CN111476373B - An artificial intelligence data service system - Google Patents

Abstract

The invention relates to an artificial intelligence data service system, which comprises a front-end interaction platform and an artificial intelligence data platform; the front-end interactive platform is characterized by comprising: the data retrieval service module is used for quickly retrieving and positioning data required by a user by inputting keywords or dividing according to the data field; the data comprises a data set, an API interface and a data model; the data set service module is used for providing the online use and downloading functions of the data set, a user can locate a required data set through the retrieval service module, try to perform online data model training through data in the data set, and download original data or data generated by training to the local through a platform through application; the data interface service module is used for the user to apply for obtaining the key and then carry out the form of parameter filling and transmission to obtain the corresponding data on line; and the model/calculation force service module is used for the user to carry out online convenient model training.

Description

An artificial intelligence data service system

technical field

The invention belongs to the field of artificial intelligence, and in particular relates to an artificial intelligence data service system.

Background technique

In the process of studying machine learning, a large amount of data resources and efficient computing power are required for repeated model training. The collection of data sets takes a lot of time and also has certain difficulties. In addition, the lack of computing resources increases the cost of machine learning research in disguise. And it is not uncommon in the research process that the model training results are too different due to the uneven quality of the dataset. The emergence of the artificial intelligence data platform system solves key problems such as difficulty in finding data and lack of training resources.

At the same time, most algorithms or application models of machine learning or artificial intelligence have certain requirements for data annotation and data quality, and the availability of data cannot be reflected through a brief introduction or a few sample data. The system provides a series of methods such as data trial function, which provides a good starting point for algorithm verification or building a better model.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the problems of difficulty in collecting and poor applicability of data sets required for the development of most algorithms or application models in the existing artificial intelligence field.

To achieve the above purpose, the present invention provides an artificial intelligence data service system, including a front-end interactive platform and an artificial intelligence data platform; wherein, the front-end interactive platform includes:

The data retrieval service module is used for users to quickly retrieve and locate the data they need by entering keywords or according to the division of data fields; the data includes data sets, API interfaces, and data models;

The data set service module is used to provide the online use and download functions of the data set. Users can locate the required data set through the retrieval service module, conduct online data model training through the data trial in the data set, or apply for the original data set. The data or data generated by training are downloaded locally through the platform;

The data interface service module is used for users to apply for obtaining keys and then fill in and transfer parameters to obtain the corresponding data online;

The model/computing service module is used for users to conduct online and convenient model training.

Further, the front-end interactive platform also includes: a data sharing module, which is used by the user to upload the generated data sets, API interfaces, and data models to the artificial intelligence data platform, and the artificial intelligence data platform filters the data through corresponding filtering algorithms.

Further, the data sharing step of the data sharing module includes:

Step 1, the data provider uploads and publishes the data set to the artificial intelligence data platform data center;

Step 2, the data provider publishes and saves the data set information, and generates a page for display;

Step 3: The data acquirer queries/brows the data set information.

Further, on the basis of the third step, when the data acquirer needs to perform data verification, the following steps are included;

Step 4: Select a typical algorithm or application model;

Step 5, select the running software/hardware environment;

Step 6: Validate the data, observe the output of the algorithm or model, and verify the applicability of the data.

Further, on the basis of the third step, when the data acquirer does not need to perform data verification, the following steps are included;

Step 7: The data acquirer submits a download application;

Step 8: After the system background review, according to the method provided by the application, feedback to the data acquirer the access method of the relevant data;

Step 9: The data acquirer acquires data according to the download address provided by the system.

Further, on the basis of the third step, when the data acquirer does not need to perform data verification and performs a data customization request, the following steps are included;

Step 10: The data acquirer submits data customization requirements, explaining further data requirements;

In step ten, the system feeds back users and provides services according to the progress of data processing.

Further, the data interface service module is specifically used for the user to apply for a key through the API interface, and to set parameters, and to obtain the corresponding result by means of parameter transmission, and the return format of the user-definable data is XML or JSON.

The invention provides artificial intelligence, especially reinforcement learning researchers, with information of relevant data sets in the field, as well as common algorithms and application model libraries. At the same time, the system provides a certain range of data and algorithm trials, which is convenient for users to observe the performance of data sets or algorithms in specific scenarios, and to build artificial intelligence systems more easily.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

1 is a schematic structural diagram of an artificial intelligence data service system according to an embodiment of the present invention;

Fig. 2 is a schematic diagram of data sharing method steps;

Figure 3 shows the interface of the error code reference table and the sample code table.

Detailed ways

The artificial intelligence data service system provided by the embodiment of the present invention mainly gathers relevant data sets in the field of artificial intelligence, and categorizes and displays them according to different dimensions, which is convenient for users to quickly search; users upload or apply for relevant data sets, algorithms and applications. Model. Application directions include computer vision, natural language, etc.; storage media include images, videos, voices, charts, etc. The system gathers algorithms and models including machine learning, statistical analysis, data processing, etc. to form an artificial intelligence algorithm library and model library. The algorithm library includes commonly used algorithms and models such as BP neural network, ART1 neural network, moving average model, and unit root test.

System maintenance personnel adapt data and algorithms to provide trial operation services. The trial operation service provided by the system is to provide users with an online experience for a specific data set, algorithm or application model. The user selects the data source according to the prompts on the webpage and as shown in the wizard, and starts the operation after the relevant variables are set. After the operation is completed, the output results are displayed on the interface, including the target variable training table (target variable training results) and the target variable prediction table (target variable prediction results).

The artificial intelligence data service system provided by the embodiment of the present invention mainly constructs a front-end display website through JAVA language + Spring framework and front-end Vue + JavaScript + Html. The back-end trial operation platform uses Docker containers to encapsulate the relevant operating environment. The website accesses the programs in Docker through the Web through the task scheduling system according to the user's operation. The system is based on the B/S architecture and the concept of cloud, and the desired data resources can be obtained online only through a browser. Users can also quickly obtain the data they want through the API interface. The platform is also equipped with an efficient GPU computing cluster, and users can perform related model training work on the platform through data + models. Users can upload the excellent data sets generated during their training to the platform for data sharing. The platform continuously filters and regenerates data through efficient filtering algorithms, thus forming a recyclable data ecosystem.

FIG. 1 is a schematic structural diagram of an artificial intelligence data service system according to an embodiment of the present invention. As shown in Figure 1, the artificial intelligence data service system includes a front-end interactive platform and an artificial intelligence data platform; wherein, the front-end interactive platform includes: a data retrieval service module, a data set service module, a data interface service module and a data interface service module.

The data retrieval service module is used by users to quickly retrieve and locate the data they need by inputting keywords or according to the division of data fields; the data includes data sets, API interfaces, and data models;

Specifically, after the user logs into the artificial intelligence data service system, the user searches through the data retrieval service module, such as retrieving data in the API retrieval box on the home page, or entering the content the user wants to search to retrieve the corresponding API.

The data set service module is used to provide the online use and download function of the data set. Users can locate the required data set through the retrieval service module, conduct online data model training through the data trial in the data set, or apply for the original data. Or the data generated by training can be downloaded locally through the platform;

Specifically, the dataset page displays all existing datasets and data field categories, and users can apply for and use data through this page.

The data interface service module is used by the user to apply for obtaining the key and then fill in and transfer the parameters to obtain the corresponding data online;

Specifically, the data interface service module can be used for users to apply for keys through the API interface, and to set parameters. The corresponding results are obtained by passing parameters. The user can define the return format of the data as XML or JSON, and the page contains errors. Code reference table and example code table (as shown in Figure 3).

The model/computing service module is used for users to conduct online and convenient model training.

Specifically, on the model/computing service module page, the user can perform model training by selecting the data folder to configure the computing parameters, and the result will be displayed to the user in the CMD window of the page.

On the model running page, the user can first select a suitable data set, here we select the Googlenet data set; then the user can choose to set the parameters, the setting parameters here are to select the running host of the corresponding parameters, and the host of the first configuration is selected to run by default. After the parameter selection is completed, there are three buttons that can be clicked, namely: provide all results, training and validation. After clicking, the training set model is run, and the user can close it and view it through the personal center later. Click the Validation Run button, click to run the validation set model, the user can close it and view it through the personal center later.

Users can publish the model that is running relatively well, publish the model that has passed the training and verification, and provide an interface for other users to use. For example, whether the image recognition is a bird model, after publishing, the system generates an API URL for access (for example, http://159.226.226.111/model/bird/). The user authenticates by sending a post request with a file (compressed package, image, etc.) attached to the parameters, and returns the result. Users can upload a picture and click the test button, and the model can automatically determine whether the picture is a bird.

Further, the front-end interactive platform also includes: a data sharing module, which is used by users to upload data sets, API interfaces, and data models to the artificial intelligence data platform, and the artificial intelligence data platform filters and regenerates data through corresponding filtering algorithms. , to improve the applicability of the dataset to data acquirers. The data sharing method steps are shown in Figure 2:

Step 1, the data provider uploads and publishes the data set to the artificial intelligence data platform data center;

Step 2, the data provider publishes and saves the data set information, and generates a page for display;

Step 3: The data acquirer queries/brows the data set information.

On the basis of step 3, when the data acquirer needs to perform data verification, the following sub-step 1 is included:

Step 4: Select a typical algorithm or application model;

Step 5, select the running software/hardware environment;

Step 6: Validate the data, observe the output of the algorithm or model, and verify the applicability of the data.

On the basis of step 3, when the data acquirer does not need to perform data verification, the following sub-step 2 is included:

Step 7: The data acquirer submits a download application;

Step 8: After the system background review, according to the method provided by the application, feedback to the data acquirer the access method of the relevant data;

Step 9: The data acquirer acquires data according to the download address provided by the system.

On the basis of step 3, when the data acquirer does not need to perform data verification and requests for data customization, the following sub-step 3 is included:

Step 10: The data acquirer submits data customization requirements, explaining further data requirements;

In step ten, the system feeds back users and provides services according to the progress of data processing.

In the embodiment of the present invention, a large number of high-quality data resources are matched with an efficient GPU computing cluster, and a user can complete a complete model training task of machine learning through simple operations on the website. Users can not only test run, apply to download existing algorithms and data on the platform, but also upload self-developed algorithms and share their knowledge products with others to create a recyclable data ecosystem.

Obviously, many variations of the invention described herein are possible without departing from the true spirit and scope of the invention. Therefore, all changes obvious to those skilled in the art should be included within the scope covered by the present claims. The scope of the claimed invention is limited only by the appended claims.

Claims (3)

1. An artificial intelligence data service system, comprising a front-end interactive platform and an artificial intelligence data platform; it is characterized in that, the front-end interactive platform comprises: The data retrieval service module is used for users to quickly retrieve and locate the data they need by entering keywords or according to the division of data fields; the data includes data sets, API interfaces, and data models; The data set service module is used to provide the online use and application functions of the data set. Users can locate the required data set through the retrieval service module, conduct online data model training through the data trial in the data set, or apply for the original data set. Data or data generated by training are downloaded locally through the platform; the data set page displays all existing data sets and data field classifications, and users can apply for and use data through this page; The data interface service module is used for users to apply for a key and then fill in and pass parameters to obtain the corresponding data online; it is specifically used for users to apply for keys through the API interface, as well as parameter settings, which are passed through parameters. To obtain the corresponding results, the user can define the return format of the data as XML or JSON, and the page contains the error code reference table and the sample code table; Model/computing service module, which is used for users to conduct online and convenient model training; users can perform model training in the form of selecting data folders to configure calculation parameters, and the results will be displayed to users in the CMD window of the page; The front-end interactive platform also includes: a data sharing module, which is used by the user to upload the generated data sets, API interfaces, and data models to the artificial intelligence data platform, and the artificial intelligence data platform filters the data through corresponding filtering algorithms; The data sharing steps of the data sharing module include: Step 1, the data provider uploads and publishes the data set to the artificial intelligence data platform data center; Step 2, the data provider publishes and saves the data set information, and generates a page for display; Step 3, the data acquirer queries/brows the data set information; On the basis of the third step, when the data acquirer needs to perform data verification, the following steps are included; Step 4: Select a typical algorithm or application model; Step 5, select the running software/hardware environment; Step 6: Validate the data, observe the output of the algorithm or model, and verify the applicability of the data. 2. The system according to claim 1, wherein, on the basis of the step 3, when the data acquirer does not need to perform data verification, the following steps are included; Step 7: The data acquirer submits a download application; Step 8: After the system background review, according to the method provided by the application, feedback to the data acquirer the access method of the relevant data; Step 9: The data acquirer acquires data according to the download address provided by the system. 3. The system according to claim 1, wherein, on the basis of the step 3, when the data acquirer does not need to carry out data verification and carries out a data customization request, the following steps are included; Step 10: The data acquirer submits data customization requirements, explaining further data requirements; In step ten, the system feeds back users and provides services according to the progress of data processing.

Images