WO2020133324A1

WO2020133324A1 - Methods and apparatuses for artificial intelligence application building and operational implementation, and machine device

Info

Publication number: WO2020133324A1
Application number: PCT/CN2018/125255
Authority: WO
Inventors: 薛俊恩; 谈国禹
Original assignee: 深圳砥脊科技有限公司
Priority date: 2018-12-29
Filing date: 2018-12-29
Publication date: 2020-07-02
Also published as: CN111819536A

Abstract

A method for building an artificial intelligence application, a method for operational implementation during artificial intelligence application building, apparatuses, and a machine device. The method comprises: receiving a selection instruction for blocks on a graphical interface, wherein the blocks are graphical representations of corresponding mathematical primitives (310); by means of the selection instruction, configuring in a building area the blocks for artificial intelligence application building, and linking the blocks with each other to form an artificial intelligence algorithm configuration of an artificial intelligence application built under the control of a user (330); according to the artificial intelligence algorithm configuration, obtaining a dictionary composed of mathematical primitive identifiers corresponding to the blocks and core parameters, the core parameters being configured to correspond to the blocks (350); by means of the dictionary, triggering the artificial intelligence application to operate on a server end. The method meets given artificial intelligence application requirements, implements artificial intelligence application building by means of components corresponding to the mathematical primitives, accurately adapts to real needs, freely implements "what you see is what you get" artificial intelligence applications, enhances interaction performance, and lowers thresholds.

Description

Artificial intelligence application construction, operation implementation method, device and machine equipment

Technical field

The present invention relates to the field of Internet application technology, and in particular, to a method for building an artificial intelligence application, a method, a device, and a machine for implementing an operation in building an artificial intelligence application.

Background technique

With the development of artificial intelligence technology, more and more artificial intelligence applications are used in many scenarios to make various decisions based on artificial intelligence technology for the data obtained. Corresponding to this, all kinds of users, for example, enterprise users and end users, can benefit from the application of artificial intelligence.

Users will have artificial intelligence application needs in many scenarios, and then develop or download the published artificial intelligence applications to achieve the required artificial intelligence technology.

For enterprise users, they often develop the artificial intelligence applications they need to obtain artificial intelligence applications that can meet their own business needs. For example, for enterprise users who need to realize their own data processing through artificial intelligence technology, they can only obtain the required artificial intelligence applications through their own development or commissioned development.

For end users, they do not have development capabilities. They can only search for resources on the Internet, search for artificial intelligence applications that meet their own needs, and download them.

However, no matter what kind of users, it is difficult to obtain artificial intelligence applications that meet their own needs, and it is urgent to provide an implementation of artificial intelligence application construction, so that all kinds of users can build artificial intelligence applications based on their own artificial intelligence application needs.

However, the construction of artificial intelligence applications is essentially the development of artificial intelligence applications. No matter what kind of users, it is very difficult to develop artificial intelligence applications independently. The complexity of the algorithms involved in artificial intelligence technology and the difficulty of programming have an insurmountable threshold for the public.

In other words, there is an urgent need to realize the free construction of artificial intelligence applications for users, so as to solve the dilemma for various users that the currently obtained artificial intelligence applications cannot accurately adapt to real needs.

Summary of the invention

In order to solve the technical problem that users cannot achieve the construction of artificial intelligence applications in related technologies, and the obtained artificial intelligence applications cannot be accurately adapted to real needs, the present invention provides a method for building artificial intelligence applications, a method for implementing operation in the construction of artificial intelligence applications, Devices and machinery.

A method for building an artificial intelligence application, the method includes:

Receiving a selection instruction for a block on the graphical interface, the block is a graphical representation of the corresponding mathematical primitive;

Configuring the building blocks of the artificial intelligence application in the building area of the graphical interface through the selection instruction, and linking the blocks to each other to form the artificial intelligence algorithm configuration of the artificial intelligence application built under user control;

Convert the included chunks into a dictionary according to the configuration of the artificial intelligence algorithm to obtain a dictionary composed of the mathematical primitive identifier corresponding to the chunk and the core parameters, the core parameters corresponding to the chunk configuration;

Initiating decoding of the artificial intelligence application built by the user to the server through the dictionary, triggering the artificial intelligence application to run on the server.

An operation implementation method in building an artificial intelligence application, the method includes:

In the construction of the artificial intelligence application selected by the user, the server receives the character string corresponding to the dictionary, and the dictionary corresponding to the character string is used to describe the block configured for the construction of the artificial intelligence application;

Decoding the character string to obtain the executable text of the artificial intelligence application;

Through the execution of the executable text, the artificial intelligence application that the user chooses to build runs on the server.

An artificial intelligence application building device, including:

The instruction receiving module is used to receive a selection instruction of a block on the graphical interface, the block is a graphical representation of the corresponding mathematical primitive.

Construction module for configuring configuration blocks for building artificial intelligence applications in the construction area of the graphical interface through the selection instruction, and linking the blocks to form an artificial intelligence algorithm structure for artificial intelligence applications built under user control type.

The conversion module is used to convert the included chunks to the dictionary according to the configuration of the artificial intelligence algorithm to obtain a dictionary composed of the mathematical primitive identification corresponding to the chunk and the core parameters, the core parameters corresponding to the chunks Configured.

The decoding initiation module is used to initiate decoding of the artificial intelligence application built by the user to the server through the dictionary, and trigger the artificial intelligence application to run on the server.

An operation realization device in the construction of manual and intelligent applications, including:

The character receiving module is used for building an artificial intelligence application selected by the user, and the server receives a character string corresponding to a dictionary, and the dictionary corresponding to the character string is used to describe a block configured for building the artificial intelligence application.

The decoding module is used for decoding the character string to obtain the executable text of the artificial intelligence application.

An execution module is configured to enable the artificial intelligence application selected by the user to run on the server through execution of the executable text.

A machine equipment, including:

Processor; and

A memory, on which computer-readable instructions are stored, which when executed by the processor implements the method described above.

The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

For the given artificial intelligence application requirements, users will freely build artificial intelligence applications in the construction area of the graphical interface. During the construction of artificial intelligence applications, they will receive the instruction to select the block on the image interface. The block is The graphical representation of the corresponding mathematical primitives, and then configure the building blocks for the construction of artificial intelligence applications in the construction area of the graphical interface by selecting instructions, and link the blocks to form the artificial intelligence algorithm configuration of artificial intelligence applications built under user control ，According to the configuration of the artificial intelligence algorithm, the conversion of the included chunks to the dictionary is obtained to obtain the dictionary composed of the mathematical primitive identification corresponding to the chunk and the core parameters. The core parameters are corresponding to the chunk configuration, and finally through the dictionary to the server Initiating the decoding of the artificial intelligence application built by the user can trigger the artificial intelligence application to run on the server, thereby satisfying the given artificial intelligence application needs, and implementing the artificial intelligence application for the user through the component corresponding to the mathematical primitive Since the obtained artificial intelligence applications are configured according to real needs, the blocks can be accurately adapted to the real needs. With the free configuration of the blocks, the WYSIWYG artificial intelligence applications are realized, which enhances the construction of artificial intelligence applications. Interactive performance, and lowered the threshold.

It should be understood that the above general description and the following detailed description are only exemplary and do not limit the present invention.

BRIEF DESCRIPTION

The drawings herein are incorporated into and constitute a part of the specification, show embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

FIG. 1 is a schematic diagram of an implementation environment involved in the present invention;

Fig. 2 is a block diagram of an apparatus according to an exemplary embodiment

Fig. 3 is a flowchart of a method for building an artificial intelligence application according to an exemplary embodiment;

FIG. 4 is a flowchart illustrating step 330 according to the embodiment corresponding to FIG. 3;

FIG. 5 is a flowchart illustrating step 350 according to the embodiment corresponding to FIG. 3;

6 is a flowchart illustrating step 370 according to the embodiment corresponding to FIG. 3;

Fig. 7 is a flowchart of a method for building an artificial intelligence application according to another exemplary embodiment;

8 is a flowchart illustrating step 470 shown in another exemplary embodiment according to the corresponding embodiment of FIG. 7;

9 is a flowchart illustrating step 430 according to the embodiment corresponding to FIG. 7;

Fig. 10 is a flowchart of a method for implementing operation in building an artificial intelligence application according to an exemplary embodiment;

FIG. 11 is a flowchart illustrating step 530 according to the embodiment corresponding to FIG. 10;

12 is a flowchart illustrating step 533 according to the embodiment corresponding to FIG. 11;

Fig. 13 is a schematic diagram showing a DAG structure corresponding to a simple mathematical primitive according to an exemplary embodiment;

14 is a flowchart illustrating step 550 according to the embodiment corresponding to FIG. 10;

15 is a flowchart illustrating step 550 according to the embodiment corresponding to FIG. 10 in another exemplary embodiment;

16 is a flowchart illustrating step 551b according to the embodiment corresponding to FIG. 15;

17 is a flowchart of a method for implementing operation in building an artificial intelligence application shown in another exemplary embodiment;

Fig. 18 is a schematic diagram of a graphical interface according to an exemplary embodiment;

Fig. 19 is a schematic diagram of a block corresponding to a convolutional neural network operation on an operation interface according to an exemplary embodiment;

20 is a schematic diagram of block distribution and linking of a three-layer convolutional neural network on an operation interface according to the corresponding embodiment of FIG. 19;

21 is a schematic diagram of interaction between the front end and the back end involved in the present invention according to an exemplary embodiment;

22 shows a schematic diagram of the implementation of the overall solution of the present invention in an exemplary embodiment;

Fig. 23 is a block diagram of a device for building an artificial intelligence application according to an exemplary embodiment;

Fig. 24 is a block diagram of a device for implementing operation in building an artificial intelligence application according to an exemplary embodiment.

detailed description

The exemplary embodiments will be explained in detail here, examples of which are shown in the drawings. When referring to the drawings below, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of devices and methods consistent with some aspects of the invention as detailed in the appended claims.

FIG. 1 is a schematic diagram of an implementation environment involved in the present invention. In an exemplary embodiment, the implementation environment includes a user terminal 110 and a server 130 configured in the background. The user terminals 110 are not limited to a single number, that is to say, all kinds of users can interact with the server 130 through the held user terminals 110 to realize the construction of the artificial intelligence application of the present invention and the operation of the built artificial intelligence application.

The server 130 is accessed by the user terminal 110 for building artificial intelligence applications and running the built artificial intelligence applications. The server 130 will be oriented to massive user terminals 110, and any user can freely build artificial intelligence applications and use the built artificial intelligence applications as long as they can access the server 130.

Under the action of the user terminal 110 and the server 130, the threshold of the artificial intelligence application is effectively lowered, and the user will be able to immediately build and run the artificial intelligence application according to the artificial intelligence application demand generated by the user.

It can be understood that through the implementation of the present invention, the user is provided with a platform capable of building the required artificial intelligence application at any time, but it is not limited to this, no matter whether it is an enterprise user or an end user, it no longer needs to be very expensive for their own needs. Independent development of artificial intelligence applications at cost.

Fig. 2 is a block diagram of a device according to an exemplary embodiment. For example, the device 200 may be the user terminal 110 in the implementation environment shown in FIG. 1. For example, the user terminal 110 is a terminal device held by a user such as a smart phone or a tablet computer, various cameras, and the like.

2, the device 200 includes at least the following components: a processing component 202, a memory 204, a power component 206, a multimedia component 208, an audio component 210, a sensor component 214 and a communication component 216.

The processing component 202 generally controls the overall operations of the device 200, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing component 202 includes at least one or more processors 218 to execute instructions to complete all or part of the steps of the method described below. In addition, the processing component 202 includes at least one or more modules to facilitate interaction between the processing component 202 and other components. For example, the processing component 202 may include a multimedia module to facilitate interaction between the multimedia component 208 and the processing component 202.

The memory 204 is configured to store various types of data to support operation at the device 200. Examples of these data include instructions for any application or method operating on the device 200. The memory 204 is implemented by at least any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, SRAM for short), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), read-only memory ( Read-Only Memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. One or more modules are also stored in the memory 204, and the one or more modules are configured to be executed by the one or more processors 218 to complete all or any of the methods shown in any of the following FIGS. 3 to 17 Partial steps.

The power supply component 206 provides power to various components of the device 200. The power component 206 includes at least a power management system, one or more power sources, and other components associated with generating, managing, and distributing power for the device 200.

The multimedia component 208 includes a screen that provides an output interface between the device 200 and the user. In some embodiments, the screen may include a liquid crystal display (Liquid Crystal) (LCD for short) and a touch panel. If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundary of the touch or sliding action, but also detect the duration and pressure related to the touch or sliding operation. The screen also includes Organic Light Emitting Display (Organic Light Emitting Display, OLED for short).

The audio component 210 is configured to output and/or input audio signals. For example, the audio component 210 includes a microphone (Microphone, MIC for short). When the device 200 is in an operation mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal may be further stored in the memory 204 or sent via the communication component 216. In some embodiments, the audio component 210 further includes a speaker for outputting audio signals.

The sensor assembly 214 includes one or more sensors for providing the device 200 with status assessments in various aspects. For example, the sensor assembly 214 detects the on/off state of the device 200, the relative positioning of the components, and the sensor assembly 214 also detects a change in the position of the device 200 or one component of the device 200 and a change in the temperature of the device 200. In some embodiments, the sensor assembly 214 further includes a magnetic sensor, a pressure sensor, or a temperature sensor.

The communication component 216 is configured to facilitate wired or wireless communication between the device 200 and other devices. The device 200 accesses a wireless network based on a communication standard, such as WiFi (WIreless-Fidelity, wireless fidelity). In an exemplary embodiment, the communication component 216 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 216 further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth technology and other technologies. .

In an exemplary embodiment, the apparatus 200 is controlled by one or more application specific integrated circuits (Application Specific Integrated Circuit (ASIC for short), digital signal processor, digital signal processing device, programmable logic device, field programmable gate array, control A microcontroller, microcontroller, microprocessor or other electronic component is implemented to perform the following method.

Fig. 3 is a flowchart of a method for building an artificial intelligence application according to an exemplary embodiment. In an exemplary embodiment, the method for building an artificial intelligence application, as shown in FIG. 3, includes at least the following steps.

In step 310, a selection instruction for a block on the graphical interface is received, and the block is a graphical representation of the corresponding mathematical primitive.

Among them, the graphical interface is a user interface that can be used for free construction of artificial intelligence applications. The artificial intelligence application construction refers to the configuration of the sequential execution of mathematical operations involved. The mathematical operations configured to perform constitute the algorithm implementation of artificial intelligence applications , And then realize artificial intelligence applications that meet the set requirements.

Optionally, the graphical interface includes an operation component selection bar, an operation interface, a toolbar, and an operation result display area. The selection command applied to the block is triggered in the selection bar of the operation component. The operation component selection bar includes a large number of blocks that can be selected, and these blocks correspond to the initially configured model and/or data. The user can complete the selection of the corresponding block by triggering an operation on the block and generating a selection instruction for the block, so that the selected block is configured on the operation interface.

For the operation component selection bar, on the one hand, it will be used to select the mathematical primitives involved for the currently built artificial intelligence application. For example, the mathematical primitives can be integrated models, on the other hand, they can also be built artificial intelligence. The application selects the data used for the iterative training. No matter what kind of selection, it will exist in the form of a block on the selection bar of the operating component of the graphical interface.

The operation interface, that is, the building area on the graphical interface is used to place the selected blocks for the construction of artificial intelligence applications, and these blocks will be placed under the control of the user to achieve the building. After the block in the operation component selection bar is triggered, it is placed on the operation interface. For example, in the selection of a block, the block in the operation component selection bar can be dragged to the operation interface by a drag operation applied to the block. The block dragged to the operation interface is used for manual operation Smart application building.

As indicated in the foregoing description, a set of blocks corresponds to a set model or data, which is a graphical representation of the set model or data. The set model can be a single mathematical operation, or a model realized by integrating more than two mathematical operations, whether it is a single mathematical operation or the integration of more than two mathematical operations, it is the integration of input, output and mathematical operations The formed complex is an independent unit, also called a mathematical primitive. Mathematical primitives correspond to chunks. Mathematical primitives are configured on the operation interface through chunks according to the algorithms involved in the artificial intelligence applications.

It should be understood that the mathematical primitives corresponding to chunks define mathematical operations on a level. Of course, with different levels, at a more subdivided level, the mathematical operations defined by the mathematical primitives will be subdivided, so that the block corresponds to several mathematical operations. Mathematical primitives will define the mathematical operations corresponding to the chunk from the input, output and the mathematical operations performed. In the construction of the artificial intelligence application, along with the configuration of the block, the configuration of the mathematical primitives involved in the artificial intelligence application is carried out, in order to deploy the mathematical operations performed in the constructed artificial intelligence application, and for the execution of each mathematical operation Control its input and output.

The toolbar on the graphical interface is used in the construction of artificial intelligence applications to adjust the corresponding data and models for the configured blocks, for example, to fine-tune the data and models.

The operation result display area on the graphical interface is used to display the results after the built artificial intelligence application runs. For example, after iterative training is performed on the currently built artificial intelligence application, the corresponding iterative training situation and the classification effect that can be obtained are displayed in the operation result display area.

In summary, in the graphical interface displayed on the user terminal for the progress of the artificial intelligence application, by configuring the building block in the initializing block, the block configured in the building area is used. For the construction of artificial intelligence applications.

From this, it can be determined that the blocks configured in the building area are the blocks selected for the artificial intelligence application currently built. Correspondingly, the mathematical primitives corresponding to the selected blocks will be used to form the current building The built artificial intelligence application, for the artificial intelligence application construction, has realized the WYSIWYG artificial intelligence application construction, whether it is the degree of freedom or the self-adaptability of the built artificial intelligence application, can be enhanced.

The graphical interface is configured for the user to initialize the building blocks for artificial intelligence applications, that is, building the optional blocks for the current artificial intelligence applications. In addition, the dragging on the required selected blocks Drag operation, selection operation, etc., have been able to achieve the construction area, that is, as mentioned above, the addition of the selected block on the operation interface, that is to say, will follow the selection instruction of the block on the graphical interface The block selected by the user is configured in the building area.

The selection instruction indicates the block selected by the user. As mentioned above, the block is a graphical representation of the corresponding mathematical primitive. Therefore, the corresponding mathematical primitive identifier can be used to uniquely identify the block. The selection instruction carries the mathematical primitive identifier to indicate the user's selection. And execute the corresponding response for this.

Whenever a block selection is performed, the process of realizing the construction of the artificial intelligence application will receive a selection instruction for the corresponding block on the graphical interface. By analogy, as the selection of chunks continues, selection instructions corresponding to different chunks will be continuously received.

In an exemplary embodiment, step 310 includes: among the several blocks of the initial configuration of the graphical interface, a user's operation applied to the chunk receives the instruction to select the chunks on the graphical interface until the artificial intelligence application is implemented All blocks are selected.

Among them, as mentioned above, the graphical interface is initially configured with a large number of blocks, for example, the existence of blocks in the operation component selection bar. Users can continue to apply user operations to the required blocks according to the needs of artificial intelligence application building, such as dragging operations to the building area, so as to generate selection instructions for the continuously selected blocks, and the corresponding process will continue to receive To the generated selection instruction until all the required blocks are selected.

In step 330, the configuration block of the artificial intelligence application is configured in the construction area of the graphical interface by selecting instructions, and the blocks are mutually linked to form an artificial intelligence algorithm configuration of the artificial intelligence application built under user control.

Among them, as mentioned above, as the selection instruction is received, the building block for the artificial intelligence application is continuously configured in the construction area of the graphical interface, and the configured block corresponds to the mathematical primitive identifier carried in the selection instruction.

Under the action of the selection instruction, the building blocks of the graphical interface are configured with blocks corresponding to the involved mathematical primitives for artificial intelligence applications, and thus there are more than two blocks in the building area of the graphical interface.

It should be understood that as the selection of blocks on the graphical interface proceeds, the selected blocks are scattered in the construction area of the graphical interface. At this time, the interconnection between the blocks will be performed under the control of the user to obtain the artificial intelligence algorithm configuration of the built artificial intelligence application.

The artificial intelligence algorithm configuration is a network topology that describes the mathematical operations performed in artificial intelligence applications, in other words, the network topology formed by the deployed mathematical primitives. For interlinked mathematical primitives, the output of the previous mathematical primitive will be used as the input of the next mathematical primitive, and so on to form the entire artificial intelligence algorithm configuration.

In an exemplary embodiment, the artificial intelligence algorithm configuration exists in the form of a graph, that is, it may be in the form of an artificial intelligence algorithm configuration graph. The configuration of the artificial intelligence algorithm indicates the blocks included in the constructed network topology and the link relationship between the contained blocks.

On the graphical interface available for building artificial intelligence applications, as the user operates to freely select and configure the blocks for the built artificial intelligence application, and build the link relationship between the selected configuration blocks, these will be operated by the user According to the needs of its own artificial intelligence applications, it is free to implement the graphical programming built by artificial intelligence applications, in order to reduce the threshold to the maximum extent.

In step 350, according to the configuration of the artificial intelligence algorithm, the included chunks are converted into dictionaries to obtain a dictionary composed of the mathematical primitive identification corresponding to the chunks and the core parameters, the core parameters corresponding to the chunk configuration.

Among them, as mentioned above, the configuration of the artificial intelligence algorithm indicates the blocks included in the network topology of the currently built artificial intelligence application and the link relationship between the blocks, and the blocks correspond to mathematical primitives, so The conversion of chunks into dictionaries according to the configuration of artificial intelligence algorithms is the transformation of chunks into dictionary data. The obtained dictionary data is used to form a dictionary for realizing artificial intelligence applications.

That is to say, for the artificial intelligence algorithm configuration, the dictionary data will be generated for the chunks to realize the conversion of the chunks contained in the artificial intelligence algorithm configuration to the dictionary. The dictionary data generated for the chunks are used to indicate the corresponding mathematical primitives, and input and output control of the mathematical primitives are performed.

Optionally, the configuration of the artificial intelligence algorithm is oriented to the network topology formed by the blocks, and the mathematical primitive identification and the core parameters configured for the mathematical primitive are obtained according to the mathematical primitives corresponding to the chunk, and the mathematical primitive identification For index items, the core parameter is the index value to construct the dictionary data corresponding to this group of blocks, and so on, the dictionary data corresponding to all the groups of blocks constitute the dictionary of the artificial intelligence application.

Here, it should be added that the core parameters correspond to the identification of mathematical primitives, that is, the core parameters are configured for the corresponding mathematical primitives. The core parameters configured corresponding to the mathematical primitives will be obtained by fine-tuning the data and models of the corresponding components in the construction area on the toolbar.

In an exemplary embodiment, for many algorithms involved in artificial intelligence applications, the core parameters corresponding to mathematical primitives include key data such as hyperparameters, input dimensions, and output dimensions applicable to the model. The configuration of the core parameters will ensure the smooth execution of the mathematical operations defined by the corresponding mathematical primitives.

In step 370, the decoding of the artificial intelligence application built by the user is initiated to the server through the dictionary, and the artificial intelligence application is triggered to run on the server.

Among them, after a dictionary is converted for the currently built artificial intelligence application, the decoding of the built artificial intelligence application can be initiated to the server through the dictionary to trigger the operation of the artificial intelligence application on the server.

The dictionary is obtained by the user building the artificial intelligence application, which will enable the server to obtain the built artificial intelligence application in the form of a dictionary through the mathematical primitive identification and core parameters of the included dictionary data records.

Through this exemplary embodiment, an artificial intelligence application based on a graphical interface is built on the user side realized by the user terminal, that is, graphical programming, and after the user completes the construction of the artificial intelligence application, the built artificial intelligence application uses block-oriented The dictionary generated by the corresponding mathematical primitives enables the user to build the artificial intelligence application that can be learned by the server and then runs on the server to meet the user's artificial intelligence needs and obtain what the built artificial intelligence application can provide. Function.

For the construction of artificial intelligence applications, under the action of the graphical interface and the blocks corresponding to the mathematical primitives, the development of complex algorithms involved in artificial intelligence applications can be realized for users, that is, through the selection of different blocks and Link to each other to build the complex algorithms needed, but users do not need to have code development capabilities and learn artificial intelligence expertise, only need to roughly understand the function of the mathematical primitives corresponding to each block, which eliminates labor for the public The threshold for building smart applications.

The exemplary embodiment described above is to encapsulate the algorithms involved in artificial intelligence applications, that is, mathematical operations and their input and output into mathematical primitives, to present to the user in the form of chunks, on the basis of which the graphical interface The construction of the application of artificial intelligence was carried out.

FIG. 4 is a flowchart illustrating step 330 according to the embodiment corresponding to FIG. 3. In an exemplary embodiment, this step 330 is shown in FIG. 4 and includes at least:

In step 331, the block indicated by the selection instruction is placed in the construction area of the graphical interface.

As the selection instruction is received, the building area will be configured with blocks according to the received selection instruction, so that the block indicated by the selection instruction is added to the construction area of the graphical interface.

The execution of step 331 is a step-by-step process of adding the required mathematical primitives for the currently built artificial intelligence application, in order to finally construct the artificial intelligence algorithm configuration that realizes the artificial intelligence application, that is, the artificial intelligence application corresponding mathematical basis Meta network topology.

In step 333, the corresponding core parameters are obtained for the blocks placed in the building area.

Among them, as the user's operation progresses, the selection instruction is continuously triggered and generated, so that the group selected by the user can be continuously added in the building area. For each block, the corresponding core parameters can be configured and adjusted to be suitable for the artificial intelligence applications currently built.

In an exemplary embodiment, for a group of blocks placed in the construction area, the configuration and adjustment of the core parameters of the corresponding mathematical primitives are performed. After the configuration and adjustment of the core parameters of the group of blocks, other The core parameter configuration and adjustment of the block.

For example, the core parameter configuration and adjustment of the block can be initiated by selecting the block in the construction area. Specifically, after selecting a group of blocks in the construction area, the toolbar on the graphical interface is used to configure and adjust the core parameters of this group of blocks. At this time, the user only needs to set parameters on this toolbar Configuration and adjustment.

The core parameters include the hyperparameters used by the model in the mathematical primitives as well as the input and output dimensions. The transformation of the link relationship of the corresponding chunks not only needs to transform the other chunks linked by the chunks in the building area, but also should adapt to the mathematical primitives of the linked chunks and adjust the input dimension and/or output dimension to adapt to the dynamics Changed artificial intelligence application building.

In an exemplary embodiment, step 333 includes: selecting the block for the graphical interface building area, and obtaining the core parameter corresponding to the block according to the core parameter configuration performed by the user on the block.

Among them, as pointed out in the foregoing description, the building area is distributed with at least one block, and any group of blocks can initiate the corresponding core parameter configuration process through its selection on the building area. Once a group of blocks is selected in the building area, the core parameter configuration of this group of blocks can be configured.

The core parameter configuration is the process of inputting parameters, adjusting parameters and selecting parameters under the control of the user. Different mathematical primitives corresponding to different blocks will also correspond to different core parameter configuration processes.

In any case, the core parameters corresponding to the selected blocks in the building area will be obtained through the core parameter configuration, and the core parameters corresponding to all the blocks in the building area will be obtained by analogy, that is, all The core parameters of the block are configured to ensure the operation of the artificial intelligence application built.

In step 335, with the configuration of more than two blocks in the building area, the interconnection between the blocks is formed to form the artificial intelligence algorithm configuration of the artificial intelligence application built under user control.

Among them, for more than two blocks stored in the building area, the links between the blocks will be controlled under the user's control, so that the blocks distributed in the building area can be built into artificial intelligence algorithms for artificial intelligence applications structure.

The link between the blocks, on the one hand, refers to the connection of the blocks in the building area. In addition to indicating the connection between the blocks, it also indicates the input and output relationship between the connected blocks; the other In terms of aspects, the link between the group blocks also indicates that the output of the previous group of blocks will be used as the input of the next group of linked blocks.

By analogy, with the end of the added blocks in the building area and the setting of links between the blocks, an artificial intelligence algorithm configuration is formed. This artificial intelligence algorithm configuration is an algorithm description of the artificial intelligence application currently built by the user. For the artificial intelligence application currently built by the user, the running process will be the process of executing the algorithm according to the mathematical primitives and the interlinking relationship indicated by the blocks in the corresponding artificial intelligence algorithm configuration.

In an exemplary embodiment, the interconnection between the blocks is realized by the wiring between the blocks under user operation. The resulting artificial intelligence algorithm configuration is realized under the control of the user. The user control referred to here refers to the block selection operation and the connection operation between the blocks that the user needs to trigger when building the artificial intelligence application. There is no limitation here. Any operation that can configure blocks for the built artificial intelligence application and build links between the configured blocks is a user operation triggered by forming an artificial intelligence algorithm configuration.

Through this exemplary embodiment, an artificial intelligence algorithm configuration is built for the currently built artificial intelligence application, so as to realize algorithm development, enhance users to build artificial intelligence algorithm configurations at will for their own needs, and at the same time increase flexibility Achieved the required artificial intelligence application building, which is no longer limited by the lack of professional knowledge such as code knowledge, artificial intelligence algorithms and mathematical representation.

FIG. 5 is a flowchart illustrating step 350 according to the embodiment corresponding to FIG. 3. In an exemplary embodiment, as shown in FIG. 5, the step 350 includes:

In step 351, for the blocks included in the configuration of the artificial intelligence algorithm, the mathematical primitive identification and core parameters corresponding to the blocks are obtained.

Among them, the artificial intelligence algorithm configuration is an algorithm logic description of the models and data used in the built artificial intelligence application, that is, the algorithm logic composed of the introduced mathematical primitives, and as pointed out in the foregoing description, the artificial intelligence algorithm configuration is The network topology composed of all the multiple blocks.

After the user completes the addition and linking of the blocks in the building area by manipulating the blocks, the blocks distributed in the building area and the relationship between them form the block architecture of the artificial intelligence application, which indicates that the artificial Mathematical primitives used in intelligent application building and the linking relationship between mathematical primitives.

In order to realize the execution of the corresponding mathematical operation, each mathematical primitive has a corresponding code description, that is, code information including core parameters, so as to achieve the execution of the corresponding mathematical operation through the execution of the corresponding code description. The configuration of artificial intelligence algorithms composed of blocks, the existence of corresponding mathematical primitives will indicate the code description information for performing a series of mathematical operations under this artificial intelligence algorithm configuration, that is, the corresponding and containing The executable text composed of the code information of the core parameters.

Based on this, for the building area where the addition of blocks and links have been completed, the construction of the artificial intelligence algorithm configuration has been completed accordingly. Under this artificial intelligence algorithm configuration, a dictionary is generated for the artificial intelligence application built by acquiring the mathematical primitive identification and core parameters corresponding to each block.

It should be understood that the generated dictionary is used to deliver the artificial intelligence algorithm configuration adaptively constructed by the user to the server, so as to obtain the artificial intelligence application running on the server.

Each block has a corresponding mathematical primitive, that is, the block is a graphical representation of the corresponding mathematical primitive. Therefore, for the block under the artificial intelligence algorithm framework, the corresponding mathematical primitive identifier can be obtained , And the core parameters are obtained along with the core parameter configuration of the chunks described above. For a group of blocks, the obtained core parameters correspond to the identification of mathematical primitives.

In step 353, a dictionary of artificial intelligence applications is constructed with the mathematical primitive identifier as the index item and the core parameter as the index value.

Among them, the dictionary data is constructed for each block under the configuration of the artificial intelligence algorithm, that is, the corresponding mathematical primitive identification is used as the index item, and the core parameter is used as the index value to construct the dictionary data for the block, and so on, The dictionary data of all the blocks under the configuration of artificial intelligence algorithm will form the dictionary of artificial intelligence application.

Under the function of the dictionary, the image information of the chunk is converted into code that exists on the server side, that is, the executable text of the artificial intelligence application is obtained, so as to solve the dilemma of the artificial intelligence algorithm is complex and difficult to adapt to user development.

In addition, by generating a dictionary, the core parameters of the artificial intelligence application built by the user will also be passed to provide the server with the core parameters of the user's personalized configuration, ensuring the accuracy of the construction of the artificial intelligence application Adapt to the user's artificial intelligence application needs.

FIG. 6 is a flowchart illustrating step 370 according to the corresponding embodiment of FIG. 3. In an exemplary embodiment, as shown in FIG. 6, the step 370 includes at least:

In step 371, character string conversion is performed on the mathematical primitive identification and indexed core parameters in the dictionary.

After generating the dictionary for the artificial intelligence application built by the user through the foregoing exemplary embodiment, the dictionary needs to be converted into a string, so that the dictionary carries the mathematical primitive identification and core parameters from the user terminal to the server.

In an exemplary embodiment, the mathematical primitive identification in the dictionary and the core parameters of the index are converted into JSON strings, and the dictionary will be transmitted to the server in the form of JSON strings.

In step 373, the artificial intelligence application built for the user transmits a character string to the server, and initiates decoding of the character string by the server through transmission of the character string to obtain executable text of the artificial intelligence application to run on the server.

After receiving the character string sent by the user terminal to the built artificial intelligence application, the server converts the character string decoding back into the form of a dictionary, and then decodes the mathematical language representation of graph theory to obtain executable text.

The executable text is the code description of the artificial intelligence application built by the user. The execution of executable text executes a series of operations configured by the user to realize the artificial intelligence application. This process is the running process of the artificial intelligence application.

Through this exemplary embodiment, the built artificial intelligence application can exist on the server side, and then the functions configured in the artificial intelligence application can be fully realized through excellent hardware conditions on the server side.

Fig. 7 is a flowchart illustrating a method for building an artificial intelligence application according to another exemplary embodiment. In an exemplary embodiment, as shown in FIG. 7, the method for building an artificial intelligence application further includes the following steps:

In step 410, the block configured in the building area receives a user's new block selection instruction, and the new block selection instruction acts on at least one block configured in the building area.

Among them, the building area is under the control of the user to add at least one block. This user-controlled process of adding blocks to the building area can be the process of building an artificial intelligence application, or it can be built for subsequent artificial intelligence applications. The process of adding blocks.

Whether it is in the construction of artificial intelligence applications, or in other cases, new blocks can be added, so that the initial configuration of the graphical interface block is added according to user needs, which is convenient for the subsequent construction of artificial intelligence applications.

In the construction of artificial intelligence applications by users, there are situations where several blocks often need to be combined together. For these blocks, they can be combined together to build a new block. The mathematical primitives corresponding to the newly added blocks are the combination of the mathematical primitives corresponding to these blocks.

In the construction area, as the user selects at least one block, a new block can be added to the selected at least one block, and a new block selection instruction can be generated. The newly added block selection instruction is applied to at least one block selected by the user in the building area.

In step 430, a dictionary corresponding to the block corresponding to the newly added block selection instruction is generated, and the dictionary includes a mathematical primitive identifier corresponding to the block and core parameters.

After receiving the new block selection instruction triggered by the user, a dictionary is generated for the block acting on the new block selection instruction.

Similar to building artificial intelligence applications, the addition of new blocks still requires the generation of a dictionary for this. The dictionary generated for adding new blocks describes the corresponding mathematical primitive identification and core information of the combined blocks. For the newly added chunks and the corresponding newly added mathematical primitives, the generated dictionary is described by data for this purpose.

In at least one of the blocks that are combined to form a new block, in the dictionary generated for the new block, the mathematical primitive identifier uniquely identifies the corresponding block, and the core parameters corresponding to the block are Contains the input dimension and output dimension corresponding to the mathematical operation, so it will indicate the link relationship between the chunks.

Specifically, the core parameters of the input dimension and the output dimension are specified for the chunks, that is, the mathematical primitives, which will be used to determine the links between chunks. If the output dimension of the mathematical primitive corresponding to one group of blocks is the input dimension of the mathematical primitive corresponding to another group of blocks, it indicates that the two groups of blocks are linked to each other.

Therefore, through the generated dictionary, the grouped blocks and the link relationship between the blocks can be accurately known without consuming excessive storage space and more calculation cost.

In step 450, the mathematical primitive identification and core parameters in the dictionary are combined and added into the newly added mathematical primitives through the mathematical language representation of graph theory.

Among them, the dictionary is used as the data description of the newly added mathematical primitives. The data description in the dictionary is represented as a mathematical language of graph theory according to the mathematical primitive identification and its corresponding core parameters, so as to obtain several block combinations New mathematical primitives formed by encapsulation.

First of all, it should be noted that the mathematical language representation of graph theory is to track the input values in the basic mathematical primitives, the mathematical operations done on the input values, and the output values, in order to reconstruct and characterize the mathematical primitives. No matter the input value, mathematical operation or output value, they will be reconstructed in the form of nodes, and then form the mathematical language representation of mathematical primitives in graph theory. Through the mathematical language representation of graph theory, the newly added mathematical primitives can be reconstructed accurately and quickly.

Understandably, for the artificial intelligence application built, after the server obtains the character string and decodes it and converts it back into a dictionary, it reconstructs the mathematical language representation of each mathematical primitive on the graph theory based on the dictionary, in order to quickly repeat Construct mathematical operations performed by artificial intelligence applications and the input and output dimensions of the performed mathematical operations.

The artificial intelligence application obtained by the construction of multiple blocks is composed of several mathematical primitives. The newly added blocks constitute an executable program, similar to it, which is also built by at least one or more blocks As a result, the reconstruction of the mathematical primitives of the newly added blocks is also based on the reconstruction of the mathematical language representation of each mathematical primitive in the graph theory based on the dictionary.

In step 470, the new mathematical primitives are updated to the graphical interface and the server.

After obtaining the mathematical primitives corresponding to the newly added chunks through the execution and reconstruction of step 450, it is necessary to update the graphical interface and the server. On the one hand, there are new blocks among the many blocks that the graphical interface can provide to users, so that users can add new blocks to the building area. On the other hand, it will also enable the server to be used by the graphical interface. The newly added chunks are configured during the initialization of the chunks, and the data related to the newly added chunks are identified in the received dictionary, so that the mathematical primitives corresponding to the newly added chunks can be run on the server.

The new mathematical primitives are updated, including the image information related to the block. In addition, the new mathematical primitives are uniquely identified and the mathematical primitives are stored for all mathematical primitives that constitute the new mathematical primitives. Identification and core parameters. Therefore, the update of the newly added mathematical primitives also includes the update of mathematical primitive definition information such as mathematical primitive identification and core parameters.

Both the user terminal and the service end update the mathematical primitive definition information, and through the updated mathematical primitive definition information, the mathematical operations performed by the newly added mathematical primitives and other information related to this can be known. The updated mathematical primitive definition information describes the newly added mathematical primitives and the execution of mathematical operations in the newly added mathematical primitives.

Through this exemplary embodiment, the function of adding new blocks is provided for the construction of artificial intelligence applications, so that users can build configuration blocks according to their own application requirements and habits, in addition to the existence of the graphical interface In addition to the original configuration blocks, there will also be user-defined configuration blocks, and the convenience and freedom of building artificial intelligence applications will be enhanced.

In this exemplary embodiment, the blocks that are frequently combined for artificial intelligence application building are packaged and packaged together to form a new block. In other words, it is to combine the original more detailed mathematical operations into a new mathematical operation, in order to build a new operation for the artificial intelligence application performed by the user, and enhance the execution performance of the artificial intelligence application.

In an exemplary embodiment, step 470 includes: adding a new block to the graphical interface for the newly added mathematical primitive according to the configured block name, so as to update the added new block among the several blocks of the initial configuration of the graphical interface Add chunks.

Among them, the newly added mathematical primitives must have corresponding newly added blocks on the graphical interface, that is to say, the newly added mathematical primitives must be represented in the form of newly added blocks on the graphical interface. In the graphical interface, the blocks that represent mathematical primitives, also called tiles, are the targets that users manipulate to build artificial intelligence applications.

It should be understood that, in order to facilitate user identification, the newly added blocks corresponding to the newly added mathematical primitive configuration on the graphical interface are marked with the names of the corresponding blocks to distinguish them from other blocks. It is convenient for users to choose.

FIG. 8 is a flowchart illustrating step 470 shown in another exemplary embodiment according to the corresponding embodiment of FIG. 7. In another exemplary embodiment, as shown in FIG. 8, this step 470 includes at least:

In step 471, a new mathematical primitive identifier is generated for the new mathematical primitive.

Among them, the newly added mathematical primitive logo is similar to the aforementioned mathematical primitive logo, and is used to uniquely identify the corresponding mathematical primitive and block.

In the configuration of core parameters for adding new mathematical primitives by adding new blocks, the configured core parameters correspond to the generated new mathematical primitive identifiers.

In step 473, the mathematical primitive definition information and the core parameter of the newly added mathematical primitive package are generated under the newly added mathematical primitive identifier.

Among them, the newly added chunks are composed of at least one chunk. Correspondingly, the newly added mathematical primitives corresponding to the newly added chunks are also formed by packaging and packaging at least one mathematical primitive. The new mathematical identifier generated for the new mathematical primitive corresponds to at least one set of mathematical primitive identifiers and core parameters.

A set of mathematical primitive identifiers and core parameters used to describe the digital primitives of the corresponding block. Together with other sets of mathematical primitive identifiers and core parameters, describe the new mathematical primitives, that is, generate math for the new mathematical primitives Primitive definition information.

It should be understood that there are corresponding mathematical primitive definition information for the blocks that are available for users to choose from in the graphical interface, so as to provide corresponding algorithms for artificial intelligence applications built by users based on this.

In step 475, the mathematical primitive definition information is updated to the server. The updating of the mathematical primitive definition information on the server enables the newly added mathematical primitive to add the corresponding block in the initialization of the graphical interface by the server.

Among them, the existence of mathematical primitive definition information on the server means that the corresponding newly added mathematical primitives and blocks are deployed in the subsequent artificial intelligence application building, and the subsequent artificial intelligence for the user through the graphical interface Application building, you can use the newly added blocks, and then configure the newly added mathematical primitives in the built artificial intelligence applications.

FIG. 9 is a flowchart illustrating step 430 according to the embodiment corresponding to FIG. 7. In an exemplary embodiment, as shown in FIG. 9, the step 430 includes at least:

In step 431, the corresponding mathematical primitive identification and core parameters are obtained respectively according to all the blocks acting on the newly added block selection instruction.

Among them, as mentioned above, the newly added chunks are implemented corresponding to at least one chunk selected by the user, which is to combine at least one mathematical primitive corresponding to the chunks and form a new group for this purpose Piece. Once the user chooses to add a new block, it is equivalent to selecting at least one corresponding block, which will enhance the convenience and efficiency of artificial intelligence application building, and can quickly build artificial intelligence applications.

For at least one block selected for building block addition in the building area, the block interconnections form a network topology capable of performing at least one mathematical operation. At this time, since the block blocks are linked to each other, The input and output dimensions of the core parameters corresponding to the chunks have been configured according to the currently constructed link relationship, and the corresponding mathematical primitive identification and core parameters can be obtained for each chunk to be used for each group The block generates dictionary data, and the dictionary data indicates the link relationship between the located block and the other linked blocks.

In step 433, for each group of blocks, a corresponding mathematical primitive identifier is used as an index item, and core parameters are index values to construct dictionary data to generate a dictionary corresponding to the newly added group of blocks.

Among them, dictionary data is constructed for each group of blocks through a key-value pair data structure, where the key-value pair data structure uses mathematical primitives as index items and core parameters as index values.

Dictionary data is constructed separately for all the blocks that make up the newly added blocks, and the dictionary corresponding to the newly added blocks is generated from all dictionary data.

In addition, corresponding to the method for building an artificial intelligence application provided above, the following is the method implementation of the present invention in the server, that is, the operation of the artificial intelligence application building applied to the server in the implementation environment shown in FIG. 1 Implementation. Fig. 10 is a flow chart showing a method for implementing operation in building an artificial intelligence application according to an exemplary embodiment. In an exemplary embodiment, as shown in FIG. 10, the operation implementation method in the construction of an artificial intelligence application includes at least the following steps.

In step 510, when the user chooses to build an artificial intelligence application, the server receives the character string corresponding to the dictionary, and the dictionary corresponding to the character string is used to describe the building blocks configured for the artificial intelligence application.

Among them, in order to realize the construction of the artificial intelligence application in the user terminal, the deployed server will respond to the manipulation of the user terminal and cooperate with the realization of the construction of the artificial intelligence application to obtain the artificial intelligence application existing on the server for the user.

The operation of artificial intelligence applications is achieved through the execution of a series of mathematical operations, that is, the operations required by users are realized in units of mathematical primitives. The execution of mathematical operations must be supported by the corresponding code information to control the execution of the mathematical operations on the server side. Therefore, the server side stores the information related to the code for the deployed blocks accordingly, which requires users to retrieve them separately under the control of the dictionary generated by the built artificial intelligence application to obtain executable text that can realize the operation of the artificial intelligence application. .

Based on this, as the user builds the artificial intelligence application on the user terminal, the dictionary generated for the built artificial intelligence application will be converted into a string and sent to the server, so that the server can obtain the user's Built artificial intelligence applications. At this time, the server will receive the character string sent by the user terminal. The dictionary converted to this character string describes and defines the corresponding mathematical primitives for the corresponding chunks through a piece of dictionary data.

It should be understood that the mathematical primitives are the operation units for implementing artificial intelligence applications in the construction of artificial intelligence applications for users. They can divide the operation units at different levels according to needs, and then configure the mathematical primitives and The corresponding chunk.

For example, according to artificial intelligence algorithms, neural network operations, dot product operations, and matrix multiplication operations are all configured as operating units. Some neural network operations include dot product operations. Therefore, you can see that both It is the division of operating units at different levels, but it does not affect the configuration of the corresponding mathematical primitives and chunks.

Therefore, the subdivided operation units can also be combined to form a new operation unit on the upper layer, that is, mathematical primitives and chunks. This is the process of adding chunks referred to in the foregoing exemplary embodiments.

In step 530, the character string is decoded to obtain the executable text of the artificial intelligence application.

Among them, the character string obtained by the dictionary conversion can be decoded to obtain the mathematical primitive identification and core parameters involved in the built artificial intelligence application. On this basis, the DAG data is continuously expressed by the mathematical language of graph theory. The structure reconstructs each mathematical primitive, and constructs the link relationship between this mathematical primitive and other mathematical primitives for this purpose, and finally forms the executable text corresponding to the built artificial intelligence application. The executable text contains all the executable sentences of the built artificial intelligence application, which is composed of the code information of all mathematical primitives.

In step 550, through execution of the executable text, the user is allowed to choose to build the artificial intelligence application to run the server.

Among them, after the executable text of the artificial intelligence application is obtained, the operation of the artificial intelligence application can be triggered. At this time, the functions deployed by the artificial intelligence application will be realized through the execution of the sentences in the executable text to satisfy the user’s Artificial intelligence needs.

For users, on the one hand, they quickly and freely build artificial intelligence applications through a graphical interface. On the other hand, because the built artificial intelligence applications exist and run on the server side, they can make the artificial Intelligent applications have obtained excellent hardware performance and powerful computing capabilities, enhancing the performance of artificial intelligence applications built.

FIG. 11 is a flowchart illustrating step 530 according to the embodiment corresponding to FIG. 10. In an exemplary embodiment, as shown in FIG. 11, the step 530 includes at least the following steps:

In step 531, the character string is decoded and converted into a dictionary, which contains the mathematical primitive identification and core parameters corresponding to the chunks configured for artificial intelligence applications.

In step 533, the code information for performing the corresponding mathematical operation is reconstructed through the mathematical primitive identifier and core parameters contained in the dictionary to obtain the executable text of the artificial intelligence application.

After decoding the converted character string to obtain a dictionary, a set of mathematical primitive identifiers and core parameters corresponding to each mathematical primitive can be obtained from the obtained dictionary.

As mentioned before, the set of dictionary data of the mathematical primitive identification and core parameters indicate the corresponding mathematical primitives and other mathematical primitives linked to this mathematical primitive, ie the input dimension is the math of the output dimension in the core parameters The primitive is the link to the current mathematical primitive.

Therefore, it is possible to reconstruct the mathematical primitives through the dictionary data contained in the dictionary. The reconstructed mathematical primitives will be expressed in the form of code information in terms of program execution, and the code information is obtained through reconstruction. This code information is the execution The programming language for mathematical operations corresponding to mathematical primitives. By analogy, all code information constitutes executable text for artificial intelligence applications.

FIG. 12 is a flowchart describing step 533 according to the embodiment corresponding to FIG. 11. In an exemplary embodiment, this step 533 is shown in FIG. 12 and includes at least the following steps.

In step 601, the data missing code of the corresponding mathematical operation is obtained according to the mathematical primitive identification in the dictionary.

Among them, the server is controlled by the artificial intelligence applications built by the user terminal to obtain a dictionary for the built artificial intelligence applications, the dictionary records the mathematical primitives corresponding to each mathematical primitive used by the built artificial intelligence applications Logo and core parameters.

In addition, the server also stores the information related to the code for the mathematical operations that need to be performed. The code-related information stored for the execution of each mathematical operation is the data-missing code with missing core parameters. The data missing codes are stored correspondingly with the index of the mathematical primitive as the index. The data missing code with missing core parameters can be adapted to fill different core parameters with the realization of different artificial intelligence applications, and realize different configurations of the mathematical operations performed, so as to fully adapt to the realization of artificial intelligence applications.

Under the effect of the data missing code stored on the server side, on the one hand, the server side provides users with artificial intelligence application algorithms supported by the data missing code, allowing users to build artificial intelligence applications simply and intuitively through the block, and the other The aspect will also obtain flexible and free algorithm implementation with the support of data missing code, which enhances the flexibility of artificial intelligence application construction.

In step 603, the core parameters corresponding to the mathematical primitive identification are filled into the obtained data missing code to obtain code information for performing the corresponding mathematical operation.

Among them, as mentioned above, the core parameters in the dictionary are corresponding to the mathematical primitive identification. Therefore, after the data missing code is obtained from the mathematical primitive identification, the corresponding core parameters will also be obtained from the mathematical primitive identification. The acquired core parameters are filled into the data-missing code to obtain complete code information for performing the corresponding mathematical operation.

The code information obtained by filling in the core parameters is a mathematical primitive description at the code level. Under the effect of missing data codes and code information, users can build artificial intelligence applications according to their own needs even without programming intelligence and programming skills.

In step 605, according to the input dimension and output dimension indicated in the core parameter corresponding to the mathematical primitive identification, the executable information of the artificial intelligence application is reconstructed from the output of the code information to the input through the mathematical language of graph theory.

Among them, as pointed out in the foregoing description, the core parameters are the corresponding mathematical operations performed, and also indicate the input and output dimensions for the implementation of the artificial intelligence algorithm through the code information. Therefore, it can be based on this and other Mathematical primitive docking.

It should be noted here that the mathematical language representation of graph theory is for mathematical primitives, and also for code information that performs corresponding mathematical operations. The mathematical language representation of graph theory is to use the DAG structure to quickly and accurately reconstruct the input, output, and mathematical operations corresponding to the mathematical primitives, and then on the basis of this, the DAG structure corresponding to the mathematical primitives is continuously carried out in the order of output to input. Until the completion of the reconstruction of all mathematical primitives corresponding to the dictionary.

After all the mathematical primitives corresponding to the dictionary are reconstructed to obtain the mathematical language representation of graph theory, the corresponding code information can be reconstructed according to the mathematical language representation of graph theory to obtain the executable text of the artificial intelligence application.

For example, Fig. 13 is a schematic diagram of a DAG structure corresponding to a simple mathematical primitive according to an exemplary embodiment. This simple mathematical primitive is "I+W=O", that is, "I" is used as the input, and "O" is used as the output, and the addition operation is performed on the input data, and "W" is the parameter obtained by the required training.

The following table shows the "I" and "O" data tables configured by the user on the block Add in the graphical interface. The block add indicates that the corresponding mathematical primitive will perform an addition operation on the input data, as shown in the following table. which is:

II	OO
11	77
22	88

33	99
44	1010
55	1111

Table 1

With this exemplary embodiment, the construction of artificial intelligence applications can be quickly realized even if complex artificial intelligence algorithms are involved by means of the DAG structure.

FIG. 14 is a flowchart illustrating step 550 according to the embodiment corresponding to FIG. 10. In an exemplary embodiment, as shown in FIG. 14, this step 550 includes at least the following steps.

In step 551a, in the training mode, sample data suitable for artificial intelligence applications is obtained, and the training mode is the mode selected by the user on the graphical interface.

In step 553a, an artificial intelligence application is run on the server through execution of executable text, and parameter training is performed through sample data during the operation of the artificial intelligence application.

This exemplary embodiment provides the training process of the artificial intelligence application for the artificial intelligence application built by the user, so that the artificial intelligence application built by the user completes the iterative training of the parameters through the set sample parameters, and then the built artificial intelligence application can be used normally .

The sample data suitable for artificial intelligence applications is configured by the user in an exemplary embodiment, for example, the user imports sample data through the toolbar set by the graphical interface, and can also be related to the data through the operation interface The set of blocks can be used to set the sample data for the artificial intelligence applications built, and the sample data can also be selected through the check box set in the toolbar.

In other words, the sample data is collected by the user, it can also be stored on the server and shared by other users, or it can be pre-configured by the server according to the needs of various artificial intelligence applications. It is not limited here.

After the construction of the artificial intelligence application is completed, all the blocks that implement the artificial intelligence application have been added and linked to the construction area, after which the training mode can be selected to iterate the parameters of the constructed artificial intelligence application training.

That is to say, for the artificial intelligence application that has been built, it can be placed in training mode or operation decision mode through user control, so as to control the operation of the artificial intelligence application.

For example, it can set a switch that is turned on in the training mode or decision mode on the graphical interface. The user only needs to manipulate the switch to select the corresponding mode.

In training mode, obtain sample data suitable for artificial intelligence applications according to the sample data configuration performed by the user, for example, obtain sample data recommended by the server, obtain sample data shared by other users, and import it in the graphical interface The required sample data, etc.

The application of artificial intelligence is based on machine learning, and solves specific problems for users through machine learning algorithms. For example, the machine learning algorithm may be a neural network algorithm. Therefore, the completed artificial intelligence application must involve the realization of training, and the parameters required for the normal operation of the artificial intelligence application can be obtained through iterative training.

Through this exemplary embodiment, the entrance of sample data acquisition is realized for the built artificial intelligence application. Even if the user is limited to various situations and cannot obtain a large amount of sample data, it can be obtained by means of the server, thereby ensuring that the built Based on the training and subsequent use of artificial intelligence applications, users can also adapt to their own needs to prepare sample data to ensure the accuracy of subsequent decisions of the artificial intelligence applications built.

In addition, the sample data submitted by the user to the server can be hidden according to their own choices, that is, stored as their own private data, but they can also be set to the public state for sharing on the server.

For the server, in addition to the pre-configured sample data and the sample data shared by the user, the sample data can also be obtained by crowdsourcing, which is not limited here.

FIG. 15 is a flowchart illustrating step 550 according to the embodiment corresponding to FIG. 10 in another exemplary embodiment. In an exemplary embodiment, as shown in FIG. 15, this step 550 includes:

In step 551b, data selected by the user for input is acquired, and the data includes target data processed by the artificial intelligence application built by the user.

In step 553b, by executing an executable file that completes parameter training, running the trained artificial intelligence application in the decision mode completes the user's selection of input data.

This exemplary embodiment realizes that the user uses the self-built artificial intelligence application to make data decisions, and the built artificial intelligence application truly solves the user's data processing needs.

The data that the user chooses to input is the data that the user needs to process with the help of the built artificial intelligence application. It should be understood that the user is building the artificial intelligence application for this data processing requirement.

Similar to the training mode, only the decision mode needs to be turned on under the user's control to process the user's selection of input data.

FIG. 16 is a flowchart illustrating step 551b according to the corresponding embodiment of FIG. 15. In an exemplary embodiment, this step 551b, as shown in FIG. 16, includes at least the following steps.

In step 601, the data uploaded by the user to the artificial intelligence application built by the user on the graphical interface is received.

Among them, after the artificial intelligence application is built and trained, the user will upload the data processed by the artificial intelligence application on the graphical interface. According to the different artificial intelligence applications, the data content and data types are different. For example, the uploaded request artificial intelligence The data processed by the application can be text data, video data or even audio data.

In step 603, the artificial intelligence application built by the user is determined according to the user identifier carried in the data.

Among them, the server cooperates with many user terminals to implement the construction of artificial intelligence applications on each user terminal. Therefore, different users have different artificial intelligence applications. Therefore, for the server, the user identification is used as an index to store the dictionary to record the artificial intelligence applications corresponding to different users.

The artificial intelligence application that has been built and trained is completed, and its dictionary is stored on the server for users to call at any time. In other words, the user can call the built and trained artificial intelligence application when he needs to access the server, and in the use of the artificial intelligence application, iteratively optimizes the parameters in the artificial intelligence application by means of decision-making results. To continuously improve the accuracy of decisions.

The user ID is used to uniquely identify the user. Therefore, the user can log in to the server through the user ID, and then call the built artificial intelligence application.

In step 605, the target data requested by the artificial intelligence application built by itself to be processed in the data is triggered to run the determined artificial intelligence application.

Wherein, through the execution of the foregoing steps, after the upload of the required processing data is completed, the artificial intelligence corresponding to the user identification is requested to process the uploaded data, in order to run the artificial intelligence application with the uploaded data as input.

Fig. 17 is a flowchart of a method for implementing operation in building an artificial intelligence application shown in another exemplary embodiment. In another exemplary embodiment, as shown in FIG. 17, the operation implementation method in the construction of artificial intelligence applications includes at least the following steps.

In step 710, the user's update of the newly added mathematical primitive is received to obtain digital primitive definition information. The mathematical primitive definition information includes the newly added mathematical primitive identifier and the newly added mathematical primitive under the newly added mathematical primitive identifier At least one mathematical base identifier and core parameters corresponding to the package.

In step 730, the mathematical primitive definition information is saved, and the graphical interface is initialized and configured with new blocks.

Among them, as the block addition in the graphical interface of the user terminal is described, that is, as described in the embodiment corresponding to FIG. 7, the server controlled by the user terminal will necessarily receive the mathematical primitive definition information sent by the user terminal. The user will save the mathematical primitive definition information for the user to call in the subsequent artificial intelligence application construction.

Of course, the newly added blocks of the user can also be shared with other users for use in the construction of artificial intelligence applications of other users, which is not limited here.

The user's artificial intelligence application needs to be short-video action recognition. Take this as an example and explain in combination with the above method. Narrate.

The user needs to realize the video action recognition through the artificial intelligence application of the present invention. Specifically, for a short video describing an action, the action can be recognized by an artificial intelligence application built by the user, and the action classification result can be output.

Here, as described in the corresponding implementation environment of FIG. 1, on the user terminal, the user can drag and drop the required blocks in the operation component selection bar of the graphical interface to drag the required blocks to the operation interface, And build a link relationship through the connection between the blocks until the construction of the artificial intelligence algorithm configuration is completed.

The neural network operation adopted in the constructed artificial intelligence algorithm configuration is the mathematical operation adopted by the user for the artificial intelligence application.

Fig. 18 is a schematic diagram of a graphical interface according to an exemplary embodiment. In an exemplary embodiment, as shown in FIG. 18, the operation component selection bar 810 is provided with a number of operation components, that is, blocks related to data and models; the operation interface 830 is a stay area where users need to drag and drop the blocks , Complete the data and model building by dragging and dropping the icon, that is, the component.

In the construction of data and models, fine-tuning of data and models will also be performed through the toolbar 850 on the right.

Through this process to complete the construction of artificial intelligence applications, and iterative training of parameters. The information related to the iterative training process, such as the number of iterations and the related iterative training results, will be displayed through the operation result display area 870, where the accuracy of the data and model training will also be displayed to ensure that the user can master artificial intelligence Real performance.

The artificial intelligence application built by the user that can realize the video action video can try different artificial intelligence algorithm configurations through continuous block dragging to find available artificial intelligence algorithm configurations.

For example, you can drag the corresponding block of the convolutional neural network operation from the operation component selection bar 810 to the operation interface to construct a first layer of convolutional neural network. FIG. 19 is a convolutional neural network according to an exemplary embodiment. Schematic diagram of the block corresponding to the operation on the operation interface. Block I represents the input, block Dense represents the operation of the convolutional neural network in the deep learning neural network, and block O represents the output. On this basis, then to the operation interface Drag two blocks corresponding to the operation of the convolutional neural network to construct a three-layer convolutional neural network to achieve the extraction of convolutional features in the built artificial intelligence application. FIG. 20 is shown according to the corresponding embodiment of FIG. 19 Schematic diagram of block distribution and links on the operation interface of the three-layer convolutional neural network.

Through the interaction between the user terminal and the server, users can continuously try to build artificial intelligence applications, and constantly verify the effect, simple operation and low cost.

Through the implementation of the present invention, it helps users quickly realize the artificial intelligence applications they need, and can call them at will when needed.

For a person with zero basic knowledge related to artificial intelligence, to realize the independent development of an artificial intelligence application, the process is roughly as follows: first, learn the code knowledge completely, have a certain code development ability, and then learn the manual Intelligent expertise, in the end, as a developer, according to the current specific needs, choose the appropriate programming language to carry out algorithm development in the compiler, the time period required for this process is at least 7 years.

It should be understood that code knowledge, deep learning algorithms related to machine learning, mathematical representation, and overall knowledge of artificial intelligence are difficult to popularize, and the code learning threshold is also insurmountable by the public.

In order to overcome many of the difficulties mentioned above, the implementation and development of the present invention first provides a user-friendly interface for the public, and encapsulates many technical details in the form of blocks and mathematical primitives to achieve standardized Artificial intelligence model building process, so as to really help the public to use artificial intelligence technology.

The artificial intelligence algorithm is different from the simple algorithm logic similar to that involved in children's programming. It is a complex natural expression and is difficult to implement with the help of graphical user programming.

In the present invention, code encapsulation is implemented through mathematical primitives to achieve the required mathematical operations, and graphical blocks are defined on the graphical interface to represent them.

Fig. 21 is a schematic diagram of interaction between a front end and a back end involved in the present invention according to an exemplary embodiment. In this exemplary embodiment, the user side implements a website graphical interface, that is to say, for the user, the artificial intelligence application needed to be implemented through a web application only needs to know the corresponding background Visit the address to run the web application through the browser.

Of course, it is not limited to this, and it can also be implemented by issuing terminal programs and many other methods.

As shown in Figure 21, through the user's manipulation on the website's graphical interface, the back-end server will convert the graphics into code according to the user-configured block, that is, to obtain the built artificial intelligence application, and rely on the server to provide Cloud computing capabilities return results to users, or artificial intelligence algorithm configuration diagrams. On the graphical interface of the website, the user constructs an AI (Artificial Intelligence) algorithm in the mathematical language of graph theory, that is, the mathematical primitive referred to above, and the algorithm information involved is stored in a dictionary implemented in JavaScript language. On the one hand, the user chooses to train the algorithm on the graphical interface of the website. After that, the dictionary can be submitted to the server in the background in the form of a JSON string. After receiving the JSON string, the server sends it to the output end When the input terminal is decoded into a runnable AI code language, the server will complete the training according to the AI code language and return the results to the website graphical interface and display it.

It should be noted that mathematical primitives correspond to mathematical operations, which can include basic mathematical operations (BMO, Basic, Mathematical Operations) such as addition, subtraction, multiplication, and division, as well as various complex mathematical operations, such as, The mathematical language expression of mathematical primitives in graph theory, such as neural network operations, is a DAG structure used to express mathematical operations, also known as MNG (Mathematical Network Graph). DAG (Directed Acyclic Graph) structure is used to describe application logic that represents expressions in the form of nodes.

22 shows a schematic diagram of the implementation of the overall solution of the present invention in an exemplary embodiment. In an exemplary embodiment, as shown in FIG. 22, for an artificial intelligence application independently developed by a user, it includes four major stages, namely: a construction stage, a stage for sending pictures, a stage for translating a dictionary to MNG, and a loopback Results stage.

The construction period is the stage for the user to build the required artificial intelligence application. At this stage, the user can build the artificial intelligence algorithm and the new MNG block through the block, whether it is the artificial intelligence algorithm or the new MNG. The construction of tiles, their corresponding mathematical primitive indexes and core parameters are marked and stored in the dictionary.

The phase of sending the diagram is carried out by the construction period. The send map stage is the stage where the dictionary is sent for the constructed artificial intelligence algorithm or new MNG block. The reason why it is called the send map stage refers to the artificial intelligence algorithm or MNG block corresponding to the dictionary. In a network topology diagram.

In the stage of sending pictures, the type conversion of the character string received by the server is realized to eliminate the character string and decode and convert it back to the dictionary.

From the stage of sending the picture, the translation dictionary is entered into the MNG stage. At this stage, the server will identify the MNG, that is, the mathematical primitive, to generate code through the DAG structure, obtain the executable AI code language, and train and Return the result.

The trained AI code language can be used to make decisions on the data. The decision-making phase for data is still running and returning results. In this process, after the server completes the construction of the artificial intelligence application method, it obtains the AI code language. At this time, it will recognize the model and check whether it has been trained. If it has been trained and contains the data to be run, use this The data runs AI code language and returns the result.

Optionally, before the decision is made on the data, it will also be confirmed to the user by showing the artificial intelligence algorithm configuration diagram to the user, and the previously trained AI code language can be run on top of this basis.

This artificial intelligence algorithm configuration diagram has been constructed by building blocks when users build artificial intelligence applications. The process of obtaining the AI code is the process of extracting important information from the artificial intelligence algorithm configuration diagram and parsing it into code, which is used to encapsulate, that is, encapsulate for each data primitive, and realize net-to-gate, that is, by Network structure to a single mathematical operation. So far, it can be understood that any mathematical operation or mathematical equation is a complex of BMO. BMOs are linked together to form a DAG structure, which is expressed as MNG. Each MNG has its own input and output, and the input and output between different MNGs can be linked to each other to form a complete logical network. Therefore, it can be used as an equivalent transition between block and code. In this way, the information corresponding to the block is accurately transferred to the code, and the conversion between the graphical interface and the text programming language is realized.

It should be understood that each mathematical primitive is realized at the beginning, and at the same time, they are also the basis for user control. The network structure built by users through blocks will be converted into a dictionary as before. Then convert the dictionary to DAG format and parse it. Then, the network structure becomes a mathematical primitive, which is then added to the server as a new mathematical primitive. The newly formed mathematical primitives can be used as other primitives to construct other networks. As this cycle repeats, more complex algorithms can be developed without increasing the difficulty of developing graphics-based programming languages.

The following is a device embodiment of the present invention, which is an embodiment of a building device for executing the above artificial intelligence application of the present invention. For details not disclosed in the device embodiment of the present invention, please refer to the embodiment of the method for building an artificial intelligence application of the present invention.

Fig. 23 is a block diagram of a device for building an artificial intelligence application according to an exemplary embodiment. In an exemplary embodiment, as shown in FIG. 23, the artificial intelligence application building device includes but is not limited to: an instruction receiving module 1010, a construction module 1030, a conversion module 1050, and a decoding initiation module 1070.

The instruction receiving module 1010 is configured to receive an instruction for selecting a block on the graphical interface, the block is a graphical representation of the corresponding mathematical primitive.

Constructing module 1030, configured to configure blocks for building artificial intelligence applications in the building area of the graphical interface through the selection instruction, and link the blocks to form an artificial intelligence algorithm for artificial intelligence applications built under user control structure.

The conversion module 1050 is configured to convert the included chunks into a dictionary according to the configuration of the artificial intelligence algorithm to obtain a dictionary composed of the mathematical primitive identification corresponding to the chunk and the core parameters, the core parameters corresponding to the group Block configuration.

The decoding initiation module 1070 is used to initiate decoding of the artificial intelligence application built by the user to the server through the dictionary, and trigger the artificial intelligence application to run on the server.

Fig. 24 is a block diagram of a device for implementing operation in building an artificial intelligence application according to an exemplary embodiment. In an exemplary embodiment, as shown in FIG. 24, the operation implementation device in the manual and intelligent application building includes, but is not limited to: a character receiving module 1110, a decoding module 1130, and an execution module 1150.

The character receiving module 1110 is used for building an artificial intelligence application selected by the user, and the server receives a character string corresponding to a dictionary, and the dictionary corresponding to the character string is used to describe a block configured for building the artificial intelligence application.

The decoding module 1130 is configured to decode the character string to obtain the executable text of the artificial intelligence application.

The execution module 1150 is configured to enable the artificial intelligence application selected by the user to run on the server through execution of the executable text.

Optionally, the present invention also provides an electronic device, which can be used in the implementation environment shown in FIG. 1 to perform all or part of the steps of any of the methods shown in FIGS. 3 to 17. The device includes: a processor; a memory for storing processor executable instructions;

Wherein, the processor is configured to perform the method mentioned above.

The specific manner in which the processor of the device in this embodiment performs operations has been described in detail in the foregoing embodiments, and will not be described in detail here.

In an exemplary embodiment, a storage medium is also provided. The storage medium is a computer-readable storage medium, for example, it may be a temporary and non-transitory computer-readable storage medium including instructions. The storage medium includes, for example, a memory 204 of instructions that can be executed by the processor 218 of the device 200 to complete the above method.

It should be understood that the present invention is not limited to the precise structure that has been described above and shown in the drawings, and various modifications and changes can be performed without departing from the scope thereof. The scope of the invention is only limited by the appended claims.

Claims

An artificial intelligence application building method, characterized in that the method includes:

Receiving a selection instruction for a block on the graphical interface, the block is a graphical representation of the corresponding mathematical primitive;

Configuring the building blocks of the artificial intelligence application in the building area of the graphical interface through the selection instruction, and linking the blocks to each other to form the artificial intelligence algorithm configuration of the artificial intelligence application built under user control;

Convert the included chunks into a dictionary according to the configuration of the artificial intelligence algorithm to obtain a dictionary composed of the mathematical primitive identifier corresponding to the chunk and the core parameters, the core parameters corresponding to the chunk configuration;

Initiating decoding of the artificial intelligence application built by the user to the server through the dictionary, triggering the artificial intelligence application to run on the server.
The method of claim 1, wherein the receiving a user selection instruction for a block on a graphical interface includes:

Among the blocks configured in the initial configuration of the graphical interface, the user's operation applied to the block receives a selection instruction for the block on the graphical interface until the blocks implementing the artificial intelligence application are all Select.
The method according to claim 1, wherein the configuration block for the artificial intelligence application is configured in the construction area of the graphical interface by the user selection instruction, and the blocks are linked to each other to form a user control The artificial intelligence algorithm configuration of artificial intelligence applications built below includes:

Configuring the block indicated by the selection instruction in the construction area of the graphical interface;

Obtain the corresponding core parameters for the blocks placed in the construction area;

With the configuration of more than two blocks in the building area, interconnecting the blocks to form an artificial intelligence algorithm configuration for artificial intelligence applications built under the control of the user.
The method according to claim 3, wherein the acquiring the corresponding core parameters of the blocks placed in the building area includes:

The block selected for the graphical interface building area is obtained according to the core parameter configuration of the block by the user to obtain the core parameter corresponding to the block.
The method according to claim 1, wherein the conversion of the included chunks into a dictionary according to the configuration of the artificial intelligence algorithm to obtain a dictionary composed of data primitive identifiers and core parameters corresponding to the chunks includes :

For the blocks included in the configuration of the artificial intelligence algorithm, obtain the mathematical primitive identification and core parameters corresponding to the blocks;

Using the mathematical primitive identifier as an index item and the core parameter as an index value, construct a dictionary of the artificial intelligence application.
The method according to claim 1, wherein the initiating decoding of the artificial intelligence application built by the user to the server through the dictionary to trigger the artificial intelligence application to run on the server includes:

Perform character string conversion on the mathematical primitive identification and index core parameters in the dictionary;

The artificial intelligence application built for the user transmits the character string to the server, and initiates decoding of the character string by the server through the transmission of the character string to obtain the executable text of the artificial intelligence application, to Run on the server.
An operation implementation method in the construction of artificial intelligence applications, characterized in that the method includes:

In the construction of the artificial intelligence application selected by the user, the server receives the character string corresponding to the dictionary, and the dictionary corresponding to the character string is used to describe the block configured for the construction of the artificial intelligence application;

Decoding the character string to obtain the executable text of the artificial intelligence application;

Through the execution of the executable text, the artificial intelligence application that the user chooses to build runs on the server.
The method according to claim 7, wherein the decoding of the character string to obtain executable text of the artificial intelligence application includes:

Decode the character string into a dictionary, the dictionary contains the mathematical primitive identification and core parameters corresponding to the blocks configured for artificial intelligence applications;

The executable text of the artificial intelligence application is obtained through the mathematical primitive identifier contained in the dictionary and the core parameter to reconstruct the code information for performing the corresponding mathematical operation.
The method according to claim 8, wherein the code information of the mathematical operation corresponding to the reconstruction and execution of the mathematical operation corresponding to the mathematical primitive identifier and the core parameters contained in the dictionary are obtained to obtain the executable text of the artificial intelligence application, include:

Obtaining the data missing code of the corresponding mathematical operation according to the mathematical primitive identification in the dictionary;

Filling the core parameters corresponding to the mathematical primitive identifier into the obtained data missing code to obtain code information for performing the corresponding mathematical operation;

According to the input dimension and the output dimension indicated in the core parameter corresponding to the mathematical primitive identification, the executable text of the artificial intelligence application is reconstructed from the output of the code information to the input through the mathematical language representation of graph theory.
The method according to claim 7, wherein the execution of the executable text to enable the artificial intelligence application selected by the user to run on the server includes:

In the training mode, obtaining sample data suitable for the artificial intelligence application, the training mode is a mode selected by a user on a graphical interface;

Run the artificial intelligence application on the server side through execution of the executable text, and perform parameter training through the sample data during the operation of the artificial intelligence application.
The method according to claim 17, wherein the execution of the executable text to enable the artificial intelligence application selected by the user to run on the server includes:

Obtain data selected by the user, including the target data processed by the artificial intelligence application built by the user;

By executing the executable file that completes the parameter training, running the trained artificial intelligence application in a decision mode completes the processing of user-selected input data.
The method according to claim 11, wherein the acquiring the data selected by the user includes:

Receiving data uploaded by the user to the artificial intelligence application built by himself on the graphical interface;

Determine the artificial intelligence application built by the user according to the user identifier carried in the data;

The target data requested by the artificial intelligence application built by itself to be processed in the data is triggered to run the determined artificial intelligence application.
An artificial intelligence application building device, which is characterized by including:

The instruction receiving module is used to receive a selection instruction of a block on the graphical interface, the block is a graphical representation of the corresponding mathematical primitive.

Construction module, configured to configure blocks for building artificial intelligence applications in the building area of the graphical interface through the user selection instruction, and link the blocks to form an artificial intelligence algorithm for artificial intelligence applications built under user control structure.

The conversion module is used to convert the included chunks to the dictionary according to the configuration of the artificial intelligence algorithm to obtain a dictionary composed of the mathematical primitive identification corresponding to the chunk and the core parameters, the core parameters corresponding to the chunks Configured.

The decoding initiation module is used to initiate decoding of the artificial intelligence application built by the user to the server through the dictionary, and trigger the artificial intelligence application to run on the server.
An operation realization device in the construction of manual and intelligent applications, which is characterized by comprising:

The character receiving module is used for building an artificial intelligence application selected by the user, and the server receives a character string corresponding to a dictionary, and the dictionary corresponding to the character string is used to describe a block configured for building the artificial intelligence application.

The decoding module is used for decoding the character string to obtain the executable text of the artificial intelligence application.

An execution module is configured to enable the artificial intelligence application selected by the user to run on the server through execution of the executable text.
A machine equipment, characterized in that it includes:

Processor; and

A memory, on which computer readable instructions are stored, which when executed by the processor implements the method according to any one of claims 1 to 10.