Data storage system and method for artificial intelligence learning patterns

Abstract

The application provides a data storage system and method for an artificial intelligence learning mode, the system comprising: a collection unit configured to collect short video data of a target object; an identification unit configured to perform artificial intelligence model identification on the short video data to obtain a first calculation result of the short video data, and determine a first storage strategy of the short video data according to the first calculation result; and an updating unit configured to mark input data of the short video according to the storage strategy to obtain marked input data after the short video is stored according to the storage strategy, and input the marked input data into an artificial intelligence model to perform learning training to obtain a trained artificial intelligence model. The application has the advantages of reducing cost and improving storage utilization.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data storage, in particular to a data storage system and method for artificial intelligence learning mode. BACKGROUND

[0002] Informationization refers to the historical process of cultivating and developing new productivity represented by intelligent tools mainly based on computers and making them benefit the society. With the deepening of the informationization process, more and more data needs to be stored. However, the storage space of any device is limited, and the effectiveness of stored data cannot be guaranteed. Therefore, intelligent processing of data storage is needed to improve the utilization of storage and reduce the cost of data storage.

[0003] With the rise of short videos, more and more users communicate through short video platforms. Therefore, short video shooting and short video storage have become the main data storage of user devices. However, the existing short video data storage determines whether to store based on the user's own selection. This method is not convenient for users and cannot intelligently identify whether the data should be stored, affecting the user experience. SUMMARY

[0004] The embodiments of the present application provide a data storage system and method for artificial intelligence learning mode, which can intelligently identify whether to perform data storage on short videos, reduce unnecessary short video storage, improve the utilization of storage space, reduce the cost of data storage, and improve the user experience.

[0005] In a first aspect, the embodiments of the present application provide a data storage system for artificial intelligence learning mode, which comprises:

[0006] A collection unit configured to collect short video data of a target object;

[0007] An identification unit configured to perform artificial intelligence model identification on the short video data to obtain a first calculation result of the short video data, and determine a first storage strategy of the short video data according to the first calculation result;

[0008] The determination of the first storage strategy according to the first calculation result specifically includes:

[0009] If the first calculation result is greater than a first threshold, the first storage strategy is determined to store the short video data; if the first calculation result is less than or equal to the first threshold, the first storage strategy is determined not to store the short video data;

[0010] The updating unit is configured to mark input data of the short video according to the storage strategy to obtain marked input data, input the marked input data into the artificial intelligence model to perform learning training to obtain a trained artificial intelligence model, input the short video data into the trained artificial intelligence model to perform artificial intelligence model recognition to obtain a second calculation result of the short video data, and if the second calculation result is better than the first calculation result, use the trained artificial intelligence model as a recognition model of a subsequent artificial intelligence learning mode.

[0011] In a second aspect, a data storage method for an artificial intelligence learning mode is provided, and the method comprises the following steps:

[0012] The terminal collects short video data of a target object, performs artificial intelligence model recognition on the short video data to obtain a first calculation result of the short video data, and determines a first storage strategy of the short video data according to the first calculation result.

[0013] The first storage strategy of the short video data is determined according to the first calculation result, and the determination specifically comprises:

[0014] If the first calculation result is greater than a first threshold value, it is determined that the first storage strategy is to store the short video data, and if the first calculation result is less than or equal to the first threshold value, it is determined that the first storage strategy is not to store the short video data.

[0015] The terminal marks input data of the short video according to the storage strategy to obtain marked input data, inputs the marked input data into the artificial intelligence model to perform learning training to obtain a trained artificial intelligence model.

[0016] The terminal inputs the short video data into the trained artificial intelligence model to perform artificial intelligence model recognition to obtain a second calculation result of the short video data, and if the second calculation result is better than the first calculation result, uses the trained artificial intelligence model as a recognition model of a subsequent artificial intelligence learning mode.

[0017] In a third aspect, a computer readable storage medium is provided, which stores a program for electronic data exchange, and the program causes a terminal to execute the method provided in the first aspect.

[0018] The embodiments of the present application have the following beneficial effects:

[0019] The technical scheme provided in the application comprises the following steps: a terminal collects short video data of a target object, performs artificial intelligence model recognition on the short video data to obtain a first calculation result of the short video data, and determines a first storage strategy of the short video data according to the first calculation result; after the terminal performs the storage strategy on the short video, the input data of the short video is marked according to the storage strategy to obtain marked input data, the marked input data is input into an artificial intelligence model to perform learning training to obtain a trained artificial intelligence model; the terminal inputs the short video data into the trained artificial intelligence model to perform artificial intelligence model recognition to obtain a second calculation result of the short video data, and if the second calculation result is better than the first calculation result, the trained artificial intelligence model is used as a recognition model of a subsequent artificial intelligence learning mode. Thus, the effectiveness of the short video is improved, the storage cost is reduced, and the user experience is improved on the premise that the short video data can be determined by the operation of the artificial intelligence model. Moreover, the learning mode of the artificial intelligence model is trained in real time, the corresponding artificial intelligence model can be updated in real time, and the accuracy of the artificial intelligence recognition is improved. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 It is a structural schematic diagram of a terminal

[0022] Figure 2 It is a flowchart of a data storage method for an artificial intelligence learning mode.

[0023] Figure 3 It is a flowchart of inputting the marked input data into an artificial intelligence model to perform learning training to obtain a trained artificial intelligence model.

[0024] Figure 4 It is a structural schematic diagram of a data storage system for an artificial intelligence learning mode. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0026] The terms "first", "second", "third", and "fourth" and the like in the description and in the claims of the present application and the accompanying drawings are used for distinguishing between similar objects, not necessarily for describing a particular sequential or chronological order. The terms "comprises", "comprising", "includes", "including" and the like are to be construed open- ended, meaning that they include the listed steps or elements, but not excluding other not listed steps or elements. For example, a process, method, article, or apparatus that comprises a list of steps or elements is not necessarily limited to the listed steps or elements, but can include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

[0027] Reference herein to "an embodiment" means that a particular feature, structure, result, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. It is expressly understood that any of the embodiments described herein can be combined with any of the other embodiments unless specifically noted otherwise.

[0028] Referring to Figure 1 , Figure 1 A terminal is provided, which can be a terminal of an IOS, Android, or the like, and can also be a terminal of other systems, such as Harmony, and the like. The present application does not limit the specific systems described above, such as Figure 1 As shown in FIG. 1, the terminal device can specifically include a processor, a memory, a display screen, a communication circuit, an audio component, and a camera. The above components can be connected by a bus, or can be connected by other means, and the present application does not limit the specific connection manner.

[0029] Referring to Figure 2 , Figure 2 A data storage method for an artificial intelligence learning mode is provided, which is executed by a terminal as shown in FIG. 2, and the method includes the following steps: Figure 2 Figure 1 The method includes the following steps:

[0030] Step S201, the terminal collects short video data of a target object, performs artificial intelligence model recognition on the short video data to obtain a first calculation result of the short video data, and determines a first storage strategy of the short video data according to the first calculation result;

[0031] For example, the first calculation result of the short video data obtained by performing artificial intelligence model recognition on the short video data can specifically include:

[0032] ​The processing is performed on the short video data to obtain input data, the input data is input into an artificial intelligence model to perform forward operation to obtain a forward operation result, and the forward operation result is subjected to activation operation to obtain the first calculation result. The forward operation includes but is not limited to multi-layer convolution operation, full connection operation, etc. The activation function of the activation operation can adopt a general activation function, and the specific type of the activation function is not limited here.

[0033] For example, the first storage strategy of the short video data is determined according to the first calculation result, which can specifically include:

[0034] If the first calculation result is greater than the first threshold value, it is determined that the first storage strategy is to store the short video data, and if the first calculation result is less than or equal to the first threshold value, it is determined that the first storage strategy is not to store the short video data.

[0035] The first threshold value can be dynamically adjusted according to the user's selection. Specifically, the second short video saved and the third short video deleted manually selected by the user are obtained, the second short video is subjected to artificial intelligence model operation to obtain a calculation result i, the third short video is subjected to artificial intelligence model operation to obtain a calculation result i+1, a first difference value between the calculation result i and the first threshold value is calculated, a second difference value between the first threshold value and the calculation result i+1 is calculated, a first ratio value between the first difference value and the second difference value is calculated, if the first ratio value belongs to a first preset range, the first threshold value is not adjusted, if the first ratio value is greater than the first preset range, the first threshold value is adjusted to recalculate the first ratio value until the first ratio value belongs to the first preset range, and the first threshold value is updated with the adjusted value. The way of adjusting the first threshold value includes but is not limited to adjusting by a preset level, for example, increasing or decreasing 0.05 each time. Of course, in actual application, the first ratio value can also be adjusted according to the size of the first ratio value. If the first ratio value is not in the first preset range and the first ratio value is greater than a ratio threshold value (a pre-set empirical value), the first ratio value is increased by 2 or 3 levels, for example, directly increased or decreased by 0.1 or 0.15.

[0036] This way can dynamically adjust the first threshold value, so that the first threshold value can better match the actual situation of the user, and whether to store the short video can better match the user, thereby improving the user experience.

[0037] In step S202, the terminal marks the input data of the short video according to the storage strategy to obtain marked input data, and inputs the marked input data into an artificial intelligence model to perform learning training to obtain a trained artificial intelligence model.

[0038] Step S203, the terminal inputs the short video data into the trained artificial intelligence model to perform artificial intelligence model recognition to obtain a second calculation result of the short video data, and if the second calculation result is better than the first calculation result, the trained artificial intelligence model is taken as a recognition model of a subsequent artificial intelligence learning mode.

[0039] The technical scheme provided in the present application comprises the following steps: a terminal collects short video data of a target object, performs artificial intelligence model recognition on the short video data to obtain a first calculation result of the short video data, and determines a first storage strategy of the short video data according to the first calculation result; the terminal performs the storage strategy on the short video, marks input data of the short video according to the storage strategy to obtain marked input data, inputs the marked input data into an artificial intelligence model to perform learning training to obtain a trained artificial intelligence model; the terminal inputs the short video data into the trained artificial intelligence model to perform artificial intelligence model recognition to obtain a second calculation result of the short video data, and if the second calculation result is better than the first calculation result, the trained artificial intelligence model is taken as a recognition model of a subsequent artificial intelligence learning mode. This scheme can improve the effectiveness of the short video, reduce the storage cost, and improve the user experience, while ensuring that the short video data can be determined by the operation of the artificial intelligence model. Moreover, the above scheme can train the learning mode of the artificial intelligence model in real time, update the corresponding artificial intelligence model in real time, and improve the accuracy of artificial intelligence recognition.

[0040] Referring to Figure 3 , Figure 3 An implementation method for inputting the marked input data into the artificial intelligence model to perform learning training to obtain the trained artificial intelligence model is provided, as shown in Figure 3 The method can specifically comprise the following steps:

[0041] Step S301, inputting the marked input data into the artificial intelligence model to perform n-layer forward operation to obtain a first forward operation result;

[0042] Step S302, inputting the first forward operation result into the artificial intelligence model as reverse input data to perform n-layer reverse operation to obtain a first n-layer weight update coefficient, and using the first n-layer weight update coefficient to update the n-layer weight to obtain an updated artificial intelligence model;

[0043] Step S303, performing activation operation on the first forward operation result to obtain a first calculation result, and calculating a first update difference value by subtracting a first threshold value from the first calculation result;

[0044] Step S304, inputting the mark input data into the updated artificial intelligence model to perform forward operation to obtain a second forward operation result, performing activation operation on the second forward operation result to obtain a second calculation result, calculating a difference between the second calculation result and the first threshold to obtain a second update difference value;

[0045] Step S305, if the second update difference value is greater than the first update difference value, ending the training, determining the un-updated artificial intelligence model as the trained artificial intelligence model, if the second update difference value is less than the first update difference value, performing the next update operation until the update difference value of the next update operation is greater than the update difference value of the last update operation, and then determining the last updated artificial intelligence model as the trained artificial intelligence model.

[0046] For example, the next update operation can include:

[0047] inputting the mark input data into the artificial intelligence model to perform n-layer forward operation to obtain a w-th forward operation result, performing activation operation on the w-th forward operation result to obtain a w-th calculation result, calculating a difference between the w-th calculation result and the first threshold to obtain a w-th difference value, if the w-th difference value is less than the w-1-th difference value, inputting the w-th calculation result as the backward input data into the artificial intelligence model to perform n-layer backward operation to obtain n-layer weight update coefficient, and updating the n-layer weight data to obtain updated n-layer weight data.

[0048] The forward operation and the backward operation can be performed in the manner of the forward operation and the backward operation, and the application does not limit the specific forms of the forward operation and the backward operation.

[0049] The technical solution is based on the distance between the update result and the first threshold to determine whether to continue updating, because according to the mark, if the update is positive, that is, the update does not overfit, the calculation result of the update should be better than the last time. Based on this principle, the technical solution can update the artificial intelligence model to achieve the purpose of positive update, and the technical solution can update the artificial intelligence model through single mark data, which improves the real-time updating of the artificial intelligence model.

[0050] For example, referring to Figure 4 , Figure 4 A data storage system for artificial intelligence learning mode is provided, and the system comprises:

[0051] The acquisition unit is configured to acquire short video data of a target object.

[0052] The identification unit is used to perform artificial intelligence model identification on the short video data to obtain a first calculation result of the short video data, and determine a first storage strategy for the short video data based on the first calculation result.

[0053] The first storage strategy for the short video data determined based on the first calculation result specifically includes:

[0054] If the first calculation result is greater than the first threshold, the first storage strategy is determined to store the short video data; if the first calculation result is less than or equal to the first threshold, the first storage strategy is determined not to store the short video data.

[0055] The update unit is used to execute the storage strategy on the short video, mark the input data of the short video according to the storage strategy to obtain marked input data, input the marked input data into the artificial intelligence model to perform learning and training to obtain the trained artificial intelligence model; input the short video data into the trained artificial intelligence model to perform artificial intelligence model recognition to obtain the second calculation result of the short video data; if the second calculation result is better than the first calculation result, the trained artificial intelligence model is used as the recognition model for subsequent artificial intelligence learning modes.

[0056] For example, the system further includes: an adjustment unit for dynamically adjusting the first threshold, specifically including:

[0057] The system retrieves the second short video manually selected and saved by the user, and the third short video deleted. It then performs an AI model operation on the second short video to obtain a calculation result i, and on the third short video to obtain a calculation result i+1. The system calculates the difference between the calculation result i and a first threshold to obtain a first difference value, and calculates the difference between the first threshold and the calculation result i+1 to obtain a second difference value. Finally, it calculates the ratio of the first difference value to the second difference value to obtain a first ratio value. If the first ratio value falls within a first preset range, the first threshold value is not adjusted. If the first ratio value is greater than the first preset range, the first threshold value is adjusted, and the first ratio value is recalculated until it falls within the first preset range. Finally, the adjusted value is used to update the first threshold value.

[0058] In an example, the updating unit is specifically configured to input the mark input data into the artificial intelligence model to perform n-layer forward operation to obtain a first forward operation result; input the first forward operation result as reverse input data into the artificial intelligence model to perform n-layer reverse operation to obtain a first n-layer weight update coefficient, use the first n-layer weight update coefficient to update the n-layer weight to obtain an updated artificial intelligence model; perform activation operation on the first forward operation result to obtain a first calculation result, calculate a difference between the first calculation result and a first threshold to obtain a first update difference; input the mark input data into the updated artificial intelligence model to perform forward operation to obtain a second forward operation result, perform activation operation on the second forward operation result to obtain a second calculation result, calculate a difference between the second calculation result and the first threshold to obtain a second update difference; if the second update difference is greater than the first update difference, end the training, and determine the un-updated artificial intelligence model as a trained artificial intelligence model; if the second update difference is less than the first update difference, perform a next update operation until the update difference of the next update operation is greater than the update difference of a previous update operation, and then determine the artificial intelligence model updated in the previous update operation as the trained artificial intelligence model.

[0059] In an example, the next update operation specifically includes:

[0060] input the mark input data into the artificial intelligence model to perform n-layer forward operation to obtain a wth forward operation result, perform activation operation on the wth forward operation result to obtain a wth calculation result, calculate a difference between the wth calculation result and the first threshold to obtain a wth difference, if the wth difference is less than a (w-1)th difference, input the wth calculation result as reverse input data into the artificial intelligence model to perform n-layer reverse operation to obtain an n-layer weight update coefficient, and use the n-layer weight data to update the n-layer weight to obtain updated n-layer weight data.

[0061] The embodiment of the present application further provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to execute part or all steps of any one of the data storage methods for artificial intelligence learning mode described in the above method embodiments.

[0062] The embodiment of the present application further provides a computer program product, which includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute part or all steps of any one of the data storage methods for artificial intelligence learning mode described in the above method embodiments.

[0063] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all described as a combination of a series of actions, but those skilled in the art should know that the present application is not limited to the order of the actions described, because according to the present application, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily necessary for the present application.

[0064] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0065] In several embodiments provided in the present application, it should be understood that the disclosed apparatus can be implemented by other means. For example, the apparatus embodiments described above are only schematic, and the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, apparatuses or units, and can be electrical or other forms.

[0066] Those of ordinary skill in the art can understand that all or part of the steps of the various methods in the above embodiments can be completed by a program instructing relevant hardware, and the program can be stored in a computer readable memory, which can include a flash disk, a read-only memory (English: Read-Only Memory, abbreviated: ROM), a random access memory (English: Random Access Memory, abbreviated: RAM), a magnetic disk or an optical disk, etc.

[0067] The above has introduced the embodiments of the present application in detail, and the principles and implementation manners of the present application have been described by applying specific examples; the above embodiment descriptions are only for helping to understand the method of the present application and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range will have changes; in view of the above, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A data storage system for artificial intelligence learning patterns, characterized by, The system comprises: a collection unit configured to collect short video data of a target object; an identification unit configured to perform artificial intelligence model identification on the short video data to obtain a first calculation result of the short video data, and determine a first storage strategy of the short video data according to the first calculation result; the first storage strategy of the short video data according to the first calculation result specifically comprises: if the first calculation result is greater than a first threshold value, the first storage strategy is determined to be storing the short video data, and if the first calculation result is less than or equal to the first threshold value, the first storage strategy is determined to be not storing the short video data; an updating unit configured to, after performing the storage strategy on the short video, mark input data of the short video according to the storage strategy to obtain marked input data, input the marked input data into the artificial intelligence model to perform learning training to obtain a trained artificial intelligence model, input the short video data into the trained artificial intelligence model to perform artificial intelligence model identification to obtain a second calculation result of the short video data, and if the second calculation result is better than the first calculation result, use the trained artificial intelligence model as an identification model of a subsequent artificial intelligence learning mode; the system further comprises an adjusting unit configured to dynamically adjust the first threshold value, specifically comprising: obtaining a second short video selected by a user to be saved and a third short video selected by the user to be deleted, performing artificial intelligence model operation on the second short video to obtain a calculation result i, performing artificial intelligence model operation on the third short video to obtain a calculation result i+1, calculating a first difference value between the calculation result i and the first threshold value, calculating a second difference value between the first threshold value and the calculation result i+1, calculating a first ratio value of the first difference value and the second difference value, if the first ratio value belongs to a first preset range, not adjusting the first threshold value, and if the first ratio value is greater than the first preset range, adjusting the first threshold value to recalculate the first ratio value until the first ratio value belongs to the first preset range, and then updating the first threshold value with the adjusted value.

2. The data storage system for an artificial intelligence learning mode according to claim 1, characterized in that: the updating unit is specifically configured to input the marked input data into the artificial intelligence model to perform n-layer forward operation to obtain a first forward operation result, input the first forward operation result into the artificial intelligence model as backward input data to perform n-layer backward operation to obtain a first n-layer weight update coefficient, use the first n-layer weight update coefficient to update n-layer weights to obtain an updated artificial intelligence model, perform activation operation on the first forward operation result to obtain a first calculation result, and calculate a first update difference value between the first calculation result and the first threshold value; input the marked input data into the updated artificial intelligence model to perform forward operation to obtain a second forward operation result, perform activation operation on the second forward operation result to obtain a second calculation result, and calculate a second update difference value between the second calculation result and the first threshold value. If the second update difference is greater than the first update difference, the training is ended, and the un-updated artificial intelligence model is determined as the trained artificial intelligence model; if the second update difference is less than the first update difference, the next update operation is performed until the update difference of the next update operation is greater than the update difference of the last update operation, and the artificial intelligence model updated last time is determined as the trained artificial intelligence model.

3. The data storage system for artificial intelligence learning pattern of claim 2, wherein, The next update operation specifically includes: The marked input data is input into the artificial intelligence model to perform n-layer forward operation to obtain a w-th forward operation result, the w-th forward operation result is subjected to activation operation to obtain a w-th calculation result, a difference between the w-th calculation result and the first threshold value is calculated to obtain a w-th difference value, if the w-th difference value is less than a w-1-th difference value, the w-th calculation result is input into the artificial intelligence model as reverse input data to perform n-layer reverse operation to obtain n-layer weight update coefficients, and the n-layer weight data is used to update the n-layer weight to obtain updated n-layer weight data.

4. A data storage method for artificial intelligence learning mode, characterized by, The method includes the following steps: The terminal collects short video data of a target object, performs artificial intelligence model recognition on the short video data to obtain a first calculation result of the short video data, and determines a first storage strategy of the short video data according to the first calculation result; The first calculation result is greater than the first threshold value, the first storage strategy is determined as storing the short video data, and if the first calculation result is less than or equal to the first threshold value, the first storage strategy is determined as not storing the short video data; After the terminal performs the storage strategy on the short video, the input data of the short video is marked as marked input data according to the storage strategy, and the marked input data is input into the artificial intelligence model to perform learning training to obtain a trained artificial intelligence model; The terminal inputs the short video data into the trained artificial intelligence model to perform artificial intelligence model recognition to obtain a second calculation result of the short video data, and if the second calculation result is better than the first calculation result, the trained artificial intelligence model is used as a recognition model of subsequent artificial intelligence learning mode; and the first threshold value is dynamically adjusted, specifically including: The saved second short video and the deleted third short video manually selected by a user are obtained, the second short video is subjected to artificial intelligence model operation to obtain a calculation result i, the third short video is subjected to artificial intelligence model operation to obtain a calculation result i+1, a difference between the calculation result i and the first threshold value is calculated to obtain a first difference value, a difference between the first threshold value and the calculation result i+1 is calculated to obtain a second difference value, a ratio of the first difference value to the second difference value is calculated to obtain a first ratio, if the first ratio belongs to a first preset range, the first threshold value is not adjusted, and if the first ratio is greater than the first preset range, the first threshold value is adjusted to recalculate the first ratio until the first ratio belongs to the first preset range, and then the first threshold value is updated with the adjusted value. The marked input data is input into the artificial intelligence model to perform n-layer forward operation to obtain a w-th forward operation result, the w-th forward operation result is subjected to activation operation to obtain a w-th calculation result, a difference between the w-th calculation result and the first threshold value is calculated to obtain a w-th difference value, if the w-th difference value is less than a w-1-th difference value, the w-th calculation result is input into the artificial intelligence model as reverse input data to perform n-layer reverse operation to obtain n-layer weight update coefficients, and the n-layer weight data is used to update the n-layer weight to obtain updated n-layer weight data.

5. The data storage method for artificial intelligence learning pattern according to claim 4, wherein, ​ ​ The first forward operation result is input as reverse input data to the artificial intelligence model to perform n-layer reverse operation to obtain a first n-layer weight update coefficient, and the first n-layer weight update coefficient is used to update the n-layer weight to obtain an updated artificial intelligence model; The first forward operation result is executed to perform activation operation to obtain a first calculation result, and a difference between the first calculation result and the first threshold value is calculated to obtain a first update difference value; The marked input data is input to the updated artificial intelligence model to perform forward operation to obtain a second forward operation result, the second forward operation result is executed to perform activation operation to obtain a second calculation result, and a difference between the second calculation result and the first threshold value is calculated to obtain a second update difference value; If the second update difference value is greater than the first update difference value, the training is ended, the un-updated artificial intelligence model is determined as a trained artificial intelligence model, if the second update difference value is less than the first update difference value, a next update operation is performed until an update difference value of the next update operation is greater than an update difference value of a previous update operation, and then the artificial intelligence model updated last time is determined as the trained artificial intelligence model.

6. The method of claim 5, wherein, The next update operation specifically includes: The marked input data is input to the artificial intelligence model to perform n-layer forward operation to obtain a wth forward operation result, the wth forward operation result is executed to perform activation operation to obtain a wth calculation result, a difference between the wth calculation result and the first threshold value is calculated to obtain a wth difference value, if the wth difference value is less than a (w-1) th difference value, the wth calculation result is input as reverse input data to the artificial intelligence model to perform n-layer reverse operation to obtain an n-layer weight update coefficient, and the n-layer weight data is used to update the n-layer weight to obtain updated n-layer weight data.

Images