirements of the destination node.

[0079] If an intermediate node satisfies the requirements of the sink node, that is, if the intermediate node contains all the model slices required by the sink node, then the intermediate node sends all the required model slices to the sink node C and terminates the backtracking search process.

[0080] If all intermediate nodes have been searched and the model slice required by the destination node C is still not found, then the source node B will send the model slice required by the current destination node to the destination node C according to the current destination node requirements.

[0081] Specifically, a sink node can initiate a request containing its own predefined sink node requirements to the previous intermediate node, allowing the intermediate node to retrieve the model slice from the sink node requirements based on the received request.

[0082] If a model slice is found in the destination node's requirements, the found model slice is sent to the destination node, the destination node's requirements are updated, and the request containing the updated destination node requirements is forwarded to the previous hop transmission node.

[0083] If the model slice required by the destination node is not found, the request is forwarded directly to the previous hop transmission node.

[0084] If the requirements of the destination node are not met when the last intermediate node is reached, the last intermediate node will forward the request containing the latest requirements of the destination node to the source node. The source node will then send the model slice from the latest requirements of the destination node to the destination node, thus ending the propagation of the model residual.

[0085] If an intermediate node possesses all the model slices required in the received request, it stops updating and forwarding requests containing the required slices after sending the corresponding model slices back to the destination node, terminates the backtracking lookup process, and ends the current model residual propagation.

[0086] For example, suppose a node B is a source node and another node C is a destination node. The two nodes form a defined routing path. The routing path also includes intermediate nodes set in at least one of the source nodes B and C. The intermediate nodes can be set as D1, D2 and D3. The routing path is defined as a routing path. There are model slices of different numbers and types at each transmission node passed through in the routing path. The transmission nodes include intermediate nodes and source node B.

[0087] When the destination node C wants to obtain some missing model slices, it does not directly send a request to the source node B for transmission. Instead, it queries each transmission node in the routing path in reverse order, starting from the previous hop transmission node of the destination node C, such as... Figure 3As shown, the routing path includes, in sequence: sink node C, intermediate node D1, intermediate node D2, intermediate node D3, and source node B.

[0088] Assign a unique number to each model slice. Assume there are ten types of model slices, labeled A1-A10.

[0089] When source node B includes not only all the model slices required by sink node C, but also all the model slices already present in sink node C;

[0090] At this time, the source node B includes: model slice A1, model slice A2, model slice A3, model slice A4, model slice A5, model slice A6, model slice A7, model slice A8, model slice A9 and model slice A10;

[0091] The sink node C at this time includes: model slice A1, model slice A2, model slice A3, model slice A4, model slice A5, model slice A6 and model slice A7;

[0092] And at this time, the intermediate node D1 includes: model slice A2, model slice A7 and model slice A8;

[0093] At this point, the intermediate node D2 includes: model slice A1, model slice A2, and model slice A9;

[0094] At this point, the intermediate node D3 includes: model slice A1, model slice A5, and model slice A6;

[0095] Based on the model slices in source node B and sink node C, we can see that:

[0096] At this point, the requirements for the sink nodes are: model slice A8, model slice A9, and model slice A10. For ease of explanation, the requirements for the sink nodes at this point are defined as the first requirement.

[0097] The specific steps are as follows:

[0098] like Figure 4 As shown, sink node C initiates a model slice request based on its model slice requirements. First, it iterates to intermediate node D1 under sink node C, confirming that intermediate node D1 currently contains the model slice A8 required in the first request, and that intermediate node D1 also contains the model slice A8 needed by the sink node. At this point, as... Figure 5 As shown, intermediate node D1 sends model slice A8 to sink node C, and determines that the sink node's requirements are not met at this time. The sink node's requirements are then updated to model slice A9 and model slice A10. For ease of explanation, the updated sink node requirements are defined as the second requirement.

[0099] like Figure 5 As shown, at this point, the request is directly forwarded to the previous hop transmission node—intermediate node D2—based on the second requirement. It is determined that intermediate node D2 contains the model slice A9 required in the second requirement, and that intermediate node D2 also contains the model slice A9 needed by the destination node. At this point, as... Figure 6 As shown, intermediate node D2 sends model slice A9 to sink node C through intermediate node D1, and determines that the sink node's requirement for model slice is not met at this time. The sink node's requirement is updated to: model slice A10. For ease of explanation, this updated sink node requirement is defined as the third requirement.

[0100] like Figure 6 As shown, at this time, the request is directly forwarded to the previous hop transmission node - intermediate node D3 according to the third requirement, and it is determined that the intermediate node D3 does not have the model slice A10 in the third requirement.

[0101] At this point, all intermediate nodes have been searched, but the model slice required for sink node C has not yet been found. Figure 7 As shown, at this point, the request is directly forwarded to the previous hop transmission node—source node B—based on the third requirement. Figure 8 As shown, based on the current third requirement, the source node B sends the remaining model slice A10 to the sink node C, at which point the current model residual propagation ends;

[0102] like Figure 9 As shown, the final sink node C includes model slices A1, A2, A3, A4, A5, A6, A7, A8, A9, and A10.

[0103] This disclosure provides a possible implementation in which model slices can be run and stored in all nodes of the network and can be transmitted between all nodes.

[0104] Specifically, source nodes, sink nodes, and intermediate nodes are all intelligent nodes in the communication system. These intelligent nodes include, but are not limited to, smartphones, tablets, laptops, and edge servers. All of these intelligent nodes have strong computing power, are capable of learning and training to generate AI models, can perform hierarchical semantic intelligent source coding and classification models, and have the ability to absorb many models on the network to achieve self-evolution.

[0105] This disclosure provides a possible implementation, wherein the network is a residual network.

[0106] It's worth noting that residual networks are convolutional neural networks proposed by four researchers from Microsoft Research. They won the image classification and object recognition categories in the 2015 ImageNet Large Scale Visual Recognition Challenge (ILSVRC). Residual networks are characterized by their ease of optimization and the ability to improve accuracy by significantly increasing depth. Their internal residual blocks use skip connections, mitigating the gradient vanishing problem that arises when increasing depth in deep neural networks.

[0107] Example 2

[0108] Figure 10 This invention discloses a residual propagation apparatus for a network model, characterized in that it is applied to a network, the network including at least one routing path, all routing paths including a source node and a destination node, and an intermediate node disposed between the source node and the destination node;

[0109] Among them, the source node stores all the model slices in the preset sink node requirements, and the sink node requirements are the sink node's requirements for model slices;

[0110] like Figure 10 As shown, the residual propagation device includes:

[0111] Path acquisition unit 201 is used to acquire the routing path;

[0112] Processing unit 201 is used to traverse from the sink node to the source node along the routing path, and intermediate nodes and / or source nodes send the model slices required by the sink node to the sink node.

[0113] This disclosure performs a reverse traversal along the routing path from the sink node to the source node. For model slices that need to be transmitted, the intermediate nodes and / or source nodes traversed transmit the required parts to the sink node according to the sink node's requirements, thereby forming model residual propagation. This not only reduces the latency for the sink node to obtain the required model slices, but also reduces data redundancy in the end-to-end communication process and improves the overall network resource utilization.

[0114] This disclosure provides a possible implementation, wherein the processing unit 202 includes:

[0115] The requirement determination module is used to determine the types and quantities of model slices that are still missing in the sink node based on the types and quantities of model slices in the source node and the sink node.

[0116] This disclosure provides a possible implementation, wherein the source node also includes all existing model slices in the destination node.

[0117] This disclosure provides a possible implementation, wherein the processing unit 202 includes:

[0118] The requirement determination module is used to assign a unique number to each model slice and determine the missing model slices in the sink node based on the number of the model slice in the source node and the number of the model slice in the sink node.

[0119] This disclosure provides a possible implementation, wherein the processing unit 202 includes:

[0120] The traversal module is used to traverse along the routing path from the sink node to the source node.

[0121] The first judgment module determines whether the intermediate node contains at least one model slice required by the sink node requirement;

[0122] If it exists, the intermediate node will send the model slices required by the existing sink node to the sink node;

[0123] If it does not exist, continue traversing along the routing path towards the source node, so that the intermediate node sends the model slices required by the existing sink node to the sink node;

[0124] The second judgment module determines whether the requirements of the sink node are met.

[0125] If the condition is met, end the iteration.

[0126] If the requirements are not met, update the sink node requirements and continue traversing along the routing path towards the source node, so that intermediate nodes send the model slices in the existing sink node requirements to the sink node. When the sink node requirements cannot be met after traversing all intermediate nodes, the source node sends the model slices in the sink node requirements to the sink node.

[0127] The beneficial effects achieved by the embodiments of this disclosure are the same as those of the residual propagation method embodiments of the network model described above, and will not be repeated here.

[0128] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0129] According to embodiments of this disclosure, this disclosure also provides an electronic device and a readable storage medium.

[0130] The electronic device includes:

[0131] At least one processor; and

[0132] A memory that communicates with at least one processor; wherein,

[0133] The memory stores instructions that can be executed by at least one processor, such that at least one processor can perform the methods described above.

[0134] The electronic device traverses the routing path from the sink node to the source node in reverse order. For the model slices that need to be transmitted, the intermediate nodes and / or source nodes traversed transmit the required parts to the sink node according to the requirements of the sink node, thereby forming model residual propagation. This not only reduces the latency for the sink node to obtain the required model slices, but also reduces data redundancy in the end-to-end communication process and improves the overall network resource utilization.

[0135] The non-transitory computer-readable storage medium stores computer instructions for causing a computer to perform the methods provided in embodiments of this disclosure.

[0136] The readable storage medium traverses the routing path in reverse order from the sink node to the source node. For model slices that need to be transmitted, the intermediate nodes and / or source nodes traversed transmit the required parts to the sink node according to the sink node's requirements, thereby forming model residual propagation. This not only reduces the latency for the sink node to obtain the required model slices, but also reduces data redundancy in the end-to-end communication process and improves the overall network resource utilization.

[0137] Figure 11 A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is 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. The electronic device may 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 examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0138] like Figure 11As shown, device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 302 or a computer program loaded from storage unit 308 into random access memory (RAM) 303. The RAM 303 may also store various programs and data required for the operation of device 300. The computing unit 301, ROM 302, and RAM 303 are connected to each other via bus 304. Input / output (I / O) interface 307 is also connected to bus 304.

[0139] Multiple components in device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of monitors, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0140] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 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 301 performs the various methods and processes described above, such as the residual propagation method for a network model. For example, in some embodiments, the residual propagation method for a network model can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 307. In some embodiments, part or all of the computer program can be loaded and / or installed on device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by computing unit 301, one or more steps of the residual propagation method for a network model described above can be performed. Alternatively, in other embodiments, computing unit 301 may be configured to perform a residual propagation method for a network model by any other suitable means (e.g., by means of firmware).

[0141] 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.

[0142] Program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, 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 may be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on the process machine, or entirely on the process machine or server.

[0143] 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. A machine-readable medium 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.

[0144] 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).

[0145] 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 implementations 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.

[0146] Computer systems can include clients and servers. Clients and servers are generally separate from each other and typically interact via a communication network. A client-server relationship is created by computer programs running on corresponding computers that have a client-server relationship with each other. Servers can be cloud servers, distributed system servers, or servers incorporating blockchain technology.

[0147] 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 performed 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.

[0148] 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 residual propagation method for a network model, characterized in that, Applied in a network, the network includes at least one routing path, and all the routing paths include a source node and a destination node, as well as an intermediate node disposed between the source node and the destination node; The source node stores all model slices in the preset sink node requirements, and the sink node requirements are the requirements of the sink node for model slices. The residual propagation method includes: Obtain the routing path; Traversing along the routing path from the destination node to the source node, the intermediate node and / or the source node send the model slices required by the destination node to the destination node.

2. The residual propagation method as described in claim 1, wherein, Determining the requirements for the destination node includes the following steps: The types and quantities of model slices that are still missing in the sink node are determined based on the types and quantities of model slices in the source node and the sink node.

3. The residual propagation method as described in claim 1, wherein, The source node also stores all existing model slices in the destination node.

4. The residual propagation method as described in claim 1, wherein, Determining the requirements for the destination node includes the following steps: Assign a unique number to each model slice; The missing model slices for the sink node are determined based on the model slice numbers in the source node and the sink node.

5. The residual propagation method as described in any one of claims 1-4, wherein, The step of traversing along the routing path from the destination node to the source node, wherein the intermediate nodes and / or the source node send the model slices required by the destination node to the destination node, includes: Traverse along the routing path from the sink node towards the source node; Determine whether the intermediate node contains at least one model slice from the sink node requirements; If it exists, the intermediate node will send the model slices in the requirements of the existing sink node to the sink node; If it does not exist, continue traversing along the routing path towards the source node, so that the intermediate node sends the model slice in the requirements of the sink node to the sink node; Determine whether the requirements of the sink node are met; If the condition is met, end the iteration. If the requirements are not met, the sink node requirements are updated, and the routing path is continued to traverse towards the source node, so that the intermediate node sends the model slices in the sink node requirements to the sink node. When the requirements of the sink node cannot be met after traversing all the intermediate nodes, the source node sends the model slices in the sink node requirements to the sink node.

6. The residual propagation method as described in claim 1, wherein, The model slices can be run and stored in all nodes of the network, and can be transmitted between all the nodes.

7. A residual propagation device for a network model, characterized in that, Applied in a network, the network includes at least one routing path, and all the routing paths include a source node and a destination node, as well as an intermediate node disposed between the source node and the destination node; The source node stores all model slices in the preset sink node requirements, and the sink node requirements are the requirements of the sink node for model slices. The residual propagation device includes: A path acquisition unit is used to acquire the routing path; The processing unit is used to traverse along the routing path from the sink node to the source node, wherein the intermediate node and / or the source node sends the model slices required by the sink node to the sink node.

8. The residual propagation device as claimed in claim 7, wherein, The processing unit includes: The requirement determination module is used to determine the types and quantities of model slices that are still missing in the sink node based on the types and quantities of model slices in the source node and the types and quantities of model slices in the sink node.

9. The residual propagation device as claimed in claim 7, wherein, The processing unit includes: The requirement determination module is used to assign a unique number to each model slice and determine the missing model slices of the sink node based on the number of the model slice in the source node and the number of the model slice in the sink node.

10. The residual propagation device according to any one of claims 7-9, wherein, The processing unit includes: The traversal module is used to traverse along the routing path from the sink node to the source node. The first judgment module determines whether the intermediate node has at least one model slice in the requirements of the sink node; If it exists, the intermediate node will send the model slices in the requirements of the existing sink node to the sink node; If it does not exist, continue traversing along the routing path towards the source node, so that the intermediate node sends the model slice in the sink node's requirement to the sink node; The second judgment module determines whether the requirements of the sink node are met. If the condition is met, end the iteration. If the requirements are not met, the sink node requirements are updated, and the routing path is continued to traverse towards the source node, so that the intermediate node sends the model slices in the sink node requirements to the sink node. When the requirements of the sink node cannot be met after traversing all the intermediate nodes, the source node sends the model slices in the sink node requirements to the sink node.

11. An electronic device, comprising: At least one processor; as well as A memory that is in communication with 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 method of any one of claims 1-6.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

Images