Model providing method, model obtaining method, information obtaining method, information processing method and network equipment

Abstract

The invention provides a model providing and obtaining method, an information obtaining and processing method and network equipment, and relates to the technical field of communication. The model providing method comprises the steps that a model request from a model training network element is received, and the model request comprises at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information and sample related information; an available model list is fed back to the model training network element according to the model request, and models in the available model list are matched with the model request.

Description

Technical Field

[0001] The present disclosure relates to the field of communication technology, and in particular to a model providing, acquiring, information acquiring and processing method and a network device. Background Art

[0002] In large-scale mobile communication networks, AI service network elements such as model training and reasoning are expected to be deployed in a distributed manner. In related technologies, each time the model training network element of the AI ​​service receives a new model requirement, it checks whether the local existing model meets the requirement. If not, it may start collecting data and training a new model. Summary of the invention

[0003] One object of the present disclosure is to improve the utilization and acquisition efficiency of models.

[0004] According to one aspect of some embodiments of the present disclosure, a model providing method is proposed, comprising: receiving a model request from a model training network element, wherein the model request includes at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; and feeding back a list of available models to the model training network element according to the model request, wherein the models in the list of available models match the model request.

[0005] In some embodiments, the model sharing method includes determining a target dimension for the matching models.

[0006] In some embodiments, the target dimension includes at least one of a sample data dimension, a model feature dimension, a model algorithm dimension, a model parameter dimension, or a correlation dimension.

[0007] In some embodiments, the method complies with at least one of the following: when the model sharing method includes a model feature dimension, the model request includes at least one of model-related information or sample-related information; when the model sharing method includes a model algorithm dimension, the model request includes a model algorithm identifier; when the model sharing method includes at least one of a sample data dimension or a correlation dimension, the model request includes sample data; when the model sharing method includes a model parameter dimension, the model request includes model parameters.

[0008] In some embodiments, the method further includes: determining a matching model in a model information library according to the model request; and determining an available model list according to the matching model.

[0009] In some embodiments, the method further includes: determining a list of potential similar models in a model information library according to a model request; and determining a list of available models according to model information of models in the list of potential similar models through a network intelligent service element.

[0010] In some embodiments, determining an available model list based on model information of a model in a potential similar model list includes: determining model information of a model in the potential similar model list; determining target model information based on a model request; determining the similarity between the model information of the model in the potential similar model list and the target model information through a predetermined calculation method; determining an available model list based on the potential similar model list and the similarity, wherein the models in the available model list are sorted in order of similarity from high to low or from low to high.

[0011] In some embodiments, the list of available models includes the model identification and model extraction information, and the model extraction information includes at least one of the scene analysis identification, target identification, model-related information, and sample-related information.

[0012] In some embodiments, the method further includes: obtaining model usage feedback information; and updating the model information according to the model usage feedback.

[0013] In some embodiments, the model usage feedback information includes at least one of scene analysis identification feedback information, target identification feedback information, sample-related feedback information, model-related feedback information, and accuracy rate.

[0014] In some embodiments, the method also includes: receiving model notification information or model registration information, the model notification information and the model registration information including at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; determining the model information based on the model notification information or the model registration information.

[0015] In some embodiments, the method complies with at least one of the following: when the model sharing method includes a model feature dimension, the model notification information and the model registration information include at least one of model-related information or sample-related information; when the model sharing method includes a model algorithm dimension, the model notification information and the model registration information include a model algorithm identifier; when the model sharing method includes at least one of a sample data dimension or a correlation dimension, the model notification information and the model registration information include sample data; when the model sharing method includes a model parameter dimension, the model notification information and the model registration information include model parameters.

[0016] In some embodiments, generating model information based on model notification information or model registration information includes: obtaining at least one item of model classification information or model association information through a network intelligent service network element based on the stored model information; and updating the stored model information based on at least one item of model classification information or model association information.

[0017] In some embodiments, obtaining at least one item of model classification information or model association information through a network intelligent service network element includes: performing association analysis on the model information based on any one dimension of scenario analysis identification, model sharing method, target identification, model-related information, sample-related information, or a combination of at least two dimensions to determine at least one item of model classification information or model association information.

[0018] In some embodiments, the method also includes: sending a model subscription request to a model training network element, the model subscription request including at least one of a scenario analysis identifier, a target identifier, a model sharing method, model-related information, and sample-related information of the subscribed target model, wherein, when the model generated by the model training network element conforms to the target model, the model notification information generated and sent by the model training network element is obtained.

[0019] According to one aspect of some embodiments of the present disclosure, a model information acquisition method is proposed, including: receiving model notification information or model registration information, the model notification information and the model registration information including at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; determining model information according to the model notification information or the model registration information.

[0020] In some embodiments, the method complies with at least one of the following: when the model sharing method includes a model feature dimension, the model notification information and the model registration information include at least one of model-related information or sample-related information; when the model sharing method includes a model algorithm dimension, the model notification information and the model registration information include a model algorithm identifier; when the model sharing method includes at least one of a sample data dimension or a correlation dimension, the model notification information and the model registration information include sample data; when the model sharing method includes a model parameter dimension, the model request includes the model parameters.

[0021] In some embodiments, generating model information based on model notification information or model registration information includes: obtaining at least one item of model classification information or model association information through a network intelligent service network element based on the stored model information; and updating the stored model information based on at least one item of model classification information or model association information.

[0022] In some embodiments, obtaining at least one item of model classification information or model association information through a network intelligent service network element includes: performing association analysis on the model information based on any one dimension of scenario analysis identification, model sharing method, target identification, model-related information, sample-related information, or a combination of at least two dimensions to determine at least one item of model classification information or model association information.

[0023] In some embodiments, the method also includes: sending a model subscription request to a model training network element, the model subscription request including at least one of a scenario analysis identifier, a target identifier, a model sharing method, model-related information, and sample-related information of the subscribed target model, wherein, when the model generated by the model training network element conforms to the target model, the model notification information generated and sent by the model training network element is obtained.

[0024] According to one aspect of some embodiments of the present disclosure, a model acquisition method is proposed, including: sending a model request to a model discovery knowledge network element according to a user request, wherein the model request includes at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; obtaining a list of available models fed back by the model discovery knowledge network element, wherein the models in the list of available models match the model request.

[0025] In some embodiments, the method further includes: searching for the target model in the local storage according to a user request; and when it is determined that the target model does not exist in the local storage, executing an operation of sending a model request to the model discovery knowledge network element according to the user request.

[0026] In some embodiments, the method further includes: determining a target model in the list of available models.

[0027] In some embodiments, the method also includes: according to user request, training based on the target model to obtain a target update model; sending the target update model to a storage network element for storage, and sending model notification information or model registration information of the target update model to the model discovery knowledge network element, the model notification information and the model registration information include at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information, so that the model discovery knowledge network element determines the model information of the target update model.

[0028] In some embodiments, the method further includes: generating model usage feedback information according to user usage feedback, and sending it to the model discovery knowledge network element, so that the model discovery knowledge network element updates the model information corresponding to the target model according to the model usage feedback.

[0029] In some embodiments, the model usage feedback information includes at least one of scene analysis identification feedback information, target identification feedback information, sample-related feedback information, model-related feedback information, and accuracy rate.

[0030] In some embodiments, the method also includes: after completing the training of the model to be stored, sending the model to be stored to a storage network element for storage; generating model notification information or model registration information for the model to be stored, the model notification information and model registration information including at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; and sending the model notification information or model registration information to the model discovery knowledge network element.

[0031] In some embodiments, the method also includes: receiving a model subscription request from a model discovery knowledge network element, the model subscription request including at least one of a scenario analysis identifier, a target identifier, a model sharing method, model-related information, and sample-related information of the subscribed target model; when the model to be stored conforms to the target model, generating model notification information and sending it to the model discovery knowledge network element.

[0032] According to one aspect of some embodiments of the present disclosure, a model information processing method is proposed, including: after completing the training of the model to be stored, sending the model to be stored to a storage network element for storage; generating model notification information or model registration information for the model to be stored, the model notification information and the model registration information including at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; and sending the model notification information or the model registration information to a model discovery knowledge network element.

[0033] In some embodiments, the method also includes: receiving a model subscription request from a model discovery knowledge network element, the model subscription request including at least one of a scenario analysis identifier, a target identifier, a model sharing method, model-related information, and sample-related information of the subscribed target model; when the model to be stored conforms to the target model, generating model notification information and sending it to the model discovery knowledge network element.

[0034] According to one aspect of some embodiments of the present disclosure, a network device is proposed, comprising: a model request receiving unit, configured to receive a model request from a model training network element, wherein the model request includes at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; a model list feedback unit, configured to feedback an available model list to the model training network element according to the model request, wherein the models in the available model list match the model request.

[0035] According to one aspect of some embodiments of the present disclosure, a network device is proposed, comprising: an information receiving unit, configured to receive model notification information or model registration information, the model notification information and the model registration information including at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; an information determination unit, configured to determine model information based on the model notification information or the model registration information.

[0036] According to one aspect of some embodiments of the present disclosure, a network device is proposed, comprising: a request sending unit, configured to send a model request to a model discovery knowledge network element according to a user request, wherein the model request includes at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; a model list acquisition unit, configured to acquire a list of available models fed back by the model discovery knowledge network element, wherein the models in the list of available models match the model request.

[0037] According to one aspect of some embodiments of the present disclosure, a network device is proposed, comprising: a model sending unit, configured to send the model to be stored to a storage network element for storage after completing the training of the model to be stored; a model information generating and sending unit, configured to: generate model notification information or model registration information for the model to be stored, the model notification information and the model registration information including at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; and send the model notification information or the model registration information to a model discovery knowledge network element.

[0038] According to one aspect of some embodiments of the present disclosure, a network device is provided, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the methods provided above based on instructions stored in the memory.

[0039] According to one aspect of some embodiments of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored, and when the instructions are executed by a processor, the steps of any one of the above methods are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0041] Figure 1 The present invention provides a flowchart of some embodiments of the model extraction method of the present invention.

[0042] Figure 2 Flowcharts of other embodiments of the model information acquisition method disclosed in the present invention.

[0043] Figure 3 The present invention is a flowchart of some embodiments of the model acquisition method of the present invention.

[0044] Figure 4 The present invention provides a flowchart of some embodiments of the model information processing method.

[0045] Figure 5 The present invention provides a signaling flow chart of some embodiments of the model information processing and acquisition method disclosed herein.

[0046] Figure 6 The present invention is a signaling flow chart of some embodiments of the model acquisition and extraction method disclosed herein.

[0047] Figures 7A-7D They are schematic diagrams of some embodiments of the network device disclosed herein.

[0048] Figure 8 Schematic diagrams of other embodiments of the network device disclosed herein.

[0049] Fig. 9 Schematic diagrams of other embodiments of the network device disclosed herein.

[0050] Fig.10 The present invention is a schematic diagram of some embodiments of the interaction between network devices disclosed herein. DETAILED DESCRIPTION

[0051] The technical solution of the present disclosure is further described in detail below through the accompanying drawings and embodiments.

[0052] In the distributed deployment of AI service network elements in large-scale networks, the objects of model training and reasoning, such as user groups, application types, network element types, slices, etc., may all present similar characteristics. The service requirements of the same network elements, the same applications, and the same user groups are similar and can be used as a reference. The model training network element only considers the local model and does not consider the use of the trained models of other AI service network elements, resulting in a waste of communication overhead and resource costs, and it is difficult to meet the timeliness requirements of AI services.

[0053] For similar analysis scenarios, if the source data of the model, the features of the model, or the model algorithm are similar in different dimensions, a new model can be quickly obtained by obtaining the trained models of other AI service network elements, as well as possible model fine-tuning, input of a small amount of source data, etc., based on AI / ML methods such as transfer learning. For example, for deployment by province or service scenario, for source and target domains with similar features, samples, model algorithms, etc., AI / ML methods such as transfer learning can be used to fine-tune the trained model in the source domain and input a small amount of possible data to meet the model requirements of the target domain.

[0054] In view of the above problems, the present disclosure proposes a model providing, acquiring, information acquiring and processing method and network device to improve the sharing and reuse rate of the model and improve the efficiency of model acquisition.

[0055] The flowcharts of some embodiments of the model providing method of the present disclosure are as follows Figure 1 In some embodiments, the model providing method may be executed by a physical network element or a virtual network element newly added to the network. In some embodiments, the network element executing the model providing method is referred to herein as a model discovery knowledge network element.

[0056] In step S11, the model discovery knowledge network element receives a model request from the model training network element, wherein the model request includes at least one of a scenario analysis identifier, a model sharing method, a target identifier, model related information, and sample related information. In some embodiments, the model request may be sent in at least one of a request and a subscription.

[0057] In some embodiments, the above-mentioned scenario analysis identifier may be a scenario ID, which is used to identify the application scenario of the model. The application scenario may be, for example, UE movement trajectory, network element load, and the like.

[0058] In some embodiments, the model sharing method includes an identifier for determining a target dimension of a matching model, and the target dimension includes at least one of a sample data dimension, a model feature dimension, a model algorithm dimension, a model parameter dimension, or a correlation dimension. In some embodiments, for different model sharing methods (target dimensions), the necessary information in the model request is different. In some embodiments, when the model sharing method includes a model feature dimension, the model request includes at least one of model-related information or sample-related information, so as to facilitate the searched model to conform to the target model feature. In some embodiments, when the model sharing method includes a model algorithm dimension, the model request includes a model algorithm identifier, so as to facilitate the searched model to conform to the target model algorithm. In some embodiments, when the model sharing method includes at least one of a sample data dimension or a correlation dimension, the model request includes sample data, so as to facilitate the searched model to conform to the target sample data. In some embodiments, when the model sharing method includes a model parameter dimension, the model request includes model parameters, so as to facilitate the searched model to conform to the target model parameters. Through the method in the above embodiment, the required dimensions for the target model can be provided when the model is requested, thereby reducing the subsequent search pressure and improving the pertinence and accuracy of model matching.

[0059] In some embodiments, the target identifier may be the object for which the model executes the task, such as the targeted user, user type, region, network type, etc.

[0060] In some embodiments, model-related information may include model algorithms, model parameters, output data characteristics of the model, labels of the output data, etc., wherein the output data characteristics and output data labels of existing models can be set through predetermined data analysis methods or in a manually specified manner.

[0061] In some embodiments, sample-related information may include sample data (all or part of the sample data), characteristics of the sample data, labels of the sample data, etc., wherein the characteristics and labels of the sample data can be set by a predetermined data analysis method or in a manually specified manner.

[0062] In step S12, the model discovery knowledge network element feeds back an available model list to the model training network element according to the model request, wherein the models in the available model list match the model request.

[0063] In some embodiments, the model discovery knowledge network element determines a matching model in the model information library according to the model request, and determines a list of available models according to the matching model, thereby providing available models to the model training network element based on the stored information, thereby improving model utilization and feedback efficiency. In some embodiments, the number of models in the list of available models is at least 1. In some embodiments, the model information stored in the model discovery knowledge network element can be obtained and stored by at least one of the following methods: collecting data in the early stage, subscribing to information about models generated by each model training network element, and obtaining information about models pushed by each model training network element.

[0064] In some embodiments, the model discovery knowledge network element may determine a list of potential similar models in the model information library according to the model request, and then determine the list of available models according to the model information of the models in the list of potential similar models through the network intelligent service network element (e.g., AI service network element). In some embodiments, the network intelligent service network element may determine the model information of the models in the list of potential similar models, determine the similarity between the model information of the models in the list of potential similar models and the target model information by a predetermined calculation method based on the target model information in the model request, and determine the list of available models according to the list of potential similar models and the similarity, wherein the models in the list of available models are sorted in order from high to low or from low to high according to the similarity. In some embodiments, the network intelligent service network element may store a predetermined algorithm, such as a similarity calculation algorithm, to determine the similarity of the models. Based on the method in this embodiment, the order of the models in the list of available models can be sorted in order from high to low or from low to high according to the similarity with the target model, so as to facilitate the model training network element to extract the model with the similarity degree that meets its needs (e.g., the model with the highest similarity), and improve the convenience of model acquisition and use.

[0065] In some embodiments, the model discovery knowledge network element can directly provide the model data itself to the model training network element, thereby improving the efficiency of model acquisition; in some embodiments, the model discovery knowledge network element can feedback the storage address or index information of the model to the model training network element, thereby facilitating the model training network element to find and obtain the corresponding model.

[0066] In some embodiments, the available model list includes the identification of the model and the model extraction information. In some embodiments, the model training network element can search for the corresponding model in the model storage network element based on the identification of the model and obtain the model data.

[0067] In some embodiments, the model extraction information also includes at least one of a scene analysis identifier, a target identifier, model-related information, and sample-related information, thereby providing the model training network element with richer information, facilitating the model training network element to further determine the target model in the list based on the model extraction information, and improving the accuracy of model acquisition.

[0068] Based on the method in the embodiment shown above, it is possible to analyze at least one item of information in the model request, including the scene analysis identifier, the model sharing method, the target identifier, the model-related information, and the sample-related information, determine a list of models that meet the requirements of the model training network element, and feed it back to the model training network element, thereby improving the utilization rate of the trained model, reducing the probability of the model training network element retraining the model, and improving the model acquisition efficiency.

[0069] In some embodiments, after the model training network element obtains the target model based on the list of available models, further training can be performed according to the needs, or it can be used directly. When the user finishes using it, feedback can be given to the model training network element. In some embodiments, the model discovery knowledge network element can receive usage feedback information from the model training network element, and then update the model information according to the model usage feedback, wherein the usage feedback information can be generated by the model training network element based on its own operation or training process, and can also include information generated after further processing based on user feedback. In some embodiments, the model usage feedback information includes at least one of the following: scene analysis identification feedback information, target identification feedback information, sample-related feedback information, model-related feedback information, and accuracy rate, wherein the scene analysis identification feedback information includes the scene analysis identification fed back by at least one of the model training network element or the user, and the target identification feedback information includes the target identification fed back by at least one of the model training network element or the user; the sample-related feedback information includes sample-related information fed back by at least one of the model training network element or the user; the model-related feedback information includes model-related information fed back by at least one of the model training network element or the user; the accuracy rate includes the accuracy rate of the model fed back by at least one of the model training network element or the user.

[0070] Based on the method in the embodiment shown above, users and model training network elements can provide feedback on the acquired and used models, thereby enriching and correcting model information, improving the accuracy of subsequent model extraction, and further improving the effective utilization and user-friendliness of the model.

[0071] In some embodiments, the model discovery knowledge network element can obtain various types of information of the trained model from each model training network element in advance and in real time, thereby enriching the stored model information. The flowcharts of other embodiments of the model information acquisition method disclosed in the present invention are as follows: Figure 2 shown.

[0072] In step S21, the model discovery knowledge network element receives model notification information or model registration information, and the model notification information and the model registration information include at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information.

[0073] In some embodiments, if the model discovery knowledge network element sends a subscription request to the model training network element, the model training network element sends model notification information to the model discovery knowledge network element after generating or obtaining information that meets the subscription request. In some embodiments, the model training network element can actively send information to the model discovery knowledge network element through model registration information. This method can improve the flexibility of information collection.

[0074] In some embodiments, the model discovery knowledge network element may send a model subscription request to the model training network element in advance. The model subscription request includes at least one of the scenario analysis identifier, target identifier, model sharing method, model-related information, and sample-related information of the subscribed target model. When the model generated by the model training network element conforms to the target model, the model notification information generated and sent by the model training network element is obtained. Through such a method, the model discovery knowledge network element can provide the features of the required model, and then obtain model information on demand, thereby improving the pertinence of information collection and reducing the amount of redundant information.

[0075] In some embodiments, the information contained in the model sharing method will affect the type of information contained in the model notification information and the model registration information. For example, when the model sharing method includes a model feature dimension, the model notification information and the model registration information include at least one of the model-related information or the sample-related information; when the model sharing method includes a model algorithm dimension, the model notification information and the model registration information include a model algorithm identifier; when the model sharing method includes at least one of the sample data dimension or the correlation dimension, the model notification information and the model registration information include sample data; when the model sharing method includes a model parameter dimension, the model notification information and the model registration information include model parameters.

[0076] Through the method in the above embodiment, data for the target dimension of the model can be collected when collecting model information, thereby reducing the search pressure for the same dimension requirements when processing model requests, and improving the pertinence and accuracy of model matching.

[0077] In step S22, the model discovery knowledge network element determines the model information according to the model notification information or the model registration information.

[0078] In some embodiments, after the model discovery knowledge network element extracts and stores the required model information from the model notification information or the model registration information, it can further classify the existing models or perform model association analysis through the network intelligent service network element according to the stored model information, obtain at least one of the model classification information or the model association information, and then update the stored model information according to at least one of the model classification information or the model association information. Through such a method, the models can be classified and associated, which can reduce the amount of analysis and calculation when determining the list of available models later, and improve feedback efficiency.

[0079] In some embodiments, obtaining model classification information or model association information may include, according to any one dimension of scene analysis identification, model sharing method, target identification, model related information, sample related information, or a combination of at least two dimensions, performing association analysis on the model information to determine at least one of the model classification information or model association information. Through such a method, it is possible to perform preliminary model information analysis and classification from the perspective of the information provided in the model request, thereby improving the matching degree between the analysis and classification results and the model request, which is conducive to improving the success rate of subsequent determination of the available model list and improving the feedback efficiency.

[0080] Based on the method in the embodiment shown above, relevant information of the models generated by each model training network element in the network can be collected and summarized in a timely manner, thereby enriching the stored model information and improving the success rate of providing available models for model training network elements.

[0081] The flowcharts of some embodiments of the model acquisition method disclosed in the present invention are as follows: Figure 3 In some embodiments, the model providing method may be performed by a physical network element or a virtual network element in the network. In some embodiments, the model providing method may be performed by a model training network element.

[0082] In step S31, a model request is sent to the model discovery knowledge network element according to the user request. In some embodiments, the user request may include the user's model requirements. In some embodiments, the model training network element may determine at least one of its scenario analysis identifier, model sharing method, target identifier, model related information, and sample related information according to the user's model requirements based on a predetermined algorithm or manual configuration, and then generate a model request. In some embodiments, the model request includes at least one of the scenario analysis identifier, model sharing method, target identifier, model related information, and sample related information.

[0083] In some embodiments, after receiving a user request, the model training network element may first search for a target model in local storage. If a target model that satisfies the user request exists locally, a service is provided based on the locally stored target model; if the target model does not exist in the local storage, an operation based on sending a model request to the model discovery knowledge network element is performed, thereby further improving the efficiency of model acquisition.

[0084] In step S32, an available model list fed back by the model discovery knowledge network element is received, wherein the models in the available model list match the model request.

[0085] In some embodiments, the model training network element may determine the target model in the list of available models based on a predetermined strategy, for example, by taking the only model in the list of available models as the target model, or by determining the model with the highest similarity as the target model according to the order of the models in the list. In some embodiments, the model training network element may further analyze the model extraction information provided in the list to determine the target model, thereby improving the matching degree between the acquired target model and the user's needs and reducing the probability of requiring secondary training and the amount of supplementary training.

[0086] In some embodiments, after determining the target model, the model training network element can also perform further training based on the target model according to the user request to obtain the target update model, thereby improving the matching degree of the model to the user request. Further, the model training network element sends the target update model to the storage network element for storage, and sends the model notification information or model registration information of the target update model to the model discovery knowledge network element. The model notification information and the model registration information include at least one of the scenario analysis identifier, the model sharing method, the target identifier, the model related information, and the sample related information, so that the model discovery knowledge network element determines the model information of the target update model. Through such a method, the model training network element can further train the model provided by the model discovery knowledge network element, and share the updated model through the model discovery knowledge network element, thereby improving the richness of the model information stored by the model discovery knowledge network element, improving the subsequent success probability of providing a list of available models to other network elements, and further improving the model utilization rate.

[0087] Based on the method in the embodiment shown above, it is possible to send a model request in response to a user request and provide information including at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information, so as to obtain a list of models that meet the requirements of the model training network element, thereby improving the utilization rate of trained models, reducing the probability of the model training network element retraining the model, and improving the efficiency of model acquisition.

[0088] In some embodiments, after providing model services to users, the model training network element may also obtain user usage feedback. The model training network element generates model usage feedback information after further processing based on the user feedback, and sends it to the model discovery knowledge network element so that the model discovery knowledge network element updates the model information corresponding to the target model according to the model usage feedback.

[0089] In some embodiments, the model usage feedback information includes at least one of scene analysis identification feedback information, target identification feedback information, sample-related feedback information, model-related feedback information, and accuracy, wherein the scene analysis identification feedback information includes the scene analysis identification fed back by at least one of the model training network elements or users, and the target identification feedback information includes the target identification fed back by at least one of the model training network elements or users; the sample-related feedback information includes sample-related information fed back by at least one of the model training network elements or users; the model-related feedback information includes model-related information fed back by at least one of the model training network elements or users; and the accuracy includes the accuracy of the model fed back by at least one of the model training network elements or users. Through such a method, users and model training network elements can be allowed to provide feedback on the models obtained and used, thereby enriching and correcting the model information, improving the accuracy of subsequent model extraction, and further improving the effective utilization rate and user-friendliness of the model; the feedback information is similar to the type of information provided in the model request, thereby reducing the difficulty of using the information and improving the accuracy of subsequent matching.

[0090] In some embodiments, the model training network element can report the trained model to the model discovery knowledge network element in advance and in real time, thereby enriching the stored model information. The flowcharts of some embodiments of the model information processing method of the present disclosure are as follows: Figure 4 shown.

[0091] In step S41, after the training of the model to be stored is completed, the model to be stored is sent to the storage network element for storage. In some embodiments, the model to be stored may be a model generated and trained by the current model training network element. In some embodiments, the model to be stored may also include a model generated after the current model training network element acquires and performs secondary training.

[0092] In step S42, model notification information or model registration information is generated for the model to be stored, and the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information.

[0093] In some embodiments, the model training network element can determine whether the model to be stored belongs to the model subscribed by the model discovery knowledge network element. If it belongs to the model type subscribed by the model discovery knowledge network element, model notification information is generated. In some embodiments, the model training network element can receive a model subscription request from the model discovery knowledge network element. The model subscription request includes at least one of the scenario analysis identifier, target identifier, model sharing method, model-related information, and sample-related information of the subscribed target model. The information in the model subscription request is used to determine whether the model to be stored belongs to the model subscribed by the model discovery knowledge network element. Through such a method, the pertinence of information collection can be improved and the amount of redundant information can be reduced.

[0094] In some embodiments, the model training network element can generate model registration information on its own according to demand and report it to the model discovery knowledge network element, thereby improving the richness of information.

[0095] In step S43, the model notification information or the model registration information is sent to the model discovery knowledge network element.

[0096] Based on the method in the embodiment shown above, relevant information of the models generated by each model training network element in the network can be collected and summarized in a timely manner, thereby enriching the stored model information and improving the success rate of providing available models for model training network elements.

[0097] The signaling flow chart of some embodiments of the model information processing and acquisition method disclosed in the present invention is as follows: Figure 5 shown.

[0098] In 501, the model discovery knowledge network element obtains the information of the trained model of the model training network element. In some embodiments, two methods can be selected. One is to configure the model training network element to actively report, and the message transmission in this step is not performed; the other is for the model discovery knowledge network element to subscribe on demand, for example, for scenarios or algorithms or goals with high possibility of model sharing, subscribe, and then transmit the message in this step. In some embodiments, the message in this step can be called a model subscription request. In some embodiments, the model subscription request carries a scenario analysis ID, a model algorithm (such as a decision tree, clustering, etc.) and a model sharing method (sample-based, model-based, feature-based, correlation-based, etc.). In some embodiments, it can optionally carry the target ID of the scenario analysis (user group, application, network element, etc.).

[0099] In 502, the model training network element performs AI or ML (Machine Learning) model training according to demand.

[0100] In 503, after the model training network element has trained the model, the model is stored in a storage network element. In some embodiments, the storage network element may be a model storage network element dedicated to sharing or storing models, or a general storage network element that stores various types of information.

[0101] In 504, the model training network element notifies the model or registers information to the model discovery knowledge network element. In some embodiments, if a subscription is made in the above 501, this step is to send model notification information; if there is no subscription, model registration information is sent. The information sent in the current step carries the scene analysis ID, model algorithm and parameters, model sharing method, feature labels (which can be labeled by self-supervised learning and other methods), sample data and information (sample data and sampling rate, data source and other information supporting this model), model address (address of the model storage network element, etc.), etc., and optionally carries the target ID. In some embodiments, different model sharing methods must carry different parameters. If it is based on features, it must carry feature labels; if it is based on models, it must carry model algorithms and parameters; if it is based on samples or correlations, it must carry sample data, etc.

[0102] In 505, the model discovery knowledge network element requests the AI ​​service network element for analysis of existing model information on demand, such as classifying and associating models through clustering algorithms such as k-means, association algorithms such as Apriori, etc. The analysis dimensions can be based on scenarios and model algorithms, based on scenarios and features, based on scenarios and sample data, etc., and combinations of the above.

[0103] In 506, the model discovery knowledge network element classifies and associates the models based on different dimensions according to the local existing analysis means and the analysis results of the AI ​​service network element. As the model training network element continuously registers the model, the model discovery knowledge is accumulated to form a knowledge base related to model discovery and selection.

[0104] The signaling flow chart of some embodiments of the model acquisition and extraction method disclosed in the present invention is as follows: Figure 6 shown.

[0105] In 601, a consumer node (e.g., a network element or user performing AI reasoning) obtains a model from a model training network element for reasoning and provides information such as an ID of an analysis scenario.

[0106] In 602, when the model training network element determines that there is no available model that meets the user's needs locally, it prepares to obtain an available model from the model discovery knowledge network element. The model training network element determines the model algorithm, model sharing method, model features, etc. If the model sharing method is based on samples or relationships, sample data can be generated based on historical data or a small amount of data collection; if the model sharing method is based on features, the features can be labeled through self-supervised learning and other methods.

[0107] In 603, the model training network element obtains the available model from the model discovery knowledge network element, carrying the analysis scenario analysis ID, model algorithm, model sharing method, feature label, sample data, etc. Different model sharing methods must carry different parameters. If it is based on features, it must carry feature labels; if it is based on models, it must carry model algorithms and parameters; if it is based on samples or correlations, it must carry sample data, etc.

[0108] In 604, the model discovery knowledge network element preliminarily screens the potential similar model list according to the matching of the model request and the accumulated model discovery knowledge information, for example, based on samples and associations, initiates similarity analysis of sample data, compares model algorithms and parameters based on models, and compares feature labels based on features, and may adjust features based on existing knowledge, such as expanding or narrowing the feature range.

[0109] In 605, the model discovery knowledge network element obtains further model similarity analysis from the AI ​​service network element on demand. The model similarity dimension comes from the model sharing method, such as feature similarity, data similarity, model parameter similarity, etc., and correspondingly carries information such as features, model parameters, data sources, number of samples, time windows, etc.

[0110] In 606, the AI ​​service network element performs similarity analysis using similarity algorithms such as cosine similarity, such as calculating the "distance" between samples, and returns similarity analysis results, including model similarity and related information such as the algorithm used.

[0111] In 607, the model discovery knowledge network element generates a list of available models, which can be prioritized according to similarity and carry the scenario analysis ID, feature list, model algorithm, source data time window, model storage address, etc. as needed.

[0112] In 608, the model discovery knowledge network element returns a list of available models to the model training network element.

[0113] In 609, the model training network element selects an available model from the available model list.

[0114] In 610, the model training network element obtains the model from the model storage network element.

[0115] In 611, if the model training network element determines that the model cannot directly meet the requirements, further model training and fine-tuning may be performed, and data may be collected if necessary. The model training method may include, for example, feature-based transfer learning based on Self-taught learning, correlation-based transfer learning based on Mapping methods, etc.

[0116] In 612, the model training network element returns the model and related information to the consumer.

[0117] In 613, the model training network element sends the trained model to the model storage network element.

[0118] In 614, the model training network element informs the model discovery knowledge network element of the new model.

[0119] In 615 , the consumer provides model usage feedback, such as target users, sample data, feature labels, accuracy, etc.

[0120] In 616, the model training network element corrects and supplements information such as feature labels based on consumer feedback, and provides it to the model discovery knowledge network element for better use of the model in the future.

[0121] In 617, the model discovery knowledge network element may update the model knowledge base after analysis based on the registration of the new model, model feedback information, etc.

[0122] In the method of the embodiment shown above, a model discovery knowledge network element is introduced into the network, and a knowledge base for sharing and reuse of subsequent models is formed through accumulation. Further analysis is performed based on the existing knowledge base and the data provided in the model requirements, and a list of models that meet the requirements is selected. The model is reused by the model training network element, and is used directly, or a small amount of samples are collected and fine-tuned, thereby improving the efficiency of providing model-related services and the utilization rate of the model.

[0123] In the method of the embodiment shown above, by adding key information transmission and judgment processing such as model sharing methods, for example, for different scenario requirements, model algorithms, etc., it is suitable for model sharing methods based on different dimensions such as sample data, model characteristics, model algorithms, and correlations. The model discovery knowledge network element supports the collection, processing and analysis of different information for different model sharing methods, and ultimately forms relevant knowledge for model discovery and selection, thereby improving the refinement of model search and matching, and improving the accuracy and efficiency of model matching.

[0124] In the method of the embodiment shown above, key information transmission and judgment processing such as model feature labels are added. For the model feature dimension, the model training network element performs feature annotation on the local model, and carries the feature label parameters when registering or requesting the model to the model discovery knowledge network element. On the one hand, the model features provided by the model training network element that provides and uses the model are more accurate, and on the other hand, it is more conducive to further feature analysis by the model discovery knowledge network element. If not carried, the model discovery knowledge network element itself can also perform feature analysis through existing knowledge and AI / ML technology, which improves universality.

[0125] In the method of the embodiment shown above, model users are supported to provide information feedback and feature annotation for the model. Model users such as model training network elements and model consumers can provide model feedback information, such as target users, sample data, feature labels, accuracy, etc., especially for model features, through joint annotation and maintenance by multiple model users, the description of model usage is more accurate, and the accumulation of model knowledge and subsequent discovery services are utilized.

[0126] In the method of the embodiment shown above, information reflecting similarity or priority is added. When the model discovery knowledge network element returns a list of recommended models to the model training network element, in addition to the basic description and data of the model, a similarity description is added, and the similarity can be used as a priority recommendation, which reduces the amount of subsequent processing operations and improves the model selection efficiency of the model training network element.

[0127] Schematic diagrams of some embodiments of the network device disclosed in the present invention are as follows Fig. 7A In some embodiments, Fig. 7A The network device shown can discover knowledge network elements for the model.

[0128] The model request receiving unit 711 is capable of receiving a model request from a model training network element, wherein the model request includes at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information. In some embodiments, the scenario analysis identifier may be a scenario ID, which is used to identify an application scenario of the model, and the application scenario may be, for example, a UE movement trajectory, a network element load, etc. In some embodiments, the model request receiving unit 701 is capable of executing the method in any of the embodiments of step S11 above.

[0129] The model list feedback unit 712 can feed back an available model list to the model training network element according to the model request, wherein the models in the available model list match the model request. In some embodiments, the model request receiving unit 712 can execute the method in any embodiment of step S12 above.

[0130] Such a network device can analyze at least one item of information in the model request, including the scene analysis identifier, model sharing method, target identifier, model-related information, and sample-related information, determine a list of models that meet the requirements of the model training network element, and feed back the list to the model training network element, thereby improving the utilization rate of the trained model, reducing the probability of the model training network element retraining the model, and improving the efficiency of model acquisition.

[0131] Schematic diagrams of some embodiments of the network device disclosed in the present invention are as follows Figure 7B In some embodiments, Figure 7B The network device shown can discover knowledge network elements for the model.

[0132] The information receiving unit 721 can receive model notification information or model registration information, wherein the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information. In some embodiments, the information receiving unit 721 can execute the method in any of the embodiments of step S21 above.

[0133] The information determination unit 722 can determine the model information according to the model notification information or the model registration information. In some embodiments, the information determination unit 722 can execute the method in any embodiment of the above step S22.

[0134] Such network equipment can timely collect and summarize relevant information of models generated by various model training network elements in the network, thereby enriching the stored model information and improving the success rate of providing available models for model training network elements.

[0135] In some embodiments, the network device may have the following features: Fig. 7A and 7B The unit shown realizes the collection, organization, and storage of information, and provides model recommendations based on the stored information, thereby reducing the workload of manual operations and improving information processing efficiency.

[0136] In some embodiments, the information collection unit 721 can also receive usage feedback information from the model training network element, and the information determination unit 722 can update the model information according to the model usage feedback, wherein the usage feedback information can be generated by the model training network element based on its own operation or training process, and can also include information generated after further processing based on user feedback. Such a network device can allow users and model training network elements to provide feedback on the models obtained and used, thereby enriching and correcting the model information, improving the accuracy of subsequent model extraction, and further improving the effective utilization rate and user-friendliness of the model.

[0137] Schematic diagrams of some embodiments of the network device disclosed in the present invention are as follows Figure 7CIn some embodiments, Figure 7C The network devices shown can be network elements for model training.

[0138] The request sending unit 731 can send a model request to the model discovery knowledge network element according to the user request. In some embodiments, the user request may include the user's model requirements. In some embodiments, the request sending unit 731 can determine at least one of its scenario analysis identifier, model sharing method, target identifier, model related information, and sample related information based on the user's model requirements, based on a predetermined algorithm or manual configuration, and then generate a model request. In some embodiments, the model request includes at least one of the scenario analysis identifier, model sharing method, target identifier, model related information, and sample related information. In some embodiments, the request sending unit 731 can execute the method in any embodiment of step S31 above.

[0139] The model list acquisition unit 732 can receive the available model list fed back by the model discovery knowledge network element, wherein the models in the available model list match the model request. In some embodiments, the model list acquisition unit 732 can execute the method in any embodiment of step S32 above.

[0140] Such a network device can send model requests in response to user requests, and provide information including at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information, so as to obtain a list of models that meet the requirements of the model training network element, thereby improving the utilization rate of trained models, reducing the probability of the model training network element retraining the model, and improving the efficiency of model acquisition.

[0141] Schematic diagrams of some embodiments of the network device disclosed in the present invention are as follows Fig.7D In some embodiments, Fig.7D The network devices shown can be network elements for model training.

[0142] The model sending unit 741 can send the model to be stored to the storage network element for storage after the current network device completes the model training. In some embodiments, the model to be stored can be a model generated and trained by the current model training network element. In some embodiments, the model to be stored can also include a model generated after the current model training network element obtains and performs secondary training. In some embodiments, the model sending unit 741 can execute the method in any embodiment of step S41 above.

[0143] The model information generation and sending unit 742 can generate model notification information or model registration information for the model to be stored, and send the model notification information or model registration information to the model discovery knowledge network element. The model notification information and the model registration information include at least one of the scenario analysis identifier, the model sharing method, the target identifier, the model related information, and the sample related information. In some embodiments, the model information generation and sending unit 742 can execute the method in any embodiment of step S42 above.

[0144] Such network equipment can timely collect and summarize relevant information of models generated by various model training network elements in the network, thereby enriching the stored model information and improving the success rate of providing available models for model training network elements.

[0145] In some embodiments, the model information generation and sending unit 742 can also obtain user usage feedback after the model training unit provides model services to users, generate model usage feedback information after further processing based on the user feedback, and send it to the model discovery knowledge network element so that the model discovery knowledge network element can update the model information corresponding to the target model according to the model usage feedback.

[0146] It should be noted that each network element in the embodiments of the present disclosure refers to a node unit in network management, which can independently complete certain information processing and transmission functions. The network element can be a device specifically deployed in the network, or a service that provides certain functions in the network.

[0147] A structural diagram of an embodiment of the network device disclosed in the present invention is as follows Figure 8 As shown. The network device includes a memory 801 and a processor 802. Among them: the memory 801 can be a disk, a flash memory or any other non-volatile storage medium. The memory is used to store the instructions in the corresponding embodiments of the model providing, acquiring, information acquiring or processing methods mentioned above. The processor 802 is coupled to the memory 801 and can be implemented as one or more integrated circuits, such as a microprocessor or a microcontroller. The processor 802 is used to execute the instructions stored in the memory, which can improve the utilization rate and acquisition efficiency of the model.

[0148] In one embodiment, it is also possible to Fig. 9 As shown, the network device 900 includes a memory 901 and a processor 902. The processor 902 is coupled to the memory 901 via a BUS 903. The network device 900 can also be connected to an external storage device 905 via a storage interface 904 to call external data, and can also be connected to a network or another computer system (not shown) via a network interface 906. No further detailed description will be given here.

[0149] In this embodiment, by storing data instructions in a memory and then processing the instructions through a processor, the utilization rate and acquisition efficiency of the model can be improved.

[0150] In another embodiment, a computer-readable storage medium stores thereon computer program instructions, which, when executed by a processor, implement the steps of the method in the corresponding embodiment of the model providing, acquiring, obtaining information or processing method. It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, devices, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] Schematic diagrams of some embodiments of the interaction between network devices disclosed in the present invention are as follows: Fig.10 As shown in the figure, the interaction between network devices is divided into two stages: model collection stage a and model discovery stage b.

[0152] In a1, the model discovery knowledge network element obtains the model registration information (or model notification information) from the model training network element 1. In some embodiments, the model training network element 1 may store the model in the model storage network element.

[0153] In a2, the model discovery knowledge network element analyzes and accumulates model knowledge for the model registration information (or model notification information). In some embodiments, the model information can be further classified and associated analyzed by the AI ​​service network element.

[0154] In b1, the consumer node sends a model acquisition user request to the model training network element 2.

[0155] In b2, the model training network element 2 sends a model request to the model discovery knowledge network element for model discovery.

[0156] In b3, the model discovery knowledge network element performs model requirement analysis according to the model request, and after determining the list of available models, performs the operation of returning the selected model to the model training network element 2. In some embodiments, after selecting the target model, the model training network element 2 obtains the model from the model storage network element.

[0157] Such network devices are based on information interaction between each other and information processing within the network devices. Under the premise of ensuring the accuracy of the model, they avoid the communication overhead and waste of computing resources caused by large-scale data collection and training from scratch, improve the timeliness of the model, and improve the accuracy provided by the model through historical analysis and accumulation of model knowledge.

[0158] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0159] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0160] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0161] So far, the present disclosure has been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0162] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0163] It should be noted that the terms "first", "second", etc. in the specification, claims, and drawings of the present disclosure are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present disclosure rather than to limit it. Although the present disclosure has been described in detail with reference to the preferred embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present disclosure can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present disclosure, which should be included in the scope of the technical solution for protection of the present disclosure.

Claims

1. A model providing method, comprising: Receiving a model request from a model training network element, wherein the model request includes at least one of a scene analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information; Feedback an available model list to the model training network element according to the model request, wherein the models in the available model list match the model request.

2. The method according to claim 1, wherein: The model sharing method includes determining a target dimension for matching models.

3. The method according to claim 2, wherein: The target dimension includes at least one of a sample data dimension, a model feature dimension, a model algorithm dimension, a model parameter dimension or a correlation dimension.

4. The method according to claim 1, wherein: The method meets at least one of the following requirements: In the case where the model sharing method includes a model feature dimension, the model request includes at least one of the model related information or the sample related information; In the case where the model sharing method includes a model algorithm dimension, the model request includes the model algorithm identifier; In the case where the model sharing method includes at least one of a sample data dimension or a relevance dimension, the model request includes sample data; In case the model sharing method includes a model parameter dimension, the model request includes the model parameters.

5. The method according to claim 1, further comprising: Determining a matching model in a model information library according to the model request; The available model list is determined according to the matched models.

6. The method according to claim 1, further comprising: Determine a list of potential similar models in a model information repository according to the model request; The available model list is determined according to the model information of the models in the potential similar model list through the network intelligent service network element.

7. The method according to claim 6, wherein: Determining the available model list according to the model information of the models in the potential similar model list includes: Determine the model information of the model in the potential similar model list; Determine target model information according to the model request; Determine the similarity between the model information of the model in the potential similar model list and the target model information by a predetermined calculation method; The available model list is determined according to the potential similar model list and the similarity, wherein the models in the available model list are sorted in order from high to low or from low to high according to the similarity.

8. The method according to any one of claims 1 to 7, wherein: The available model list includes a model identifier and model extraction information, and the model extraction information includes at least one of a scene analysis identifier, a target identifier, model-related information, and sample-related information.

9. The method according to claim 1, further comprising: Get feedback on model usage; Model information is updated based on the model usage feedback.

10. The method according to claim 9, wherein: The model usage feedback information includes at least one of scene analysis identification feedback information, target identification feedback information, sample-related feedback information, model-related feedback information, and accuracy rate.

11. The method according to claim 1, further comprising: Receiving model notification information or model registration information, wherein the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The model information is determined according to the model notification information or the model registration information.

12. The method according to claim 11, wherein: The method meets at least one of the following requirements: In the case where the model sharing method includes a model feature dimension, the model notification information and the model registration information include at least one of the model related information or the sample related information; In the case where the model sharing method includes a model algorithm dimension, the model notification information and the model registration information include the model algorithm identifier; In the case where the model sharing method includes at least one of a sample data dimension or a relevance dimension, the model notification information and the model registration information include sample data; In the case where the model sharing method includes a model parameter dimension, the model notification information and the model registration information include model parameters.

13. The method according to claim 9, wherein: The generating model information according to the model notification information or the model registration information comprises: According to the stored model information, at least one of the model classification information or the model association information is obtained through the network intelligent service network element; The stored model information is updated according to at least one of the model classification information or the model association information.

14. The method according to claim 13, wherein: The acquiring at least one of the model classification information or the model association information through the network intelligent service network element comprises: According to any one dimension of the scenario analysis identifier, the model sharing method, the target identifier, the model-related information, and the sample-related information, or a combination of at least two of the dimensions, the model information is subjected to association analysis to determine at least one item of the model classification information or the model association information.

15. The method according to claim 11, further comprising: A model subscription request is sent to the model training network element, wherein the model subscription request includes at least one of a scenario analysis identifier, a target identifier, a model sharing method, model-related information, and sample-related information of the subscribed target model, wherein when the model generated by the model training network element conforms to the target model, the model notification information generated and sent by the model training network element is obtained.

16. A method for obtaining model information, comprising: Receiving model notification information or model registration information, wherein the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The model information is determined according to the model notification information or the model registration information.

17. The method according to claim 16, wherein: The method meets at least one of the following requirements: In the case where the model sharing method includes a model feature dimension, the model notification information and the model registration information include at least one of the model related information or the sample related information; In the case where the model sharing method includes a model algorithm dimension, the model notification information and the model registration information include the model algorithm identifier; In the case where the model sharing method includes at least one of a sample data dimension or a relevance dimension, the model notification information and the model registration information include sample data; In case the model sharing method includes a model parameter dimension, the model request includes the model parameters.

18. The method according to claim 16, wherein: The generating model information according to the model notification information or the model registration information comprises: According to the stored model information, at least one of the model classification information or the model association information is obtained through the network intelligent service network element; The stored model information is updated according to at least one of the model classification information or the model association information.

19. The method according to claim 18, wherein: The acquiring at least one of the model classification information or the model association information through the network intelligent service network element comprises: According to any one dimension of the scenario analysis identifier, the model sharing method, the target identifier, the model-related information, and the sample-related information, or a combination of at least two of the dimensions, the model information is subjected to association analysis to determine at least one item of the model classification information or the model association information.

20. The method of claim 16, further comprising: A model subscription request is sent to the model training network element, wherein the model subscription request includes at least one of a scenario analysis identifier, a target identifier, a model sharing method, model-related information, and sample-related information of the subscribed target model, wherein when the model generated by the model training network element conforms to the target model, the model notification information generated and sent by the model training network element is obtained.

21. A model acquisition method, comprising: Sending a model request to a model discovery knowledge network element according to a user request, wherein the model request includes at least one of a scenario analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; Obtaining an available model list fed back by the model discovery knowledge network element, wherein the models in the available model list match the model request.

22. The method according to claim 21, further comprising: Searching for a target model in local storage according to the user request; In the case where it is determined that the target model does not exist in the local storage, the operation of sending a model request to the model discovery knowledge network element according to the user request is performed.

23. The method of claim 20, further comprising: A target model is determined in the list of available models.

24. The method of claim 22, further comprising: According to the user request, training is performed based on the target model to obtain a target update model; The target update model is sent to a storage network element for storage, and model notification information or model registration information of the target update model is sent to the model discovery knowledge network element, wherein the model notification information and the model registration information include at least one of a scenario analysis identifier, a model sharing method, a target identifier, model-related information, and sample-related information, so that the model discovery knowledge network element determines the model information of the target update model.

25. The method according to claim 23 or 24, further comprising: Model usage feedback information is generated according to the user's usage feedback and sent to the model discovery knowledge network element, so that the model discovery knowledge network element updates the model information corresponding to the target model according to the model usage feedback.

26. The method according to claim 25, wherein: The model usage feedback information includes at least one of scene analysis identification feedback information, target identification feedback information, sample-related feedback information, model-related feedback information, and accuracy rate.

27. The method of claim 21, further comprising: After completing the training of the model to be stored, sending the model to be stored to the storage network element for storage; Generate model notification information or model registration information for the model to be stored, wherein the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The model notification information or the model registration information is sent to the model discovery knowledge network element.

28. The method according to claim 27, further comprising: Receiving a model subscription request from the model discovery knowledge network element, wherein the model subscription request includes at least one of a scenario analysis identifier, a target identifier, a model sharing method, model related information, and sample related information of a subscribed target model; In the case where the model to be stored conforms to the target model, the model notification information is generated and sent to the model discovery knowledge network element.

29. A model information processing method, comprising: After completing the training of the model to be stored, sending the model to be stored to the storage network element for storage; Generate model notification information or model registration information for the model to be stored, wherein the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The model notification information or the model registration information is sent to the model discovery knowledge network element.

30. The method of claim 29, further comprising: Receiving a model subscription request from the model discovery knowledge network element, wherein the model subscription request includes at least one of a scenario analysis identifier, a target identifier, a model sharing method, model related information, and sample related information of a subscribed target model; In the case where the model to be stored conforms to the target model, the model notification information is generated and sent to the model discovery knowledge network element.

31. A network device comprising: A model request receiving unit is configured to receive a model request from a model training network element, wherein the model request includes at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The model list feedback unit is configured to feed back an available model list to the model training network element according to the model request, wherein the models in the available model list match the model request.

32. A network device comprising: An information receiving unit is configured to receive model notification information or model registration information, wherein the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The information determining unit is configured to determine the model information according to the model notification information or the model registration information.

33. A network device comprising: A request sending unit is configured to send a model request to the model discovery knowledge network element according to a user request, wherein the model request includes at least one of a scenario analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The model list acquisition unit is configured to acquire the available model list fed back by the model discovery knowledge network element, wherein the models in the available model list match the model request.

34. A network device comprising: A model sending unit is configured to send the model to be stored to the storage network element for storage after completing the training of the model to be stored; The model information generating and sending unit is configured as follows: Generate model notification information or model registration information for the model to be stored, wherein the model notification information and the model registration information include at least one of a scene analysis identifier, a model sharing method, a target identifier, model related information, and sample related information; The model notification information or the model registration information is sent to the model discovery knowledge network element.

35. A network device comprising: Memory; as well as A processor coupled to the memory, the processor being configured to execute the method according to any one of claims 1 to 30 based on instructions stored in the memory.

36. A computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 30.

Images