A model propagation method and device based on semantic capability
By using nodes that need updating as senders in the intelligent simplified network, broadcasting and optimizing the model of the receiver, the problems of large differences in node semantic capabilities and low efficiency are solved, and the efficient operation of the network is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2023-05-16
- Publication Date
- 2026-04-21
AI Technical Summary
In the Zhijian Network, the semantic capabilities of some network nodes have not been updated for a long time, resulting in low efficiency and inability to meet user needs. There are also large differences in semantic capabilities between nodes.
The nodes that need to update their models and algorithms are designated as the sending end, broadcasting the original semantic information. The receiving end recovers the semantic information from its local knowledge base and sends it back. The sending end selects the optimal receiving end and requests model propagation to optimize the models and algorithms of the nodes in the network.
It enables timely updates of semantic capabilities, balances the semantic capabilities among nodes in the network, reduces redundant information, and improves network efficiency.
Smart Images
Figure CN116723543B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of communication technology, and in particular to a model propagation method and apparatus based on semantic capabilities. Background Technology
[0002] The core of semantic communication is extracting the "meaning" of the information sent by the sender. By leveraging a matching knowledge base (KB) between the sender and receiver, the semantic information can be successfully "interpreted" at the receiver. Therefore, semantic communication is essentially a communication scheme primarily based on artificial intelligence. Semantic communication is proposed as an intelligent communication scheme that focuses on the meaning of the transmitted message, rather than the precise transmission of bitstreams. Therefore, semantic information sources are compression and escaping of traditional syntactic information sources (bitstreams), which can reduce the amount of data that needs to be transmitted in the communication system and improve the efficiency of the communication system. The nodes in the intelligent simplified network are responsible for tasks such as compression and escaping of syntactic information sources and model training.
[0003] In semantic communication, the degree to which the receiver can discern the semantics recovered from the sender can be represented as the receiver's semantic capability in this type of semantic information. The nodes in the intelligent and simplified network act as senders and receivers, transmitting various types of information. As the network expands and application scenarios become more complex, significant differences emerge between network nodes. Some nodes may be more widely used in certain areas, thus training more efficient models; conversely, some nodes may not have been updated in these areas for a long time, leading to decreased efficiency or even failure to meet user needs. Summary of the Invention
[0004] In view of this, embodiments of the present invention provide a model propagation method and apparatus based on semantic capabilities to eliminate or improve one or more defects existing in the prior art, and solve the problems of low efficiency, inability to meet user needs, and large differences in semantic capabilities between nodes caused by the long-term lack of updating of the semantic capabilities of some network nodes in the intelligent and simplified network semantic communication.
[0005] On one hand, this invention provides a model propagation method based on semantic capabilities, characterized in that a deep learning model is embedded in a 6G mobile communication network to form a simplified intelligent network, the nodes in the simplified intelligent network that need to update the model and algorithm are used as sending ends, and the nodes within the broadcast range of the sending ends are used as receiving ends; the method is executed at the sending end and includes the following steps:
[0006] The original semantic information is broadcast to each receiving end; the original semantic information is extracted from the knowledge base of the sending end.
[0007] Receive the restored semantic information fed back by each receiving end; wherein, the restored semantic information is obtained by each receiving end's model from the original semantic information when each receiving end can find information similar to the original semantic information in its local knowledge base;
[0008] According to the preset evaluation criteria, the recovered semantic information that is closest to the original semantic information is selected, and the corresponding receiving end is taken as the optimal receiving end.
[0009] Send a model propagation request to the optimal receiver;
[0010] The model received from the optimal receiver broadcasts is then used to optimize the original model.
[0011] In some embodiments of the present invention, the number of nodes to be broadcast or the broadcast time are preset before broadcasting the original semantic information to each receiving end, so as to ensure the efficiency of the intelligent and simplified network.
[0012] In some embodiments of the present invention, before broadcasting the original semantic information to each receiving end, the sending end first sends a probe signal to each receiving end to detect the data format and structure that each receiving end can recognize.
[0013] In some embodiments of the present invention, the original semantic information includes semantic information from multiple knowledge bases, wherein when the multiple knowledge bases correspond to different models, it further includes:
[0014] According to the preset evaluation criteria, select multiple restored semantic information that are closest to the semantic information of multiple knowledge bases in the original semantic information, and take each receiving end corresponding to the selected restored semantic information as the optimal receiving end.
[0015] Send model propagation requests to each optimal receiver;
[0016] It receives the models broadcast by each optimal receiver and optimizes its own original models.
[0017] In some embodiments of the present invention, the method is further comprising the following steps prior to execution:
[0018] In the intelligent simplified network, each node is pre-set to a state according to a preset semantic service quality; the state includes an accept state and a reject state.
[0019] When a node can meet the preset semantic service quality requirements, it is set to the rejection state and the original semantic information sent by the corresponding sender is discarded.
[0020] When a node fails to meet the preset semantic service quality requirements, it is set to the accept state and accepts the original semantic information sent by the corresponding sender.
[0021] In some embodiments of the present invention, the method further includes:
[0022] When the sending end broadcasts the original semantic information to the designated receiving end, and the information needs to pass through other nodes in the intelligent and simplified network, the other nodes only forward the original semantic information.
[0023] In some embodiments of the present invention, the original semantic information includes node hardware requirements, and if the receiving end does not meet the node hardware requirements, the original semantic information is ignored.
[0024] In some embodiments of the present invention, the original semantic information includes node computing power and power requirements. If the receiving end does not meet the node computing power and power requirements, the original semantic information is ignored.
[0025] In some embodiments of the present invention, the preset evaluation criteria are determined according to the semantic capabilities required by the corresponding nodes, and the preset evaluation criteria include at least model computing capabilities and model recognition accuracy.
[0026] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the program is executed by a processor, it implements the steps of the semantic capability-based model propagation method as described in any of the preceding claims.
[0027] The beneficial effects of the present invention are at least as follows:
[0028] This invention provides a model propagation method and apparatus based on semantic capabilities, comprising: designating nodes whose models and algorithms need updating as senders and broadcasting their original semantic information; each receiver searching for similar information in its local knowledge base; if similar information is found, inputting the original semantic information into the receiver's model for recovery, outputting the recovered semantic information and feeding it back to the sender; otherwise, directly ignoring the original semantic information; the sender selecting the recovered semantic information closest to the original semantic information according to a preset evaluation criterion, designating its corresponding receiver as the optimal receiver, and sending a model propagation request to the optimal receiver; the optimal receiver broadcasting its own model to the network, updating the models and algorithms of other nodes in the network. The method provided by this invention can promptly update and optimize nodes whose semantic capabilities have not been updated for a long time, are inefficient, and cannot meet user needs, balancing the semantic capabilities among nodes in the network.
[0029] Furthermore, the number of nodes or the time for broadcasting by the sending end are limited to ensure the efficiency of the intelligent and simplified network and reduce redundant information in the network. At the same time, before executing the model propagation method, the receiving or rejecting state of each node in the network is set according to the semantic service quality. Nodes that can achieve the semantic service quality are set to the rejecting state, and the original semantic information sent by the sending end is discarded to reduce redundant information in the network.
[0030] Furthermore, considering that different nodes have different requirements for semantic capabilities, i.e., different requirements for model capabilities, the sending end judges the quality of recovered semantic information based on evaluation criteria such as effectiveness (model computational capability) and reliability (model recognition accuracy), and selects the optimal model accordingly.
[0031] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.
[0032] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description
[0033] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, are not intended to limit the scope of the invention. In the drawings:
[0034] Figure 1 This is a schematic diagram illustrating the steps of a model propagation method based on semantic capabilities in one embodiment of the present invention.
[0035] Figure 2 This is a flowchart illustrating the steps of a model propagation method based on semantic capabilities in one embodiment of the present invention. Detailed Implementation
[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.
[0037] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.
[0038] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.
[0039] It should also be noted that, unless otherwise specified, the term "connection" in this article can refer not only to a direct connection, but also to an indirect connection involving an intermediary.
[0040] In the following description, embodiments of the invention will be illustrated with reference to the accompanying drawings. In the drawings, the same reference numerals represent the same or similar parts, or the same or similar steps.
[0041] It should be emphasized here that the step markers mentioned below are not a limitation on the order of the steps, but should be understood as meaning that the steps can be executed in the order mentioned in the embodiments, or in a different order than in the embodiments, or several steps can be executed simultaneously.
[0042] To address the issues of low efficiency, inability to meet user needs, and significant differences in semantic capabilities between nodes caused by the long-term lack of updates to the semantic capabilities of some network nodes in the intelligent simplified network semantic communication, this invention provides a model propagation method based on semantic capabilities. This method uses nodes in the intelligent simplified network that require model and algorithm updates as senders, and nodes within the broadcast range of the senders as receivers. Figure 1 As shown, this method is executed at the sending end and includes the following steps S101 to S105:
[0043] Step S101: Broadcast the original semantic information to each receiving end. The original semantic information is extracted from the knowledge base of the sending end.
[0044] Step S102: Receive the restored semantic information fed back by each receiving end. The restored semantic information is obtained by each receiving end's model from the original semantic information, provided that each receiving end can find information similar to the original semantic information in its local knowledge base.
[0045] Step S103: According to the preset evaluation criteria, select the recovered semantic information that is closest to the original semantic information, and take its corresponding receiver as the optimal receiver.
[0046] Step S104: Send a model propagation request to the optimal receiver.
[0047] Step S105: Receive the model broadcast by the optimal receiver and optimize its own original model.
[0048] Intelligent and simplified networks represent an evolution of the sixth-generation (6G) mobile communication network. "Intelligent and simplified" can be understood as intelligent evolution and inherent simplicity. In intelligent and simplified networks, deep learning models are embedded into the 6G mobile communication network to provide users with customized services and experiences through real-time monitoring and analysis of network traffic and data.
[0049] In the intelligent simplified network, any node can act as a sender or receiver. When a node needs to update and optimize its own model and algorithm, it can act as a sender and send a semantic capability update request to each receiver. The sender uses a broadcast method, so there are multiple receivers.
[0050] In step S101, the sending end extracts raw semantic information from the corresponding knowledge base according to the model to be updated. For example, when the model is a handwritten Arabic numeral 0-9 recognition model, its corresponding knowledge base should contain at least multiple handwritten images of Arabic numerals 0-9; when the model is a text summary model, its corresponding knowledge base should contain multiple text segments; when the model is a video semantic information extraction model, its corresponding knowledge base should contain multiple video segments. The raw semantic information is extracted from the sending end's knowledge base and tagged. The raw semantic information can be adjusted according to the size of the knowledge base to achieve the desired effect with as little raw semantic information as possible.
[0051] The sending end broadcasts the original semantic information to each receiving end. For example, the original semantic information consists of five handwritten images of each of the Arabic numerals 0 to 9, and the number of images can be adjusted according to the actual situation.
[0052] In some embodiments, in order to ensure the efficiency of the intelligent simplified network, the number of broadcast nodes or the broadcast time can be preset, and the redundancy information in the intelligent simplified network can be controlled by the demand size, thereby reducing the possibility of network congestion caused by model propagation in the future.
[0053] In some embodiments, considering that the data processing methods of each node in the intelligent simplified network are not the same, before the sender broadcasts the original semantic information to each receiver, the sender first sends a probe signal to each receiver to detect the data format and structure that each receiver can recognize, so that the original semantic information subsequently sent to each receiver is valid and avoids model propagation failure due to data format and structure problems.
[0054] In some embodiments, considering that the hardware of each node in the intelligent simplified network is not the same, the original semantic information may include node hardware requirements. When the receiving end does not meet the node hardware requirements, the receiving end automatically ignores the original semantic information.
[0055] In some embodiments, considering that the computing power and power requirements of each node in the intelligent simplified network are not the same, the original semantic information may include the node computing power and power requirements. When the receiving end does not meet the node computing power and power requirements, the receiving end automatically ignores the original semantic information.
[0056] In some embodiments, if the sender broadcasts the original semantic information to a designated receiver and it needs to pass through other nodes in the intelligent simplified network, the other nodes do not need to understand and recover the original semantic information; they only need to forward the original semantic information to the node of the designated receiver.
[0057] In step S102, after receiving the original semantic information, each receiving end searches for similar information in its local knowledge base. If similar information can be found, it means that the receiving end can understand the original semantic information. The original semantic information is then input into the receiving end's model for recovery, and the recovered semantic information is output and fed back to the sending end. For example, if the handwritten image of the Arabic numeral 8 in the original semantic information is input into the receiving end's model, the model can output the Arabic numeral 8 (the bitstream of the Arabic numeral 8 may be represented as 1000). The output Arabic numeral 8 is the recovered semantic information obtained by the model. If no similar information can be found, it means that the receiving end cannot understand the original semantic information, and the original semantic information is automatically ignored without any response.
[0058] In step S103, after receiving the recovered semantic information fed back by each receiver, the transmitting end determines the receiver with the best recovery capability according to the preset evaluation criteria, that is, the receiver whose recovered semantic information is closest to the original semantic information, and takes it as the optimal receiver.
[0059] In some embodiments, different nodes have different needs for semantic information; some nodes prioritize effectiveness, while others prioritize reliability. Therefore, the requirements for the model processing semantic information also differ. When propagating the model, the model to be propagated should be determined according to these different needs. When effectiveness is dominant, the computational complexity of the model is more important; when reliability is dominant, the semantic information recognition accuracy of the model is more important.
[0060] In step S104, the sending end sends a model propagation request to the optimal receiving end.
[0061] In step S105, after the optimal receiver receives the model propagation request, it broadcasts its own model to the intelligent simplified network. The sender and other non-optimal receivers can update and optimize their original models.
[0062] For example, such as Figure 2As shown, node A is the sender, and nodes B, C, and D are the receivers, respectively. Step A corresponds to step S101, step B corresponds to step S102, step C corresponds to step S103, step D corresponds to step S104, and step E corresponds to step S105. Specifically, in step A, the sending end broadcasts the original semantic information to receiving ends B, C, and D; in step B, receiving ends B and C recover the original semantic information and each generate recovered semantic information to feed back to the sending end. Since receiving end D does not contain information similar to the original semantic information in its local knowledge base, it automatically ignores the original semantic information and does not respond; in step C, the sending end compares the recovered semantic information fed back by receiving ends B and C, selects the one most similar to the original semantic information according to the preset evaluation criteria, and finally determines that the node with the best recovery capability is receiving end B; in step D, the sending end sends a model propagation request to receiving end B; in step E, the optimal receiving end (i.e., receiving end B) broadcasts its model to the network, and both the sending end and receiving end C can update and optimize their own original models.
[0063] In some embodiments, before performing the semantic capability-based model propagation method, each node in the simplified intelligent network can pre-set a state according to a preset semantic quality of service (QoS). For example, the state can include an accept state and a reject state. When a node can meet the preset semantic quality of service requirements, that is, when it does not need to update or optimize its own model, it can refuse to propagate the model. Specifically, the node is set to the reject state, and when it receives the original semantic information sent by the sender, it is directly discarded without any processing or response, which can reduce redundant information in the simplified intelligent network. When a node cannot meet the preset semantic quality of service requirements, that is, when it needs to update or optimize its own model, it can accept the model propagation. Specifically, the node is set to the accept state, and when it receives the original semantic information sent by the sender, it is processed normally according to the steps of the semantic capability-based model propagation method provided by the present invention.
[0064] In some embodiments, when the sending end has multiple different knowledge bases, the original semantic information sent can simultaneously contain a set of unrelated semantic information from multiple knowledge bases. For example, if the sending end contains three knowledge bases: the first knowledge base contains images of cats and dogs, the second knowledge base contains images of chickens and ducks, and the third knowledge base contains images of flowers and plants, then the original semantic information sent by the sending end can contain 5 images of cats and dogs, 5 images of chickens and ducks, and 5 images of flowers and plants.
[0065] When these multiple knowledge bases correspond to one model, only one model needs to be propagated when using the semantic capability-based model propagation method; when these multiple knowledge bases correspond to multiple models, multiple models need to be propagated when using the semantic capability-based model propagation method.
[0066] Each knowledge base corresponds to at least one model, but a model may correspond to more than one knowledge base. Because some models are more powerful, they can correspond to multiple knowledge bases. For example, if model X can simultaneously recognize cats and dogs, chickens and ducks, and flowers and plants, then model X can simultaneously correspond to the first knowledge base, the second knowledge base, and the third knowledge base mentioned above.
[0067] When the original semantic information contains semantic information from multiple knowledge bases, and the multiple knowledge bases correspond to different models, after executing steps S101 to S102, according to the preset evaluation criteria, select multiple restored semantic information that are closest to the semantic information of the multiple knowledge bases in the original semantic information. If the corresponding multiple models belong to one receiver, that receiver is regarded as the optimal receiver, and the optimal receiver needs to broadcast and send the corresponding multiple models; if the corresponding multiple models belong to multiple receivers, each of the corresponding receivers is regarded as the optimal receiver, and each optimal receiver broadcasts and sends the corresponding model.
[0068] The present invention will be further described below with reference to a specific embodiment:
[0069] Node A needs to update its handwritten Arabic numeral recognition capability for 0-9, i.e., update its handwritten Arabic numeral recognition model for 0-9. Node A extracts 5 images of each of the 5 handwritten Arabic numerals from its local knowledge base and packages them as raw semantic information. Node A, as the sender, broadcasts the raw semantic information to other nodes. Based on a preset number of broadcasting nodes or a preset broadcast time, Node A sends the raw semantic information to Nodes B, C, and D. After receiving the raw semantic information, Nodes B, C, and D retrieve similar information from their local knowledge bases to determine if they can understand the raw semantic information. If Node D does not retrieve similar information, it automatically ignores the raw semantic information and does not respond. Node C retrieves similar information, inputs the original semantic information into its respective model, generates corresponding restored semantic information, and feeds it back to Node A. Based on the accuracy of the recognition, Node A determines that Node B's restoration ability is better than Node A's, that is, Node B's restored semantic information is closer to the original semantic information. Node A sends a propagation request for its Arabic numeral 0-9 handwritten recognition model to Node B. After receiving Node A's propagation request, Node B broadcasts its own Arabic numeral 0-9 handwritten recognition model to the network, thereby optimizing the Arabic numeral 0-9 handwritten recognition models of Node A and Node C in the network and improving the Arabic numeral 0-9 handwritten recognition capabilities of Node A and Node C.
[0070] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a semantic capability-based model propagation method.
[0071] Corresponding to the above method, the present invention also provides an apparatus comprising a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor executing the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the apparatus performs the steps of the method as described above.
[0072] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the aforementioned edge computing server deployment method. The computer-readable storage medium can be a tangible storage medium, such as random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disks, removable storage disks, CD-ROMs, or any other form of storage medium known in the art.
[0073] In summary, this invention provides a model propagation method and apparatus based on semantic capabilities, comprising: designating nodes whose models and algorithms need updating as senders and broadcasting their original semantic information; each receiver searching for similar information in its local knowledge base; if similar information is found, inputting the original semantic information into the receiver's model for recovery, outputting the recovered semantic information and feeding it back to the sender; otherwise, directly ignoring the original semantic information; the sender selecting the recovered semantic information closest to the original semantic information according to a preset evaluation criterion, and designating its corresponding receiver as the optimal receiver, and sending a model propagation request to the optimal receiver; the optimal receiver broadcasting its own model to the network, updating the models and algorithms of other nodes in the network. The method provided by this invention can promptly update and optimize nodes whose semantic capabilities have not been updated for a long time, are inefficient, and cannot meet user needs, balancing the semantic capabilities among nodes in the network.
[0074] Furthermore, the number of nodes or the time for broadcasting by the sending end are limited to ensure the efficiency of the intelligent and simplified network and reduce redundant information in the network. At the same time, before executing the model propagation method, the receiving or rejecting state of each node in the network is set according to the semantic service quality. Nodes that can achieve the semantic service quality are set to the rejecting state, and the original semantic information sent by the sending end is discarded to reduce redundant information in the network.
[0075] Furthermore, considering that different nodes have different needs for semantic information, i.e., different requirements for model capabilities, the sending end judges the quality of recovered semantic information based on evaluation criteria such as effectiveness (model computing power) and reliability (model recognition accuracy), and selects the optimal model accordingly.
[0076] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.
[0077] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.
[0078] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.
[0079] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A semantic capability-based model propagation method, characterized in that, A deep learning model is embedded in a 6G mobile communication network to form a simplified intelligent network. Nodes in this network that require model and algorithm updates are designated as transmitters, and nodes within the broadcast range of these transmitters are designated as receivers. The method is executed at the transmitter and includes the following steps: The original semantic information is broadcast to each receiving end; the original semantic information is extracted from the knowledge base of the sending end. Receive the restored semantic information fed back by each receiving end; wherein, the restored semantic information is obtained by each receiving end's model from the original semantic information when each receiving end can find information similar to the original semantic information in its local knowledge base; According to the preset evaluation criteria, the recovered semantic information that is closest to the original semantic information is selected, and the corresponding receiving end is taken as the optimal receiving end. Send a model propagation request to the optimal receiver; The model received from the optimal receiver broadcasts is then used to optimize the original model. 2.The semantic capability-based model propagation method according to claim 1, characterized in that, Before broadcasting the original semantic information to each receiving end, the number of nodes to be broadcast or the broadcast time are preset to ensure the efficiency of the intelligent and simplified network. 3.The semantic capability-based model propagation method of claim 1, wherein, Before broadcasting the original semantic information to each receiving end, the sending end first sends a probe signal to each receiving end to detect the data format and structure that each receiving end can recognize. 4.The semantic capability-based model propagation method of claim 1, wherein, The original semantic information includes semantic information from multiple knowledge bases. When the multiple knowledge bases correspond to different models, it also includes: According to the preset evaluation criteria, select multiple restored semantic information that are closest to the semantic information of multiple knowledge bases in the original semantic information, and take each receiving end corresponding to the selected restored semantic information as the optimal receiving end. Send model propagation requests to each optimal receiver; It receives the models broadcast by each optimal receiver and optimizes its own original models.
5. The semantic capability-based model propagation method according to claim 1, wherein, Before executing the method, the following is also included: In the intelligent simplified network, each node is pre-set to a state according to a preset semantic service quality; the state includes an accept state and a reject state. When a node can meet the preset semantic service quality requirements, it is set to the rejection state and the original semantic information sent by the corresponding sender is discarded. When a node fails to meet the preset semantic service quality requirements, it is set to the accept state and accepts the original semantic information sent by the corresponding sender.
6. The semantic capability-based model propagation method according to claim 1, wherein, The method further includes: When the sending end broadcasts the original semantic information to the designated receiving end, and the information needs to pass through other nodes in the intelligent and simplified network, the other nodes only forward the original semantic information.
7. The semantic capability-based model propagation method according to claim 1, wherein, The original semantic information includes node hardware requirements. If the receiving end does not meet the node hardware requirements, the original semantic information is ignored. 8.The semantic capability based model propagation method according to claim 1, wherein, The original semantic information includes node computing power and power requirements. If the receiving end does not meet the node computing power and power requirements, the original semantic information is ignored. 9.The semantic capability-based model propagation method of claim 1, wherein, The preset evaluation criteria are determined based on the semantic capabilities required by the corresponding nodes, and the preset evaluation criteria include at least the model's computational capability and the model's recognition accuracy.
10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1 to 9.