Methods, apparatus, electronic devices, and storage media for model updates
By segmenting and transmitting model slices in the intelligent simplified network and then merging and updating them at the receiving end, the problem of low model propagation efficiency in the intelligent simplified network is solved, achieving efficient and reliable model updates and improving network intelligence and user experience.
Patent Information
- Application Number
- CN202210119401.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-02-08
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2042-02-08
AI Technical Summary
In intelligent simplified networks, the model propagation process requires an efficient and reliable update method to ensure smooth communication, a requirement that is difficult to meet with existing technologies.
By dividing the model into multiple slices, transmitting these slices using paths in the intelligent simplified network, and fusing them at the receiving end, the new model replaces the old model once the similarity between the new model and the old model reaches a preset threshold, thus achieving model updates.
It improves model transmission efficiency, reduces network overhead and resource waste, enhances network intelligence, and ensures a consistent user experience.
Smart Images

Figure CN116629335B_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of network technology, and in particular to model update methods, apparatuses, electronic devices, and storage media in intelligent and simplified networks. Background Technology
[0002] Based on the background of highly intelligent nodes, a simplified network, a combination of virtual and physical elements, and digital twins in the Intelligent Simplified Network, we believe that what will be propagated in the Intelligent Simplified Network will no longer be just traditional content data, but rather relatively stable AI models generated computationally. The network has storage capabilities; the models reside on the network, possibly on the end-user side or in the cloud. Each node can absorb numerous models on 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. The technical problem this disclosure aims to solve is that as the Intelligent Simplified Network gradually evolves, communication processes with models as the propagation entity will become dominant, requiring suitable model update methods to ensure their efficient and reliable operation. Summary of the Invention
[0003] This disclosure provides a method, apparatus, electronic device, and storage medium for model updating.
[0004] According to a first aspect of this disclosure, a method for model updating is provided, comprising:
[0005] In response to the first user's model update request, the sending end divides the first new model into N slices;
[0006] The sending end finds a path based on the network location of the first user and sends N slices through the path;
[0007] After receiving N slices, the first user merges the N slices into a third new model; the similarity between the third new model and the first new model is greater than a preset first threshold; N is a positive integer;
[0008] The first user replaces the first old model with the third new model.
[0009] Preferred options also include:
[0010] The first user broadcasts the N slices to M second users within a first specified range; where M is a positive integer.
[0011] The M second users respectively fuse the N slices to obtain the third new model;
[0012] The M second users replace their respective first old models with the third new model.
[0013] According to a second aspect of this disclosure, a method for model updating is also provided, comprising:
[0014] In response to model update requests from K third users within the second specified range, the sending end divides the second new model into T slices; K and T are both positive integers;
[0015] The sending end finds the first shortest path based on the network location within the second specified range, and sends T slices through the first shortest path;
[0016] After receiving at least one of the T slices, any one of the K third users will broadcast at least one of the T slices within the second specified range;
[0017] In response to any one of the K third users obtaining all T slices via the broadcast method, any one of the K third users fuses the T slices to obtain a fourth new model; the similarity between the fourth new model and the second new model is greater than a preset second threshold;
[0018] Any one of the K third users replaces the second old model with the fourth new model.
[0019] Preferably, the sending end divides the second new model into T slices, including: horizontal layer-by-layer segmentation, vertical cross-layer segmentation, random segmentation, segmentation based on data volume, segmentation based on function, or segmentation based on purpose.
[0020] Preferably, the sending end finds a third shortest path based on the network location within the second specified range, and sends T slices through the third shortest path, including:
[0021] The sending end searches for the first shortest path based on the network locations within the second specified range;
[0022] The sending end sends any one of the T slices to any third user within the second specified range via the first shortest path;
[0023] The sending end searches for the second shortest path based on the network location of any of the third users;
[0024] The sending end will send any one of the T-1 slices to any one of the third users via the second shortest path;
[0025] The sending end repeats the above steps to send the remaining T-2 slices to any one of the third users through S paths; where S is a positive integer.
[0026] Preferably, the sending end searches for the first shortest path based on the network locations within the second specified range, including:
[0027] Obtain the second shortest path from the sending end to each of the third users within the second specified range;
[0028] The shortest path among all the second shortest paths is taken as the first shortest path.
[0029] According to a third aspect of this disclosure, a model updating apparatus is also provided, comprising:
[0030] The first slicing module: In response to the model update request from the first user, the sending end slices the first new model into N slices;
[0031] First sending module: used by the sending end to find a path based on the network location of the first user, and send N slices through the path;
[0032] First fusion module: used to fuse N slices into a third new model after the first user receives N slices; the similarity between the third new model and the first new model is greater than a preset first threshold; N is a positive integer;
[0033] First update module: used by the first user to replace the first old model with the third new model.
[0034] Preferred options also include:
[0035] First broadcast module: used by the first user to broadcast the N slices to M second users within a first specified range; where M is a positive integer;
[0036] The second fusion module is used by the M second users to fuse the N slices respectively to obtain the third new model;
[0037] The second update module is used by the M second users to replace their respective first old models with the third new model.
[0038] According to a fourth aspect of this disclosure, a model updating apparatus is also provided, comprising:
[0039] The second segmentation module is used to respond to model update requests from K third users within a second specified range. The sending end segments the second new model into T slices; K and T are both positive integers.
[0040] The fourth sending module is used by the sending end to find the first shortest path based on the network location within the second specified range, and to send T slices through the first shortest path;
[0041] Second broadcast module: used to broadcast at least one of the T slices within the second specified range after any one of the K third users receives at least one of the T slices;
[0042] The second fusion module is used to respond to any one of the K third users obtaining all T slices through the broadcast method, and any one of the K third users fusing the T slices to obtain a fourth new model; the similarity between the fourth new model and the second new model is greater than a preset second threshold.
[0043] The third update module is used by any one of the K third users to replace the second old model with the fourth new model.
[0044] Preferably, the second segmentation module includes: horizontal layer-by-layer segmentation, vertical cross-layer segmentation, random segmentation, segmentation based on data volume, segmentation based on function, or segmentation based on purpose.
[0045] Preferably, the fourth transmitting module includes:
[0046] First addressing module: used by the sending end to find the first shortest path based on the network location within the second specified range;
[0047] Fifth sending module: used by the sending end to send any one of the T slices to any third user in the second specified range through the first shortest path;
[0048] The second addressing module is used by the sending end to find the second shortest path based on the network location of any of the third users.
[0049] The sixth sending module is used by the sending end to send any one of the T-1 slices to any one of the third users through the second shortest path;
[0050] Repeat module: used by the sending end to reuse the above steps to send the remaining T-2 slices to any one of the third users through S paths; where S is a positive integer.
[0051] Preferably, the first addressing module includes:
[0052] The third addressing module is used to obtain the second shortest path from the sending end to each of the third users within the second specified range;
[0053] Shortest path module: used to take the shortest path in all the second shortest paths as the first shortest path.
[0054] According to a fifth aspect of this disclosure, an electronic device is also provided, comprising:
[0055] At least one processor; and
[0056] A memory communicatively connected to the at least one processor; wherein,
[0057] 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 method described in any of the above technical solutions.
[0058] According to a sixth aspect of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is also provided, wherein the computer instructions are used to cause the computer to perform any one of the methods described above.
[0059] According to the seventh aspect of this disclosure, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the method described in any one of the above technical solutions.
[0060] 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
[0061] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:
[0062] Figure 1 This is a schematic diagram of a first embodiment of a model updating method according to the present disclosure;
[0063] Figure 2 This is a schematic diagram of a second embodiment of a model updating method according to the present disclosure;
[0064] Figure 3 This is a schematic diagram of an embodiment of another model updating method according to the present disclosure;
[0065] Figure 4 This is a schematic diagram of an embodiment of the present disclosure in which the sending end finds a first shortest path based on the network location within the second specified range, and sends T slices through the first shortest path;
[0066] Figure 5 This is a schematic diagram of an embodiment in which the sending end of this disclosure searches for the first shortest path based on the network location within the second specified range;
[0067] Figure 6 This is a schematic diagram of a first embodiment of a model updating apparatus according to the present disclosure;
[0068] Figure 7 This is a schematic diagram of a second embodiment of a model updating apparatus according to the present disclosure;
[0069] Figure 8 This is a schematic diagram of an apparatus updated according to another model of this disclosure;
[0070] Figure 9 This is a schematic diagram based on the fourth transmitting module of this disclosure;
[0071] Figure 10 This is a schematic diagram of the first addressing module according to this disclosure;
[0072] Figure 11 This is a block diagram of an electronic device used to implement the model update method of the embodiments of this disclosure;
[0073] Explanation of reference numerals in the attached figures:
[0074] 3. A model updating device 4. Another model updating device
[0075] 301 First Segmentation Module 302 First Transmission Module
[0076] 303 First Fusion Module; 304 First Update Module
[0077] 305 First Broadcast Module; 306 Second Fusion Module
[0078] 307 Second Update Module
[0079] 401 Second Segmentation Module; 402 Fourth Transmission Module
[0080] 403 Second Broadcast Module; 404 Second Fusion Module
[0081] 405 Third Update Module
[0082] 4021 First Addressing Module; 4022 Fifth Transmission Module
[0083] 4023 Second Addressing Module; 4024 Sixth Sending Module
[0084] 4025 Repeat Module
[0085] 40211 Third Addressing Module; 40212 Shortest Path Module Detailed Implementation
[0086] 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.
[0087] The so-called intelligent and simplified network refers to a network that inherently possesses elements of "intelligent simplicity." Nodes within the network will become new types of intelligent nodes, while the network's protocol structure will tend towards extreme simplicity. Supported by characteristics such as native intelligence and cognitive reshaping, the network builds a targeted intelligent service ecosystem around different communication objects, forming a state of "extremely simple network, extremely intelligent nodes," ultimately achieving a self-evolving, self-optimizing, and self-balancing state where "simplicity arises from intelligence, and intelligence is promoted through simplicity." In the intelligent and simplified network, business information is primarily disseminated through artificial intelligence models. By using these models, the first business information to be transmitted is compressed into second business information related to the AI model, significantly reducing the amount of data communication in the network, 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 service information and uses a pre-configured second model to recover the second service information to obtain the third service information. The third service information recovered by the second model has a slight quality difference compared to the original first service information, but the content is consistent, providing a virtually identical user experience. Before the sending device transmits the second service information to the receiving device, the system further includes an update module that 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 service information using pre-trained artificial intelligence models, the amount of data transmitted in communication services can be significantly reduced, greatly improving information transmission efficiency. These models are relatively stable and possess reusability and propagation capabilities. The propagation and reuse of models 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 merged into a model. The model slices can be distributed and stored on multiple network nodes. When a network node discovers that it lacks or needs to update a certain model or model slice, it can request it from nearby nodes that may have that slice. The network nodes along the path 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 source nodes to destination nodes, with multiple paths existing between them. The model transmission function of intelligent routers can transmit model slices, improving the model transmission rate by rationally arranging model slices to travel through multiple paths. Before being merged into a model, the model slices cannot be used by the network node to process business information.
[0088] Specifically, this invention mainly provides a method and apparatus applied in intelligent simplified networks. In intelligent simplified networks, business information is mainly propagated through artificial intelligence models. By using the artificial intelligence model, the first business information to be propagated is compressed into second business information related to the artificial intelligence model, greatly reducing the data communication volume in the network, and the compression efficiency far exceeds that of 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 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 restore the second business information to obtain third business information. The third business information restored by the second model has a slight quality difference compared to the original first business information, but the content is consistent, providing a virtually identical user experience. Before the sending device transmits the second business information to the receiving device, the method further includes: an update module determining whether the receiving device needs to update the second model, and if an update is required, transmitting a pre-configured third model to the receiving device; the receiving device then uses the third model to update the second model. The core of the intelligent simplified communication standard is the model standard; therefore, model security is crucial. The nonlinear propagation of intelligent and simplified networks primarily involves the propagation of models. These models are relatively stable and possess reusability and propagation capabilities. The propagation and reuse of models will help enhance network intelligence while reducing overhead and resource waste.
[0089] like Figure 1 As shown, according to a first aspect of this disclosure, a method for model updating is provided, comprising:
[0090] S101: In response to the first user's model update request, the sending end divides the first new model into N slices; in the intelligent simplified network, the model is the main object of transmission; and the "volume" of the model, i.e., the number of bits, is quite large. If it is not divided into N slices for transmission, it will inevitably occupy too much network bandwidth and network usage time.
[0091] S102: The sending end finds a path based on the network location of the first user and sends N slices through the path; the network location of the first user is determined, so multiple algorithms can be used to determine the path; the so-called "path" refers to which routing nodes between the sending end and the first user serve as sending and relaying nodes for the N slices.
[0092] S103: After receiving N slices, the first user merges the N slices into a third new model; the similarity between the third new model and the first new model is greater than a preset first threshold; the first threshold, for example, is 96%; N is a positive integer; after receiving N slices, the first user also needs to merge all N slices to achieve the desired purpose. A slice of the model is a data packet that does not have independent execution functionality. The third new model is composed of the re-merged N slices.
[0093] S104: The first user replaces the first old model with the third new model. After obtaining the third new model, the first user can replace the first old model, thus achieving the purpose of model updating.
[0094] like Figure 2 As shown, preferably, it also includes:
[0095] S105: The first user broadcasts the N slices to M second users within a first specified range; M is a positive integer; in this embodiment, the N slices are sent to the M second users within the first specified range; the first specified range is typically within the local area network where the first user's network location is located. Since the first user and the M second users are on the same local area network, the first user can broadcast all the N slices it receives to the M second users respectively.
[0096] S106: The M second users respectively fuse the N slices to obtain the third new model; the M second users each fuse the N slices together to obtain the same third new model as the first user.
[0097] S107: The M second users replace their respective first old models with the third new model. After obtaining the third new model, each of the M second users can replace their respective first old models with the third new model. The first old models of the M second users are not necessarily the same; they can be old models with different functions.
[0098] like Figure 3 As shown, according to a second aspect of this disclosure, another method for model updating is also provided, comprising:
[0099] S201: In response to model update requests from K third users within a second specified range, the sending end divides the second new model into T slices; K and T are both positive integers; the second specified range refers to the local area network (LAN) where the K third users are located; of course, it can also be a combination of several specified LANs as the second specified range. In this embodiment, it is not just one first user who submits the model update request, but K third users within the second specified range who submit the model update request. Similarly, the sending end also needs to divide the second new model into T slices to facilitate the transmission of the second new model.
[0100] S202: The sending end searches for a first shortest path based on the network location within the second specified range, and sends T slices through the first shortest path. At this time, the third shortest path searched by the sending end is different from the first shortest path. The first shortest path is the shortest path between the sending end and the first user; while the third shortest path is the shortest path from the sending end to the second specified range. The second specified range, as mentioned above, is a combination of multiple local area networks (LANs) or a single LAN. Of course, there are many algorithms for finding the third shortest path, such as Dijkstra's algorithm.
[0101] S203: After receiving at least one of the T slices, any one of the K third users broadcasts at least one of the T slices within the second specified range. Since any one of the T slices sent by the sending end arrives at the network location within the second specified range, the K third users within the second specified range randomly obtain at least one of the T slices. To ensure that each of the K third users can obtain all T slices, each third user broadcasts any one of the T slices it receives to all third users within the second specified range. In this way, all K third users can obtain all T slices.
[0102] S204: In response to any one of the K third users obtaining all T slices via the broadcast method, any one of the K third users fuses the T slices to obtain a fourth new model; the similarity between the fourth new model and the second new model is greater than a preset second threshold; the second threshold is, for example, 90%; when each of the K third users has obtained T slices, the T slices can be fused to obtain the fourth new model.
[0103] S205: Any one of the K third users replaces the second old model with the fourth new model. Ultimately, all K third users replace the second old model with the fourth new model.
[0104] Preferably, the transmitting end divides the first new model into N slices, including: horizontal layer-by-layer slicing, vertical cross-layer slicing, random slicing, slicing based on data volume, slicing based on function, or slicing based on purpose. In this embodiment, the model is described as a deep neural network. Horizontal layer-by-layer slicing is used because a deep neural network contains many layers, and horizontal model slicing divides the deep neural network by layers, for example, two layers per slice. Vertical cross-layer slicing is used to equally divide the parameters of each layer of the deep neural network into several parts, with each node storing one part, which is equivalent to each node storing the entire neural network structure, but each layer only stores a portion of the parameters. Random slicing means that each node of the deep neural network stores a skeleton network, and when the working nodes communicate with each other, they randomly transmit some neuron parameters belonging to the non-skeleton network, thereby playing a role in exploring the global topology of the original network.
[0105] like Figure 4 As shown, preferably, the sending end finds a first shortest path based on the network location within the second specified range, and sends T slices through the first shortest path, including:
[0106] S2021: The sending end searches for a first shortest path based on the network locations within the second specified range. Since there are numerous routing nodes between the sending end and the second specified range of network locations, and the delays of different routing nodes and network connections vary, an algorithm is typically needed to find the shortest path between the sending end and the second specified range of network locations—that is, the path with the minimum delay between the sending end and the second specified range of network locations. Many algorithms exist for calculating the first shortest path, such as Dijkstra's algorithm.
[0107] S2022: The sending end sends any one of the T slices to any third user in the second specified range through the first shortest path; once the first shortest path is found, the sending end can send out any one of the N slices, and with a certain delay, the slice can reach the network location of the first user.
[0108] S2023: The sending end searches for a second shortest path based on the network location of any of the third users. In this embodiment, since each of the aforementioned slices has a certain delay t0 before reaching the first user, sending another slice through the first shortest path using the same method within t0 may cause network congestion and is also detrimental to the timely arrival of the other slice at the network location of the first user. Therefore, in this embodiment, in addition to the first shortest path, the sending end needs to find a second shortest path besides the first shortest path after sending a slice through the first shortest path using Dijkstra's algorithm or other better algorithms.
[0109] S2024: The sending end sends any one of the T-1 slices to any one of the third users via the second shortest path; after excluding any slice sent via the first shortest path, N-1 slices remain. Any one of the remaining N-1 slices is sent to the network location of the first user via the second shortest path. The delay of the second shortest path is not necessarily longer than the delay of the first shortest path, and may have a shorter delay.
[0110] S2025: The sending end repeats the above steps, sending the remaining T-2 slices to any one of the third users through S paths; where S is a positive integer. Repeating the above steps does not necessarily mean that N-2 paths need to be found; it is possible that after the sending end has sent any slice using a certain shortest path (i.e., after the first user has received the slice), the sending end will again use the aforementioned shortest path to send another slice. Therefore, S is a positive integer less than or equal to N-2. Typically, S is less than N-2.
[0111] like Figure 5 As shown, preferably, the sending end searches for the first shortest path based on the network locations within the second specified range, including:
[0112] S20211: Obtain the second shortest path from the sending end to each of the third users within the second specified range; as mentioned above, there are K third users within the second specified range; the sending end has a first shortest path to each third user, that is, the sending end has K first paths to all the third users within the second specified range.
[0113] S20212: Take the shortest path among all the second shortest paths as the first shortest path. Among the K first paths, there must be a path with the smallest delay, and this path with the smallest delay is the second shortest path.
[0114] like Figure 6As shown, according to a third aspect of this disclosure, a model updating apparatus 3 is also provided, comprising:
[0115] First Segmentation Module 301: In response to the model update request from the first user, the sending end segments the first new model into N slices;
[0116] First sending module 302: Used for the sending end to find a path based on the network location of the first user, and send N slices through the path;
[0117] First fusion module 303: used to fuse N slices into a third new model after the first user receives N slices; the similarity between the third new model and the first new model is greater than a preset first threshold; N is a positive integer;
[0118] First update module 304: used by the first user to replace the first old model with the third new model.
[0119] like Figure 7 As shown, preferably, it also includes:
[0120] First broadcast module 305: used by the first user to broadcast the N slices to M second users within a first specified range; where M is a positive integer;
[0121] Second fusion module 306: used by the M second users to fuse the N slices respectively to obtain the third new model;
[0122] Second update module 307: used by the M second users to replace their respective first old models with the third new model.
[0123] like Figure 8 As shown, according to a fourth aspect of this disclosure, another model updating device 4 is also provided, comprising:
[0124] The second segmentation module 401 is used to respond to model update requests from K third users within the second specified range, and the sending end segments the second new model into T slices; K and T are both positive integers;
[0125] The fourth sending module 402 is used by the sending end to find the first shortest path according to the network location within the second specified range, and send T slices through the first shortest path;
[0126] Second broadcast module 403: After any one of the K third users receives at least one of the T slices, it broadcasts at least one of the T slices within the second specified range;
[0127] The second fusion module 404 is configured to respond to any one of the K third users obtaining all T slices via the broadcast method, wherein any one of the K third users fuses the T slices to obtain a fourth new model; the similarity between the fourth new model and the second new model is greater than a preset second threshold.
[0128] Third update module 405: used by any one of the K third users to replace the second old model with the fourth new model.
[0129] Preferably, the second segmentation module 401 includes: horizontal layer-by-layer segmentation, vertical cross-layer segmentation, random segmentation, segmentation based on data volume, segmentation based on function, or segmentation based on purpose.
[0130] like Figure 9 As shown, preferably, the fourth transmitting module 402 includes:
[0131] First addressing module 4021: used by the sending end to find the first shortest path based on the network location within the second specified range;
[0132] Fifth sending module 4022: used by the sending end to send any one of the T slices to any third user in the second specified range through the first shortest path;
[0133] Second addressing module 4023: used by the sending end to find the second shortest path based on the network location of any third user;
[0134] The sixth sending module 4024 is used by the sending end to send any one of the T-1 slices to any one of the third users through the second shortest path;
[0135] Repeat module 4025: used by the sending end to repeatedly use the above steps to send the remaining T-2 slices to any one of the third users through S paths; where S is a positive integer.
[0136] like Figure 10 As shown, preferably, the first addressing module 4021 includes:
[0137] The third addressing module 40211 is used to obtain the second shortest path from the sending end to each of the third users within the second specified range;
[0138] Shortest path module 40212: used to take the shortest path in all the second shortest paths as the first shortest path.
[0139] According to a fifth aspect of this disclosure, an electronic device is also provided, comprising:
[0140] At least one processor; and
[0141] A memory communicatively connected to the at least one processor; wherein,
[0142] 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 method described in any of the above technical solutions.
[0143] According to a sixth aspect of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is also provided, wherein the computer instructions are used to cause the computer to perform any one of the methods described above.
[0144] According to the seventh aspect of this disclosure, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the method described in any one of the above technical solutions.
[0145] 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.
[0146] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.
[0147] Figure 11 A schematic block diagram of an example electronic device 1100 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 illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.
[0148] like Figure 11 As shown, device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in read-only memory (ROM) 1102 or a computer program loaded from storage unit 1108 into random access memory (RAM) 1103. The RAM 1103 may also store various programs and data required for the operation of device 1100. The computing unit 1101, ROM 1102, and RAM 1103 are interconnected via bus 1104. Input / output (I / O) interface 1105 is also connected to bus 1104.
[0149] Multiple components in device 1100 are connected to I / O interface 1105, including: input unit 1106, such as keyboard, mouse, etc.; output unit 1107, such as various types of monitors, speakers, etc.; storage unit 1108, such as disk, optical disk, etc.; and communication unit 1109, such as network card, modem, wireless transceiver, etc. Communication unit 1109 allows device 1100 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0150] The computing unit 1101 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1101 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 1101 performs the various methods and processes described above, such as the model update method. For example, in some embodiments, the model update method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program may be loaded and / or installed on device 1100 via ROM 1102 and / or communication unit 1109. When the computer program is loaded into RAM 1103 and executed by the computing unit 1101, one or more steps of the model update method described above may be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to perform the model update method by any other suitable means (e.g., by means of firmware).
[0151] 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.
[0152] The 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 a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0153] 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.
[0154] 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).
[0155] 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.
[0156] 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.
[0157] 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.
[0158] 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 method for updating a model, characterized in that, The method is applied to intelligent and simplified networks, including: In response to the first user's model update request, the sending end divides the first new model into N slices; the first new model is a deep neural network model; The sending end finds a path based on the network location of the first user and sends N slices through the path; After receiving N slices, the first user merges the N slices into a third new model; the similarity between the third new model and the first new model is greater than a preset first threshold; N is a positive integer; the N slices are data packets that do not have independent execution functions. The first user replaces the first old model with the third new model; The first user broadcasts the N slices to M second users within a first specified range; M is a positive integer; the first specified range is the local area network where the first user is located, and the first user and the M second users are in the same local area network; The M second users respectively fuse the N slices to obtain the third new model; The M second users replace their respective first old models with the third new model.
2. A method for updating a model, characterized in that, The method is applied to intelligent and simplified networks, including: In response to model update requests from K third users within a second specified range, the sending end divides the second new model into T slices; K and T are both positive integers; the second new model is a deep neural network model; The sending end finds the first shortest path based on the network location within the second specified range, and sends T slices through the first shortest path; After receiving at least one of the T slices, any one of the K third users broadcasts at least one of the T slices within the second specified range; the T slices are data packets that do not have independent execution functions. In response to any one of the K third users obtaining all T slices via broadcast, any one of the K third users merges the T slices to obtain a fourth new model; the similarity between the fourth new model and the second new model is greater than a preset second threshold; Any one of the K third users replaces the second old model with the fourth new model.
3. The method according to claim 2, characterized in that, The sending end divides the second new model into T slices, including: horizontal layer-by-layer segmentation, vertical cross-layer segmentation, random segmentation, segmentation based on data volume, segmentation based on function, or segmentation based on purpose.
4. The method according to claim 2, characterized in that, The sending end searches for a first shortest path based on the network location within the second specified range, and sends T slices through the first shortest path, including: The sending end searches for the first shortest path based on the network locations within the second specified range; The sending end sends any one of the T slices to any third user within the second specified range via the first shortest path; The sending end searches for the second shortest path based on the network location of any of the third users; The sending end will send any one of the T-1 slices to any one of the third users via the second shortest path; The sending end repeats the above steps to send the remaining T-2 slices to any one of the third users through S paths; where S is a positive integer.
5. The method according to claim 4, characterized in that, The sending end searches for the first shortest path based on the network locations within the second specified range, including: Obtain the second shortest path from the sending end to each of the third users within the second specified range; The shortest path among all the second shortest paths is taken as the first shortest path.
6. A model updating apparatus, characterized in that, The device is used in intelligent and simplified networks, including: The first slicing module: In response to the model update request from the first user, the sending end slices the first new model into N slices; the first new model is a deep neural network model; First sending module: used by the sending end to find a path based on the network location of the first user, and send N slices through the path; First fusion module: used to fuse N slices into a third new model after the first user receives N slices; the similarity between the third new model and the first new model is greater than a preset first threshold; N is a positive integer; the N slices are data packets that do not have independent execution functions; First update module: used by the first user to replace the first old model with the third new model; First broadcast module: used by the first user to broadcast the N slices to M second users within a first specified range; M is a positive integer; the first specified range is the local area network where the first user is located, and the first user and the M second users are in the same local area network. The second fusion module is used by the M second users to fuse the N slices respectively to obtain the third new model; The second update module is used by the M second users to replace their respective first old models with the third new model.
7. A model updating apparatus, characterized in that, The device is used in intelligent and simplified networks, including: The second segmentation module is used to respond to model update requests from K third users within a second specified range. The sending end segments the second new model into T slices; K and T are both positive integers; the second new model is a deep neural network model. The fourth sending module is used by the sending end to find the first shortest path based on the network location within the second specified range, and to send T slices through the first shortest path; The second broadcast module is used to broadcast at least one of the T slices within the second specified range after any one of the K third users receives at least one of the T slices; the T slices are data packets that do not have independent execution functions. The second fusion module is used to respond to any one of the K third users obtaining all T slices via broadcast, and the K third users fusing the T slices to obtain a fourth new model; the similarity between the fourth new model and the second new model is greater than a preset second threshold. The third update module is used by any one of the K third users to replace the second old model with the fourth new model.
8. The apparatus according to claim 7, characterized in that, The second segmentation module includes: horizontal layer-by-layer segmentation, vertical cross-layer segmentation, random segmentation, segmentation based on data volume, segmentation based on function, or segmentation based on purpose.
9. The apparatus according to claim 7, characterized in that, The fourth sending module includes: First addressing module: used by the sending end to find the first shortest path based on the network location within the second specified range; Fifth sending module: used by the sending end to send any one of the T slices to any third user in the second specified range through the first shortest path; The second addressing module is used by the sending end to find the second shortest path based on the network location of any of the third users. The sixth sending module is used by the sending end to send any one of the T-1 slices to any one of the third users through the second shortest path; Repeat module: used by the sending end to reuse the above steps to send the remaining T-2 slices to any one of the third users through S paths; where S is a positive integer.
10. The apparatus according to claim 9, characterized in that, The first addressing module includes: The third addressing module is used to obtain the second shortest path from the sending end to each of the third users within the second specified range; Shortest path module: used to take the shortest path in all the second shortest paths as the first shortest path.
11. 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 method of any one of claims 1-5.
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 1-5.
13. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-5.
Citation Information
Patent Citations
Data transmission method and device, electronic equipment and storage medium
CN109412753A
Data transmission method and device and electronic equipment
CN111263402A
