Method for inhibiting collapse of mode in zigzag network, electronic device, and storage medium
By performing pattern collapse detection and restricting the propagation of pattern collapse models in the intelligent simplified network, the problem of data monotony in the intelligent simplified network is solved, and the network's communication performance is improved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING UNIV OF POSTS & TELECOMM
- Filing Date
- 2023-07-14
- Publication Date
- 2026-06-09
Smart Images
Figure CN116886547B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of semantic communication technology for intelligent and simplified networks, and in particular to a method for suppressing pattern collapse in intelligent and simplified networks, an electronic device, and a storage medium. Background Technology
[0002] Given that some nodes in the Intelligent Simplified Network are intelligent nodes, these nodes can be highly integrated with neural networks, storing neural network models and sharing them as needed. In the Intelligent Simplified Network, traditional data sharing among network nodes transforms into neural network model sharing. Each network node shifts from storing data to storing neural network models, and autonomously learns and updates these models based on internal and external network requirements. This achieves cognition and learning based on network-native intelligence, forming a highly autonomous network with swarm intelligence.
[0003] The essence of pattern collapse is the inconsistency between the distribution of real data and the distribution of data generated after model training. In other words, a neural network model exhibiting pattern collapse generates only a single type of sample after training, or only covers a small portion, failing to meet the diversity requirements of the original real data. For example, if the training dataset of a neural network model contains multiple categories of data, but after training, the model only generates data for one or a few categories, even if the generated category data is of high quality and accuracy, the output category data still cannot cover all categories corresponding to the training dataset. This fails to meet the initial training intent and purpose of the neural network model, thus exhibiting pattern collapse.
[0004] Neural network models with pattern collapse in network nodes in the intelligent simplified network are prone to generating data that is monotonous and fails to meet the diverse needs of the original real data applications. Summary of the Invention
[0005] In view of this, embodiments of the present invention provide a mode collapse suppression method, electronic device and storage medium for intelligent and simplified networks to eliminate or improve one or more defects existing in the prior art.
[0006] One aspect of the present invention provides a method for suppressing mode collapse in intelligent simplified networks, comprising:
[0007] Based on the target functional requirements of any network node in the intelligent simplified network, search for neural network models that satisfy the target functional requirements among the other network nodes of the intelligent simplified network;
[0008] The neural network model is subjected to mode collapse detection using the semantic pilot test data set to determine whether mode collapse exists in the neural network model; the semantic pilot test data set includes test samples of all sample types corresponding to the neural network model;
[0009] The propagation of the neural network model that exhibits pattern collapse is restricted among the network nodes of the intelligent simplified network.
[0010] In some embodiments of the present invention, the step of searching for a neural network model that satisfies the target functional requirements of any network node in the intelligent simplified network includes:
[0011] Receive model requests sent by each network node in the intelligent simplified network;
[0012] Extract and store the functional requirements corresponding to each network node from each of the model requests;
[0013] The various functional requirements stored locally are thus taken as the current target functional requirements, and a neural network model corresponding to the target functional requirement is searched among the network nodes of the network.
[0014] In some embodiments of the present invention, the step of performing mode collapse detection on the neural network model using the semantic pilot test data set of the neural network model includes:
[0015] The upper-layer server requests the semantic pilot test data set of the neural network model, wherein the upper-layer server periodically updates the test samples in the semantic pilot test data set corresponding to each neural network model;
[0016] Each test sample containing its own type label in the semantic pilot test data set is then input into the neural network model so that the neural network model outputs the sample type corresponding to each test sample.
[0017] Based on the consistency comparison results between the sample types and type labels corresponding to each test sample in the semantic pilot test data set output by the neural network model, it is determined whether the neural network model has mode collapse.
[0018] In some embodiments of the present invention, the step of performing mode collapse detection on the neural network model using the semantic pilot test data set of the neural network model further includes:
[0019] The semantic pilot test data set is temporarily stored locally.
[0020] In some embodiments of the present invention, after searching for a neural network model that satisfies the target functional requirements among other network nodes of the intelligent simplified network, the method further includes:
[0021] If it is determined that the neural network model for the target functional requirement is currently marked as exempt from inspection, then it is determined whether the number of times the neural network model has been inspected within a preset time period is less than the sum of the number of times exempted from inspection specified by the exemption mark and the threshold number of times exempted from inspection conditions. If so, then the neural network model is subjected to mode collapse detection exemption processing within the current detection cycle.
[0022] In some embodiments of the present invention, after searching for a neural network model that satisfies the target functional requirements among other network nodes of the intelligent simplified network, the method further includes:
[0023] If it is determined that the neural network model for the target functional requirement is not currently marked for exemption from inspection, then it is determined whether the number of times the neural network model has been inspected within a preset time period is greater than the threshold number of times the exemption condition is set. If so, then an exemption mark is set for the neural network model so that the neural network model can undergo mode collapse detection exemption processing within the number of exemptions specified by the exemption mark.
[0024] In some embodiments of the present invention, restricting the propagation of the neural network model exhibiting pattern collapse among the network nodes of the intelligent simplified network includes:
[0025] For neural network models exhibiting pattern collapse, a restriction label is applied to the neural network model, and the information is reported to the upper-layer server and the network node storing the neural network model. This is to prevent the neural network model with the restriction label from propagating among the network nodes of the intelligent simplified network within at least one preset detection period.
[0026] In some embodiments of the present invention, restricting the propagation of the neural network model exhibiting mode collapse among the network nodes of the intelligent simplified network further includes:
[0027] For neural network models that meet the preset collapse identification conditions, the neural network model is marked for deletion. The neural network model with the deletion mark is reported to the upper-layer server and the network node storing the neural network model, and the neural network model is deleted from each network node.
[0028] Another aspect of the present invention provides an electronic device including a processor and a memory, wherein the memory stores computer instructions, and the processor is configured to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device implements the steps of the mode collapse suppression method of the intelligent simplified network described above.
[0029] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the mode collapse suppression method of the intelligent simplified network described above.
[0030] The pattern collapse suppression method for intelligent simplified networks of the present invention detects pattern collapse in neural network models that meet functional requirements in a timely manner; and effectively reduces communication errors caused by neural network models with pattern collapse by restricting the propagation of neural network models with pattern collapse in intelligent simplified networks, thereby effectively improving the communication performance of intelligent simplified networks.
[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 provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.
[0034] Figure 1 Flowchart for pattern collapse suppression in intelligent and simplified networks.
[0035] Figure 2 This is a diagram of the model request passing structure in one embodiment of the present invention.
[0036] Figure 3 This is a structural diagram of the semantic pilot test data set in one embodiment of the present invention.
[0037] Figure 4 This is a structural diagram of mode collapse detection in one embodiment of the present invention. Detailed Implementation
[0038] 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.
[0039] 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.
[0040] 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.
[0041] 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.
[0042] 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.
[0043] As mentioned in the background technology, each network node in the intelligent simplified network stores multiple different neural network models. Each network node can call and copy the neural network models stored in other network nodes according to its own functional requirements. This can save storage space and expand the functional coverage of each network node. However, if the neural network model stored in a network node of the intelligent simplified network has pattern collapse, the continuous sharing and copying of this neural network model among the network nodes can easily exacerbate the propagation of neural network models with pattern collapse in the intelligent simplified network, thereby reducing the reliability of the intelligent simplified network and reducing its network performance. To address the communication errors caused by the propagation of neural network models with pattern collapse in the intelligent simplified network, this invention provides a pattern collapse detection device, an electronic device, and a storage medium for the intelligent simplified network. By performing pattern collapse detection on the neural network model before processing the data to be processed, only sharing and propagation of neural network models that have passed pattern collapse detection are allowed, thereby effectively reducing communication errors caused by neural network models with pattern collapse.
[0044] This invention provides a mode collapse suppression method based on intelligent simplified networks, such as... Figure 1 As shown, steps S110-S130 are included:
[0045] In step S110, based on the target functional requirements of any network node in the intelligent simplified network, a neural network model that satisfies the target functional requirements is searched among the other network nodes of the intelligent simplified network.
[0046] In step S110 above, the intelligent simplified network includes multiple network nodes that perform data processing functions, intermediate nodes that communicate with each network node, and an upper-layer server for each intermediate node. Each network node stores various neural network models, such as multiple neural network models for processing various data types including digital data, image data, voice data, and text data. Correspondingly, the target functional requirements of each network node in the intelligent simplified network may include: directly obtaining the corresponding data type of the input digital data, image data, voice data, or text data based on the input data.
[0047] In one or more embodiments of this application, when a network node in the intelligent simplified network receives a model request to call a neural network model for a target functional requirement, and the network node does not have a neural network model for the target functional requirement, the intermediate node in the intelligent simplified network executes the above step S110, which specifically includes: receiving model requests issued by each network node in the intelligent simplified network; extracting and storing the functional requirements corresponding to each network node from each model request; taking each of the functional requirements stored locally as the current target functional requirement, and searching for the neural network model corresponding to the target functional requirement in each network node of the intelligent simplified network.
[0048] In one or more embodiments of this application, the intelligent simplified network includes multiple network nodes, a router serving as an intermediate node among the network nodes, and an upper-layer server for the router. The intelligent simplified network performs the above-described step S110, as follows: Figure 2 As shown, a network node S in this intelligent simplified network sends a model request carrying the functional requirement f of the network node to the corresponding router. The functional requirement f is: to generate animal image data based on semantic features, such as cats, dogs, and rabbits. Upon receiving the model request, the router searches among its corresponding network nodes for the neural network model F corresponding to the functional requirement of the model request.
[0049] In step S120, the neural network model is subjected to mode collapse detection using the semantic pilot test data set of the neural network model in order to determine whether the neural network model has mode collapse.
[0050] The semantic pilot test data set includes test samples of all sample types corresponding to the neural network model. The semantic pilot test data sets corresponding to each neural network model stored in each network node of the intelligent simplified network are all stored in the upper-layer server. Furthermore, to ensure the generalization of pattern collapse detection for the neural network model, the upper-layer server periodically updates the test samples in the semantic pilot test data sets corresponding to each neural network model. After the neural network model undergoes pattern collapse detection, the semantic pilot test data set corresponding to that neural network model can be directly stored in the local intermediate node. If a neighboring intermediate node also needs to access the semantic pilot test data set, the local intermediate node can directly send the semantic pilot test data set to that neighboring router.
[0051] Corresponding to the above embodiments, when the intermediate node of the intelligent simplified network receives model requests from various network nodes, the intermediate node executes the above step S120, which specifically includes: requesting the semantic pilot test data set of the neural network model from the upper-layer server; and inputting each test sample containing its own type label in the semantic pilot test data set into the neural network model, so that the neural network model outputs the sample type corresponding to each test sample; and determining whether the neural network model has mode collapse based on the consistency comparison result between the sample type and type label corresponding to each test sample in the semantic pilot test data set output by the neural network model. If the sample type corresponding to the test sample output by the neural network model is inconsistent with the type label of the test sample, then the neural network model has mode collapse and fails the mode collapse detection of the current detection period; otherwise, the neural network model does not have mode collapse and fails the mode collapse detection of the current detection period.
[0052] Furthermore, for each neural network model stored in each network node of the intelligent simplified network, a pattern collapse detection step must be performed before the neural network model is used to process the current input data, and the number of times the neural network model performs pattern collapse detection is recorded within a preset time period. If a neural network model passes the pattern collapse detection multiple times consecutively, reaching a preset threshold for the number of times it can be exempted from detection, an exemption flag is set in the neural network model to instruct it to perform pattern collapse detection exemption processing within the number of exemptions specified by the exemption flag in the current detection cycle. Accordingly, in one or more embodiments of this application, after executing step S110 above, it is necessary to first determine whether the searched neural network model has an exemption mark. If it is determined that the neural network model for the target functional requirement currently has an exemption mark, it is determined whether the number of times the neural network model has been detected within a preset time period is less than the sum of the exemption number specified by the exemption mark and the exemption condition number threshold. If yes, then the neural network model is subjected to pattern collapse detection exemption processing within the current detection cycle; if no, then the exemption mark of the neural network model is deleted, and the pattern collapse detection shown in step S120 above continues to be executed. If it is determined that the neural network model for the functional requirement currently does not have an exemption mark, it is determined whether the number of times the neural network model has been detected within a preset time period is greater than the exemption condition number threshold. If yes, then an exemption mark is set for the neural network model so that the neural network model is subjected to pattern collapse detection exemption processing within the exemption number specified by the exemption mark; if no, then the pattern collapse detection shown in step S120 above is executed directly.
[0053] Corresponding to the above embodiments, step S120 is performed in a simplified network comprising multiple network nodes, a router serving as an intermediate node for each network node, and an upper-layer server of the router. Figure 3 As shown, the satellite node P storing the neural network model copies and uploads the neural network model F to the router; the router requests the semantic pilot test data set corresponding to the model request from the upper-layer server; the upper-layer server then sends the semantic pilot test data set to the router. Figure 4 As shown, the router inputs each test sample containing its own type label from the semantic pilot test data set into the neural network model F, so that the neural network model F outputs the sample type corresponding to each test sample. If the output sample type of each test sample is inconsistent with the type label contained in each test sample, it is considered that the neural network model F has mode collapse; otherwise, it is considered that the neural network model F does not have mode collapse. The test samples include test samples containing type label a (cat), type label b (dog), and type label c (rabbit).
[0054] Step S130: Restrict the propagation of the neural network model with mode collapse among the network nodes of the intelligent simplified network.
[0055] Corresponding to the above embodiments, when the intermediate node of the intelligent simplified network completes the pattern collapse detection of the neural network model, the intermediate node executes the above step S130, specifically including: for neural network models that are determined to have pattern collapse, restricting the propagation of the neural network model among the network nodes of the intelligent simplified network; for neural network models that are determined not to have pattern collapse, copying the neural network model and forwarding it to the network node that sent its corresponding model request, so as to complete the sharing and propagation of the neural network model among the network nodes of the intelligent simplified network. The restriction on the propagation of the neural network model among the network nodes of the intelligent simplified network includes: for neural network models that have failed the collapse pattern detection less than a threshold number of times the pattern collapse determination condition is met, prohibiting the sharing and propagation of the neural network model among the network nodes of the intelligent simplified network only within at least one preset detection period, including the current detection period; for neural network models that have failed the collapse pattern detection more than a threshold number of times the pattern collapse determination condition is met, directly setting a deletion flag in the neural network model to instruct each network node storing the neural network model in the intelligent simplified network to directly delete the neural network model.
[0056] Accordingly, in one or more embodiments of this application, after performing the pattern collapse detection shown in step S120 above, it is necessary to first determine whether the neural network model passes the pattern collapse detection. If yes, it is determined that the neural network model does not have pattern collapse, and the neural network model is forwarded and copied to the network node that issued the model request corresponding to the neural network model, so as to complete the sharing and propagation of the neural network model among the network nodes of the intelligent simplified network within the current detection period. If no, it is determined that the neural network model has pattern collapse, and step S130 above is performed to restrict the propagation of the neural network model with pattern collapse among the network nodes of the intelligent simplified network. During the execution of step S130 above, it is necessary to determine whether the number of consecutive failures of the pattern collapse detection in the neural network model is less than the threshold number of the pattern collapse identification condition. If so, it is considered that the neural network model has a pattern collapse within the current detection period. The neural network model with the stored pattern collapse is marked with a restriction and reported to the upper-level server of the current intermediate node and the network node storing the neural network model, so as to prevent the neural network model with the restriction mark from propagating among the network nodes of the intelligent network within at least one preset detection period. If not, it indicates that the neural network model has reached the preset pattern collapse identification condition. The neural network model that has reached the preset pattern collapse identification condition is marked with a deletion mark. The neural network model with the deletion mark is reported to the upper-level server of the current intermediate node and the network node storing the neural network model, and the neural network model in the corresponding network node is deleted.
[0057] This invention first performs pattern collapse detection on the required neural network model, thereby promptly identifying neural network models with pattern collapse and restricting the propagation of such neural network models among the network nodes of the intelligent simplified network; thus effectively reducing communication errors caused by pattern collapse of neural network models and improving the communication performance of the intelligent simplified network.
[0058] The present invention also provides an intermediate node for implementing the mode collapse suppression method of the intelligent simplified network. Specifically, the embodiment of the intermediate node can execute the processing flow of the embodiment of the mode collapse suppression method of the intelligent simplified network in the above embodiment. Its function will not be repeated here, but can be referred to the detailed description of the embodiment of the mode collapse suppression method of the intelligent simplified network above.
[0059] This invention also provides a simplified intelligent network for modal collapse suppression, comprising network nodes, intermediate nodes, and an upper-layer server. Each network node stores multiple neural network models and sends model requests corresponding to current functional requirements to the intermediate nodes. The intermediate nodes, based on the model requests received from the network nodes, request a semantic pilot test data set for the corresponding neural network model from the upper-layer server and use the semantic pilot test data set to perform modal collapse detection on the neural network model. The upper-layer server stores the semantic pilot test data sets corresponding to each neural network model and periodically updates the sample data in the semantic pilot test data sets. The embodiments of the simplified intelligent network can specifically execute the processing flow of the embodiments of the modal collapse suppression method of the simplified intelligent network described above. Its functions will not be repeated here, but can be referred to the detailed description of the embodiments of the modal collapse suppression method of the simplified intelligent network described above.
[0060] Corresponding to the above method, the present invention also provides an electronic device, which includes a computer device, the computer device including a processor and a memory, the memory storing computer instructions, the processor being used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the electronic device implements the steps of the mode collapse suppression method of the intelligent simplified network as described above.
[0061] 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 mode collapse suppression method for intelligent simplified networks. 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.
[0062] 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.
[0063] 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.
[0064] 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.
[0065] 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 method for suppressing pattern collapse in intelligent simplified networks, characterized in that, include: Based on the target functional requirements of any network node in the intelligent simplified network, search for neural network models that satisfy the target functional requirements among the other network nodes of the intelligent simplified network; The upper-layer server requests the semantic pilot test data set of the neural network model, wherein the semantic pilot test data set includes test samples of all sample types corresponding to the neural network model; the test samples carry their respective corresponding type labels; the upper-layer server periodically updates the test samples in the semantic pilot test data set corresponding to each neural network model; Each test sample containing its own type label in the semantic pilot test data set is then input into the neural network model so that the neural network model outputs the sample type corresponding to each test sample. Based on the consistency comparison results between the sample types and type labels corresponding to each test sample in the semantic pilot test data set output by the neural network model, it is determined whether the neural network model has mode collapse. By comparing the sample type output by the neural network model with the type label of the test sample, if the output of the neural network model for the sample type corresponding to the test sample in the semantic pilot test data set is inconsistent with the corresponding type label, it is determined that the neural network model has mode collapse. For neural network models exhibiting pattern collapse, a restriction label is applied to the neural network model, and the information is reported to the upper-layer server and the network node storing the neural network model. This is to prevent the neural network model with the restriction label from propagating among the network nodes of the intelligent simplified network within at least one preset detection period.
2. The mode collapse suppression method according to claim 1, characterized in that, The step of searching for a neural network model that satisfies the target functional requirements of any network node in the intelligent simplified network, among other network nodes in the intelligent simplified network, includes: Receive model requests sent by each network node in the intelligent simplified network; Extract and store the functional requirements corresponding to each network node from each of the model requests; Each of the aforementioned functional requirements stored locally is taken as the current target functional requirement, and the neural network model corresponding to the target functional requirement is searched in each network node of the intelligent simplified network.
3. The mode collapse suppression method according to claim 1, characterized in that, Performing mode collapse detection on the neural network model using the semantic pilot test data set of the neural network model further includes: The semantic pilot test data set is temporarily stored locally.
4. The mode collapse suppression method according to claim 1, characterized in that, After searching for a neural network model that meets the target functional requirements among the other network nodes of the intelligent simplified network, the process further includes: If it is determined that the neural network model for the target functional requirement is currently marked as exempt from inspection, then it is determined whether the number of times the neural network model has been inspected within a preset time period is less than the sum of the number of times exempted from inspection specified by the exemption mark and the threshold number of times exempted from inspection conditions. If so, then the neural network model is subjected to mode collapse detection exemption processing within the current detection cycle.
5. The mode collapse suppression method according to claim 1, characterized in that, After searching for a neural network model that meets the target functional requirements among the other network nodes of the intelligent simplified network, the process further includes: If it is determined that the neural network model for the target functional requirement is not currently marked for exemption from inspection, then it is determined whether the number of times the neural network model has been inspected within a preset time period is greater than the threshold for the number of times the exemption condition is met. If so, then an exemption mark is set for the neural network model so that the neural network model can undergo mode collapse detection exemption processing within the number of exemptions specified by the exemption mark.
6. The mode collapse suppression method according to claim 1, characterized in that, Restricting the propagation of the neural network model exhibiting pattern collapse among the network nodes of the intelligent simplified network further includes: For neural network models that meet the preset collapse identification conditions, the neural network model is marked for deletion. The neural network model with the deletion mark is reported to the upper-layer server and the network node storing the neural network model, and the neural network model is deleted from each network node.
7. An electronic device comprising a processor and a memory, characterized in that, The memory stores computer instructions, and the processor executes the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device implements the steps of the mode collapse suppression method of the intelligent simplified network as described in any one of claims 1 to 6.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the steps of the mode collapse suppression method for the intelligent simplified network as described in any one of claims 1 to 6.