RSS
Логотип
Баннер в шапке 1
Баннер в шапке 2

EZDL (the platform for machine learning)

Product
Developers: Baidu
Date of the premiere of the system: September, 2018
Branches: Information technologies

2018: Start of the platform

At the beginning of September, 2018 the Baidu company developing of the same name the search system, largest in China, issued the platform for software development of EZDL. It allows to create and bring into commercial operation models of machine learning without knowledge of programming.

According to the press release of Baidu, the new service platform is addressed to the companies representing small and medium business which have no specialists in such projects as TensorFlow PyTorch and Microsoft Cognitive Toolkit.

Baidu started the platform for machine learning without knowledge of programming

Through the EZDL company can build models of machine learning, using the drag-and-drop interface (simple drag and drop of objects by means of a mouse). To train the system of deep learning constructed especially for accomplishment of specific business challenges, the company needs to do only four steps, the technical lead of the project of Baidu of EZD Yongkang Xie says.

File:Aquote1.png
Even if you were never engaged in programming, you can quickly construct models on this platform without any obstacles. EZDL can help the companies which lack experts in artificial intelligence and computing resources, quickly and effectively to hold a training of model of deep learning and to implement it with a small amount of data — noted Hsi.
File:Aquote2.png

In EZDL AI models with very small set of training data which often includes only tens or hundreds of examples, but not thousands and millions, as at Internet giant can be developed.

By the beginning of September, 2018 the Baidu platform is limited three types of models: pattern discrimination, detection of objects and classification of a sound. Clients began to use EZDL for check of shelves of supermarkets, monitoring of security, recognition of a voice in wild the nature, control of quality of production of keyboards, etc. It is noted that the accuracy of the models created by means of the platform varies from 88% to 95%.[1]

Notes