Capability processing method and device, storage medium and electronic equipment

Abstract

The embodiment of the invention relates to a capability processing method and device, a computer readable storage medium and electronic equipment, and relates to the technical field of computers, and the method comprises the steps: obtaining capability evaluation index data of each capability in a plurality of capabilities of the same function type; determining a plurality of attribute features of each capability based on the capability evaluation index data and the time delay of the artificial intelligence model in each capability; fitting the plurality of attribute features through a deep learning model to score each capability, and obtaining a scoring result of each capability; and selecting a target capability from a plurality of capabilities of the same function type according to the scoring result. According to the invention, the capability selection accuracy can be improved.

Description

Background Art

[0002] Cloud network operations can collect the AI ​​capabilities of each subsidiary, and then classify all reported AI capabilities into 5 domains, 12 scenarios and 46 capabilities.

[0003] During the AI ​​capability maintenance process, it is necessary to optimally select AI capabilities (including feature sets + AI models) of the same functional category (such as IP New City L3 layer link traffic prediction models) to avoid wasting resources associated with AI capabilities, and to facilitate the recommendation of an optimal AI capability based on user needs, thereby reducing the trial and error costs of applying multiple AI capabilities.

[0004] In the related technology, there are a variety of methods for selecting the optimal strategy or recommending the optimal model, but it is impossible to select the optimal AI capabilities for the unified management and sharing platform of AI capabilities, which has certain limitations. Summary of the invention

[0005] The purpose of the present disclosure is to provide a capability processing method, a capability processing device, a computer-readable storage medium, and an electronic device, thereby overcoming the problem of inaccurate capability processing caused by the limitations and defects of related technologies at least to a certain extent.

[0006] According to the first aspect of the present disclosure, a capability processing method is provided, including: obtaining capability evaluation index data for each capability of a plurality of capabilities of the same functional type; determining a plurality of attribute characteristics of each capability based on the capability evaluation index data and the time delay of an artificial intelligence model in each capability; fitting the plurality of attribute characteristics through a deep learning model to score each capability and obtain a scoring result for each capability; and selecting a target capability from the plurality of capabilities of the same functional type according to the scoring result.

[0007] In an exemplary embodiment of the present disclosure, the capability evaluation index data includes the recommendation degree, the actual total number of calls of the same functional type capabilities, the actual total number of calls of each capability, the actual usage accuracy after the capability is called and used, and the computing power occupancy of the capability; based on the capability evaluation index data and the latency of each artificial intelligence model, multiple attribute characteristics of each capability are determined, including: determining the proportion of call times according to the actual total number of calls of each capability in the capability evaluation index data and the actual total number of calls of the same functional type capabilities; determining the latency of each capability according to the latency of each artificial intelligence model; determining the computing power resource occupancy rate according to the computing power occupancy of each capability and the overall computing power resource occupancy of the same functional type; taking the recommendation degree, the proportion of call times, the latency of each capability, the computing power resource occupancy rate and the actual usage accuracy after the capability is used as the multiple attribute characteristics.

[0008] In an exemplary embodiment of the present disclosure, the multiple attribute features are fitted through a deep learning model to score each capability and obtain a scoring result for each capability, including: determining positive features and negative features based on the multiple attribute features; performing element-wise product operations on the positive features and the negative features to determine a fusion vector; determining a predicted scoring result based on the fusion vector, and determining the scoring result in combination with the predicted scoring result and the actual scoring result.

[0009] In an exemplary embodiment of the present disclosure, the positive features and negative features are determined based on the multiple attribute features, including: encapsulating the recommendation degree, the proportion of call times of each capability, and the actual usage accuracy of each capability after being called and used in the multiple attribute features to determine the positive features; encapsulating the latency of each capability and the computing resource occupancy rate of each capability to determine the negative features.

[0010] In an exemplary embodiment of the present disclosure, determining the predicted scoring result based on the fusion vector includes: performing logical processing on a weight parameter, the fusion vector, and a mean deviation value of a total acquired feature of each capability to obtain the predicted scoring result.

[0011] In an exemplary embodiment of the present disclosure, determining the scoring result by combining the predicted scoring result and the actual scoring result includes: determining a loss function according to the predicted scoring result and the actual scoring result; updating the predicted scoring result based on the loss function until the loss function satisfies an end condition to determine the scoring result.

[0012] In an exemplary embodiment of the present disclosure, selecting a target capability from a plurality of capabilities of the same functional type according to the scoring result includes: selecting a capability with the highest scoring result as the target capability from the plurality of capabilities of the same functional type.

[0013] In an exemplary embodiment of the present disclosure, the method further includes: retaining the target capability and deleting capabilities that are duplicated with the target capability.

[0014] According to a second aspect of the present disclosure, a capability processing device is provided, including: a capability evaluation index data acquisition module, used to acquire multiple capabilities, and acquire capability evaluation index data for each capability in multiple capabilities of the same functional type; an attribute feature determination module, used to determine multiple attribute features of each capability based on the capability evaluation index data and the time delay of the artificial intelligence model in each capability; a scoring result determination module, used to fit the multiple attribute features through a deep learning model to score each capability and obtain a scoring result for each capability; a target capability selection module, used to select a target capability from multiple capabilities of the same functional type according to the scoring result.

[0015] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the capability processing method described in any one of the above items is implemented.

[0016] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute any one of the capability processing methods described above by executing the executable instructions.

[0017] In the technical solution provided in the embodiments of the present disclosure, on the one hand, multiple attribute characteristics of each AI capability can be determined based on the capability evaluation index data of each capability of the same functional type and the delay of the artificial intelligence model, and then each capability can be scored based on multiple attribute characteristics to select the target capability based on the scoring result. The attribute characteristics of each capability can be determined based on different dimensions to determine the scoring result, and then the target capability can be selected based on the scoring result, avoiding the limitation of the related technology that the required target capability cannot be selected, and improving the accuracy and scope of application. On the other hand, by fitting multiple attribute characteristics through a deep learning model to obtain a scoring result, the accuracy of the scoring result can be improved, so that the target capability can be accurately screened, thereby reducing resource usage and reducing the high cost of trial and error for users.

[0018] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0020] Figure 1 A flowchart of a capability processing method in an embodiment of the present disclosure is schematically shown.

[0021] Figure 2 A schematic diagram schematically illustrates the attribute features in an embodiment of the present disclosure.

[0022] Figure 3 The following is a schematic diagram schematically showing a method of determining a scoring result for each capability based on attribute characteristics in an embodiment of the present disclosure.

[0023] Figure 4 A schematic diagram of determining fusion features in an embodiment of the present disclosure is schematically shown.

[0024] Figure 5 The overall process flow diagram of capability selection in an embodiment of the present disclosure is schematically shown.

[0025] Figure 6 A schematic diagram schematically illustrates a system architecture for extracting AI capabilities in an embodiment of the present disclosure.

[0026] Figure 7 The block diagram of the capability processing device in the embodiment of the present disclosure is schematically shown.

[0027] Figure 8 The block diagram of the electronic device according to the embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0028] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the example embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0029] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0030] In some embodiments, the automation and intelligence of cloud network operations are the key points for thoroughly implementing the strategy of building a cyber power, building a high-speed, ubiquitous, space-ground integrated, cloud-network integrated, intelligent and agile, green and low-carbon, safe and controllable intelligent comprehensive digital information infrastructure, and practicing the digital transformation of cloud network operations lines.

[0031] With the rapid development of network intelligence, a variety of AI platforms or AI systems that manage AI capabilities (including AI models, feature label data, etc.) have been developed, resulting in duplicate construction and serious fragmentation of AI capabilities. It is necessary to unify the AI ​​capabilities scattered on single-point platforms / systems and call them on other platforms / systems to provide more efficient services and effectively improve the efficiency of network operation and management. However, there is currently no platform / system that provides unified management and sharing of AI capabilities.

[0032] During the AI ​​capability maintenance process, it is necessary to optimally select AI capabilities (including feature sets + AI models) of the same functional type (such as IP New City L3 layer link traffic prediction models) to avoid wasting resources associated with AI capabilities, and to facilitate the recommendation of an optimal AI capability based on user needs, thereby reducing the trial and error costs of applying multiple AI capabilities.

[0033] In some embodiments, there are multiple methods for selecting the optimal strategy or recommending the optimal model, but the above methods cannot select the optimal AI capabilities for the unified management and sharing platform of AI capabilities.

[0034] In the embodiments of the present disclosure, in order to solve the above-mentioned technical problems, a capability processing method is provided to select the target capability that best meets the user's needs from multiple capabilities of the same functional type, so as to perform subsequent processing based on the target capability. The capability processing method can include an AI capability update module, which realizes the upgrade and iteration of AI capabilities and deletes the old version of AI capabilities. Specific functional requirements may include: requiring random collection of evaluation index data such as recommendation degree, total number of calls, accuracy / absolute mean square error, etc. of artificial intelligence capabilities in the same functional class within a period of time. It is necessary to calculate the training delay, running delay and computing power resources used by the artificial intelligence model. It is required to score similar artificial intelligence capabilities according to artificial intelligence algorithms and feature data, select the artificial intelligence capabilities with the highest comprehensive scores to retain, and remove redundant / duplicate artificial intelligence capabilities to obtain the required target capabilities.

[0035] Next, refer to Figure 1 The capability processing method in the embodiment of the present disclosure is specifically described in detail.

[0036] In step S110 , capability evaluation index data of each capability among a plurality of capabilities of the same functional type is obtained.

[0037] In the embodiments of the present disclosure, multiple AI platforms or AI systems developed can manage multiple capabilities, and the multiple capabilities can belong to the same functional type or different functional types. In order to reduce resource usage, multiple capabilities of the same functional type can be optimally selected. Exemplarily, the capability evaluation index data of each capability can be obtained among all capabilities corresponding to the same functional type. For all capabilities, multiple capabilities of the same functional type refer to multiple capabilities belonging to the same functional type. For example, all capabilities corresponding to functional type 1 include capabilities A-capability F, and all capabilities corresponding to functional type 2 include capabilities M and capabilities N. Then, the capability evaluation index data of each capability in capabilities A-capability F corresponding to functional type 1 can be obtained, and the capability evaluation index data of capabilities M and capabilities N corresponding to functional type 2 can be obtained.

[0038] The capability evaluation index data can be used to evaluate the performance of each AI capability. The capability evaluation index data may include one or more of the following: the recommendation degree of each AI capability, the actual total number of calls for each type of AI capability (i.e., the actual total number of calls for all AI capabilities of the same functional type), the actual total number of calls for each AI capability, and the actual accuracy / mean square absolute error after the AI ​​capability is called and used. In addition, the capability evaluation index data may also include the computing power occupied by each AI capability, which may specifically include the occupation of resources for AI capability training and algorithm operation environment.

[0039] In some embodiments, in the AI ​​capability management and sharing system, evaluation indicators such as the recommendation degree of each AI capability of the same functional type of multiple customers within a period of time, the actual total number of calls for each type of AI capability, the actual total number of calls for each AI capability, the actual accuracy / mean square absolute error after the AI ​​capability is called and used, can be randomly collected to obtain capability evaluation indicator data for each capability.

[0040] Next, continue to refer to Figure 1 As shown in , in step S120, multiple attribute characteristics of each capability are determined based on the capability evaluation index data and the delay of the artificial intelligence model in each capability.

[0041] In the disclosed embodiment, the AI ​​capability may include information such as an AI model and feature data. Each AI model, that is, the latency of the AI ​​model, may be determined based on the training latency of the AI ​​model and the computation latency of the AI ​​model.

[0042] After obtaining the capability evaluation index data of each AI capability and the latency of each AI model, multiple attribute characteristics of each capability can be determined based on the capability evaluation index data of each AI capability and the latency of the AI ​​model in each capability. The attribute characteristics are used to describe the calling status, latency status, and computing resource usage of the capability. Figure 2 As shown in , the attribute characteristics may include the proportion of the number of calls of each AI capability, the latency of each AI capability, and the computing resource occupancy rate of each AI capability. In addition, the attribute characteristics also include the recommendation degree of each AI capability, the actual accuracy or mean square absolute error, absolute error, variance, etc. after the AI ​​capability is called and used. The actual accuracy is used as an example for explanation. Among them, the proportion of the number of calls of each AI capability = the number of calls of each AI capability / the total number of calls of each type of AI capability; the latency of each AI capability = 0.5×(training latency + inference latency); the computing resource occupancy rate of each AI capability = the computing resource occupancy of each AI capability (the resource capacity occupied by the training model + the resource capacity occupied by the inference algorithm) / the total computing resource occupancy of each type of AI capability.

[0043] Continue to refer Figure 1 As shown in , in step S130, the multiple attribute features are fitted by a deep learning model to score each capability and obtain a scoring result for each capability.

[0044] In the disclosed embodiment, on the basis of obtaining multiple attribute characteristics of each AI capability, multiple attribute characteristics can be fitted to score each capability in multiple capabilities of the same functional type to obtain a scoring result for each capability. Exemplarily, each capability can be scored based on a deep learning model by integrating multiple attribute characteristics of each capability mentioned above to obtain a scoring result for each capability. The deep learning model can implement a deep learning algorithm, which can be, for example, an RMRT (rating matrix and review text based recommendation model) algorithm or other types of deep learning algorithms, which are not specifically limited here.

[0045] Figure 3 The process of fitting multiple attribute features through deep learning algorithms is schematically shown in Figure 3 As shown in , it mainly includes the following steps:

[0046] In step S310, multiple attribute characteristics of each capability are obtained;

[0047] In step S320, a fusion vector is calculated according to the attribute features;

[0048] In step S330, the fusion vector is passed to the AI ​​capability scoring layer to obtain a prediction scoring result;

[0049] In step S340, the loss function is calculated based on the actual rating results fed back by the user, and positive feedback is performed to correct the predicted rating results;

[0050] In step S350, the final scoring result is obtained.

[0051] In the disclosed embodiment, the attribute characteristics may include but are not limited to the recommendation degree of each capability, the proportion of call times of each capability and the actual usage accuracy of each capability, the latency of each capability and the computing resource occupancy rate of each capability.

[0052] First, since the influence and effect of each attribute feature are different, multiple attribute features can be divided into two types: positive features and negative features, depending on whether the attribute feature has a positive effect. Positive features can provide a positive effect on the ability score result, and the positive features may be positively correlated with the score result, that is, the larger the positive feature, the higher the score result may be. Negative features can provide a negative effect on the ability score result, and the negative features may be negatively correlated with the score result, that is, the larger the negative feature, the lower the score result may be. Reference Figure 4 As shown in , positive features and negative features can be determined based on multiple attribute features.

[0053] For example, reference Figure 4 As shown in , since the attribute features such as recommendation degree, call frequency ratio and actual use accuracy have a positive effect on the scoring result, while the latency of AI capability and computing resource utilization rate have a negative effect on the scoring result, the recommendation degree, call frequency ratio of each capability and actual use accuracy rate of each capability among multiple attribute features of each capability can be encapsulated as positive feature U, and the latency of each AI capability and computing resource utilization rate of each capability can be encapsulated as negative feature N. Moreover, the larger the positive feature, the better, and the smaller the negative feature, the better.

[0054] It should be noted that if the AI ​​capability is not actually used, the actual usage accuracy described in the attribute characteristics can be replaced by the accuracy of the model retraining.

[0055] Furthermore, after encapsulating multiple attribute features into positive features and negative features, the positive features and the negative features can be element-wise multiplied to obtain a fusion vector. The element-wise multiplication operation refers to multiplying the elements at the same position in the positive features and the negative features. For example, the element-wise multiplication operation can be performed on the positive features and the negative features according to formula (1) to obtain a fusion vector H:

[0056] H=multiply(U,N) Formula (1)

[0057] On this basis, the fusion vector H can be passed to the capability scoring layer for comprehensive scoring. For example, the weight parameters of the capability scoring layer, the fusion vector, and the mean deviation value of the total acquired features of each capability can be logically processed to obtain a predicted scoring result. Specifically, the weight parameters of the capability scoring layer and the fusion vector can be multiplied and added to the mean deviation value of the total acquired features of each capability to obtain a predicted scoring result, as shown in formula (2):

[0058] R P (U,N)=WH+b Formula (2)

[0059] Among them, w represents the weight parameter of the capability scoring layer, and b represents the mean deviation value of the total acquired features of each AI capability.

[0060] After obtaining the predicted scoring result, the loss function can be determined according to the predicted scoring result and the actual scoring result. Specifically, the loss function can be determined according to the logical processing result of the predicted scoring result and the actual scoring result, wherein the loss function can be calculated according to the difference between the predicted scoring result and the actual scoring result combined with the regularization parameter. The calculation formula of the loss function can refer to formula (3):

[0061] loss=∑ U,N∈I(R P -R r ) 2 +σ||w|| 2 Formula (3)

[0062] Among them, I represents the total sample set in the training set, R P Represents the prediction score result of comprehensive calculation, R r Represents the true rating result of user feedback. In order to prevent overfitting, σ is used to control the regularization degree.

[0063] Next, based on the loss function, the final scoring result of each capability can be obtained according to the actual scoring result and the predicted scoring result. Exemplarily, the predicted scoring result can be updated based on the loss function until the loss function meets the end condition to determine the scoring result. In some embodiments, the predicted scoring result when the loss function meets the end condition can be determined as the final scoring result. Among them, the end condition can be that the loss function is 0 or reaches a pre-set threshold. The pre-set threshold can be set according to actual needs and is not specifically limited here.

[0064] For example, if the loss function is 0 and the predicted score result is 0.6, 0.6 can be used as the final score result.

[0065] Furthermore, the target capability can be selected from all capabilities of the same functional type according to the scoring results. The number of target capabilities can be one or more, which is determined according to actual needs. When the number of target capabilities is one, the capability with the highest scoring result among all capabilities can be used as the target capability. For example, if functional type 1 includes capabilities A-capability F, among which capability B has the highest scoring result, capability B can be used as the target capability corresponding to functional type 1. When the number of target capabilities is multiple, the required number of capabilities can be selected as target capabilities in the order of the scoring results from high to low.

[0066] After determining the target capability, the target capability can be retained, and other capabilities except the target capability in the same function type can be deleted to delete redundant or duplicate AI capabilities. For example, if function type 1 includes capabilities A-F, and capability B has the highest score, capability B can be used as the target capability corresponding to function type 1, and capability A and capabilities C-F can be deleted. Furthermore, the operation corresponding to the task request can be performed by retaining the filtered target capabilities.

[0067] In the disclosed embodiment, since the scoring result of each capability is used, a capability that meets the requirements can be selected from multiple capabilities belonging to the same functional type, thereby reducing resource usage and reducing the high cost of trial and error for users when there are duplicate capabilities.

[0068] Figure 5 The overall flow chart for capability selection is schematically shown in Figure 5 As shown in , it mainly includes the following steps:

[0069] In step S510, capability evaluation index data is collected;

[0070] In step S520, feature fusion calculation is performed to obtain multiple attribute features;

[0071] In step S530, scoring is performed based on the deep learning algorithm to obtain a predicted scoring result;

[0072] In step S540, the final scoring result is obtained;

[0073] In step S550, AI capability maintenance is performed based on the scoring results.

[0074] In the disclosed embodiment, based on the recommended degree of AI capabilities of the same functional type, AI model training delay, AI model reasoning delay, computing power occupancy and other parameters, multiple attribute characteristics of each capability are determined from multiple dimensions such as capability evaluation index data and AI model delay, and the same AI capabilities are comprehensively scored based on the above attribute characteristics based on the deep learning algorithm RMRT, so as to select the optimal AI capability based on multiple attribute characteristics of each capability. The technical solution in the disclosed embodiment can select the AI ​​capability with the best effect as the target capability under the premise of meeting user needs, calculate the scoring result of each capability according to the attribute characteristics of each capability of the same type, further determine the target capability according to the scoring result, and extract the target capability by combining multi-dimensional features, thereby improving the accuracy and comprehensiveness of the selected target capability, and reducing resource occupancy and high-cost trial and error of users. In addition, the device for selecting the target capability can be encapsulated as a tool for various AI capability management and sharing platforms / systems to apply, thereby improving the scope of application and providing convenience for enterprises.

[0075] In the embodiment of the present disclosure, a system architecture for selecting AI capabilities is provided. Figure 6As shown in , the system architecture may include an external system and an AI capability management system. The external system can send user intent to the AI ​​capability management system, which parses user application intent based on the intent parsing algorithm and feeds the results back to the external system. The system architecture for selecting AI capabilities may include an AI capability request module, an AI capability subscription module, an AI model effect evaluation module, an AI model retraining module, an AI capability publishing module, an AI capability application value evaluation module, an optimal AI capability selection module, an AI capability management and sharing system improvement plan formulation module, and an intelligent asset full life cycle management module. Among them:

[0076] AI capability request module: Users input AI capability application intentions into the AI ​​capability management system. The AI ​​capability management system parses the user's application intentions based on the intent parsing algorithm and provides feedback to the user for verification.

[0077] AI capability subscription module: After the user verifies that the parsed intent is correct, he / she submits a subscription request for the AI ​​capability. After the subscription request is approved, the AI ​​capability can be downloaded to the local system in the form of an image or a model.

[0078] AI model effectiveness evaluation module: Evaluate the effectiveness of AI models based on value evaluation indicators (such as accuracy, variance, etc.).

[0079] AI capability sharing can include capability publishing and model retraining. Specific steps include AI capability request, AI capability subscription, and AI model effect evaluation. The AI ​​model retraining module can compare the AI ​​model effect evaluation result with the threshold, and retrain the AI ​​model below the threshold until the AI ​​model effect evaluation result is greater than or equal to the threshold. AI capability publishing module: Based on user intent, the AI ​​model that has passed the best AI model effect evaluation (highest accuracy / lowest computing latency, etc.) will be shared through one or more sharing methods. Sharing methods can include standardized sharing, retraining sharing, inference call, model embedded reference, and image download, among which:

[0080] Standardized sharing: Standardized AI capabilities can be directly shared through intelligent asset management;

[0081] Retraining and sharing: Retrain the managed AI capabilities to make them reusable in different regions, professions, and scenarios;

[0082] Reasoning call: call through the API interface to perform model reasoning in the cloud;

[0083] Model embedded reference: encapsulate and package AI models and embed them into other systems;

[0084] Image download: The AI ​​inference model and its operating environment are packaged and downloaded to other systems, which then provide computing resources.

[0085] AI capability application value assessment module: Based on capability evaluation indicator data such as the application scope of AI capabilities, number of calls, actual application accuracy of AI models, etc., the actual application value of AI capabilities is comprehensively assessed.

[0086] Optimal AI capability selection module: Based on the recommendation degree, AI model training latency, AI model inference latency, computing power usage, etc. of the AI ​​capabilities of the same functional category, the optimal AI capabilities in the same functional category are comprehensively selected for retention and redundant AI capabilities are deleted. The target capability can be obtained through feature selection, feature fusion calculation, score settlement based on the RMRT model, and AI capability selection.

[0087] AI capability management and sharing system improvement plan formulation module: Based on the results of AI capability application value assessment, improvement plans are proposed from the aspects of improving AI model accuracy, optimal selection of AI capabilities with the same functional category, and optimization of recommended plans.

[0088] Smart asset full life cycle management module: optimizes and updates the managed smart assets based on the retraining results, performs maintenance management on the addition, deletion and modification of newly managed smart assets, and provides query of the full amount of smart assets.

[0089] In the disclosed embodiments, through the optimal AI capability selection module in the above-mentioned system architecture for selecting AI capabilities, the optimal AI capability is selected based on a collection of deep learning RMRT algorithms such as the recommendation degree of AI capabilities of the same functional type managed, AI model training delay, AI model inference delay, computing power occupancy, etc., which can recommend the most effective AI capabilities while meeting user needs, thereby improving the accuracy of capability selection.

[0090] The platform or system in the disclosed embodiment is a separate platform or system that can interact with the internal and external systems of the enterprise through interfaces, thereby providing efficient and high-quality intelligent services, improving the efficiency of network operation and management, and improving customer perception. In addition, the optimal AI capability selection module can be encapsulated as a tool for various AI capability management and sharing platforms to apply, which improves convenience and versatility and expands the scope of application.

[0091] The present disclosure also provides a capability processing device. Figure 7 As shown, the capability processing device 700 mainly includes the following modules:

[0092] A capability evaluation index data acquisition module 701 is used to acquire capability evaluation index data of each capability among multiple capabilities of the same functional type;

[0093] An attribute feature determination module 702, for determining a plurality of attribute features of each capability based on the capability evaluation index data and the latency of the artificial intelligence model in each capability;

[0094] A scoring result determination module 703 is used to fit the multiple attribute features through a deep learning model to score each capability and obtain a scoring result for each capability;

[0095] The target capability selection module 704 is used to select a target capability from multiple capabilities of the same functional type according to the scoring result.

[0096] In an exemplary embodiment of the present disclosure, the capability evaluation index data includes the recommendation degree, the actual total number of calls of the same functional type capabilities, the actual total number of calls of each capability, the actual usage accuracy after the capability is called and used, and the computing power occupancy of the capability; the attribute feature determination module includes: a call number ratio determination module, which is used to determine the call number ratio according to the actual total number of calls of each capability in the capability evaluation index data and the actual total number of calls of the same functional type capabilities; a delay determination module, which is used to determine the delay of each capability according to the delay of each artificial intelligence model; a computing power resource occupancy rate determination module, which is used to determine the computing power resource occupancy rate according to the computing power occupancy of each capability and the overall computing power resource occupancy of the same functional type; an attribute feature calculation module, which is used to use the recommendation degree, the call number ratio, the delay of each capability, the computing power resource occupancy rate and the actual usage accuracy after the capability is used as the multiple attribute features.

[0097] In an exemplary embodiment of the present disclosure, the scoring result determination module includes: a feature classification module, used to determine positive features and negative features based on the multiple attribute features; a fusion vector determination module, used to perform element-wise product operations on the positive features and the negative features to determine the fusion vector; a scoring result calculation module, used to determine a predicted scoring result based on the fusion vector, and determine the scoring result in combination with the predicted scoring result and the actual scoring result.

[0098] In an exemplary embodiment of the present disclosure, the feature classification module includes: a positive feature determination module, which is used to encapsulate the recommendation degree, the proportion of call times of each capability, and the actual usage accuracy of each capability after being called and used in the multiple attribute features to determine the positive features; a negative feature determination module, which is used to encapsulate the delay of each capability and the computing resource occupancy rate of each capability to determine the negative features.

[0099] In an exemplary embodiment of the present disclosure, the scoring result calculation module includes: a predicted scoring result determination module, which is used to logically process the weight parameter, the fusion vector and the mean deviation value of the total acquired features of each capability to obtain the predicted scoring result.

[0100] In an exemplary embodiment of the present disclosure, the scoring result calculation module includes: a loss function determination module, used to determine the loss function according to the predicted scoring result and the actual scoring result; a scoring result update module, used to update the predicted scoring result based on the loss function until the loss function satisfies the end condition to determine the scoring result.

[0101] In an exemplary embodiment of the present disclosure, the target capability selection module is configured to: select the capability with the highest score result from multiple capabilities of the same functional type as the target capability.

[0102] In an exemplary embodiment of the present disclosure, the device further includes: a capability deletion module, configured to retain the target capability and delete capabilities that are repeated with the target capability.

[0103] It should be noted that the specific details of each module in the above-mentioned capability processing device have been described in detail in the corresponding capability processing method, so they will not be repeated here.

[0104] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0105] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0106] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0107] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods or program products. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as "circuits", "modules" or "systems".

[0108] Refer to the following Figure 8 The electronic device 800 according to this embodiment of the present disclosure is described. Figure 8 The electronic device 800 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0109] like Figure 8 As shown, the electronic device 800 is in the form of a general computing device. The components of the electronic device 800 may include, but are not limited to: the at least one processing unit 810, the at least one storage unit 820, a bus 830 connecting different system components (including the storage unit 820 and the processing unit 810), and a display unit 840.

[0110] The storage unit stores program codes, which can be executed by the processing unit 810, so that the processing unit 810 performs the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processing unit 810 can perform the following steps: Figure 1 Follow the steps shown in .

[0111] The storage unit 820 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 8201 and / or a cache memory unit 8202 , and may further include a read-only memory unit (ROM) 8203 .

[0112] The storage unit 820 may also include a program / utility 8204 having a set (at least one) of program modules 8205, such program modules 8205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0113] Bus 830 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0114] The electronic device 800 may also communicate with one or more external devices 900 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 800, and / or communicate with any device that enables the electronic device 800 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 850. Furthermore, the electronic device 800 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 860. As shown, the network adapter 860 communicates with other modules of the electronic device 800 via a bus 830. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 800, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0115] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or an electronic device, etc.) to execute the method according to the implementation of the present disclosure.

[0116] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary implementations of the present disclosure described in the above "Exemplary Method" section of the present specification.

[0117] According to the program product for implementing the above method in the embodiment of the present disclosure, it can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0118] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0119] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0120] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0121] Program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0122] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0123] Other embodiments of the present disclosure will be readily apparent to those skilled in the art after considering the specification and practicing the inventions invented herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not invented by the present disclosure. The specification and examples are to be considered merely exemplary, and the true scope and spirit of the present disclosure are indicated by the claims.

Claims

1. A capability processing method, characterized in that: include: Obtain capability evaluation index data for each capability among multiple capabilities of the same functional type; Determining multiple attribute characteristics of each capability based on the capability evaluation indicator data and the latency of the artificial intelligence model in each capability; Fitting the multiple attribute features through a deep learning model to score each capability and obtain a scoring result for each capability; A target capability is selected from multiple capabilities of the same functional type according to the scoring result.

2. The capability processing method according to claim 1, characterized in that: The capability evaluation index data includes the recommendation degree, the actual total number of calls of the same function type capability, the actual total number of calls of each capability, the actual usage accuracy after the capability is called and used, and the computing power usage of the capability; The method of determining multiple attribute characteristics of each capability based on the capability evaluation index data and the delay of each artificial intelligence model includes: Determine the call count ratio according to the actual total call count of each capability in the capability evaluation index data and the actual total call count of capabilities of the same functional type; Determine the latency of each capability based on the latency of each AI model; Determine the computing power resource utilization rate based on the computing power utilization of each capability and the overall computing power resource utilization of the same function type; The recommendation degree, the proportion of the number of calls, the delay of each capability, the computing resource occupancy rate, and the actual usage accuracy of the capability after use are used as the multiple attribute features.

3. The capability processing method according to claim 1, characterized in that: The multiple attribute features are fitted by the deep learning model to score each capability, and the scoring result of each capability is obtained, including: Based on the multiple attribute features, determining positive features and negative features; Performing an element-wise multiplication operation on the positive feature and the negative feature to determine a fusion vector; A predicted scoring result is determined based on the fusion vector, and the scoring result is determined in combination with the predicted scoring result and the actual scoring result.

4. The capability processing method according to claim 3, characterized in that: The determining of positive features and negative features based on the multiple attribute features includes: Encapsulate the recommendation degree, the proportion of the number of calls of each capability, and the actual usage accuracy of each capability after being called and used in the multiple attribute features to determine the positive feature; The latency of each capability and the computing resource occupancy rate of each capability are encapsulated to determine the negative feature.

5. The capability processing method according to claim 3, characterized in that: The step of determining a prediction score result based on the fusion vector comprises: The weight parameter, the fusion vector and the deviation value mean of the total acquired features of each capability are logically processed to obtain the predicted scoring result.

6. The capability processing method according to claim 1, characterized in that: The step of combining the predicted scoring result and the actual scoring result to determine the scoring result includes: Determine a loss function according to the predicted scoring result and the actual scoring result; The predicted scoring result is updated based on the loss function until the loss function satisfies an end condition to determine the scoring result.

7. The capability processing method according to claim 1, characterized in that: The step of selecting a target capability from a plurality of capabilities of the same functional type according to the scoring result comprises: From multiple capabilities of the same functional type, the capability with the highest score result is selected as the target capability.

8. The capability processing method according to claim 1, characterized in that: The method further comprises: The target capability is retained, and capabilities that are duplicated with the target capability are deleted.

9. A capability processing device, characterized in that: include: A capability evaluation index data acquisition module is used to acquire capability evaluation index data of each capability among multiple capabilities of the same functional type; An attribute feature determination module, for determining a plurality of attribute features of each capability based on the capability evaluation index data and the latency of the artificial intelligence model in each capability; A scoring result determination module, used to fit the multiple attribute features through a deep learning model to score each capability and obtain a scoring result for each capability; The target capability selection module is used to select a target capability from multiple capabilities of the same functional type according to the scoring result.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the capability processing method according to any one of claims 1 to 8 is implemented.

11. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to execute the capability processing method according to any one of claims 1 to 8 by executing the executable instructions.

Images