The name of the base system (platform): | Microsoft Azure |
Developers: | Microsoft |
Date of the premiere of the system: | May 10, 2017 |
Technology: | Cloud Computing |
2017: Announcement
On May 10, 2017 Microsoft provided service for training of neural networks in cloud infrastructure of the company. The new solution was received by the name Azure Batch AI Training.
Azure Batch AI Training will become a part of the Azure platform and will allow third-party developers to train own neural network models, using the most popular frameworks for deep training of TensorFlow from Google, Cognitive Toolkit from Microsoft, Caffe and other libraries. Developers will be able to execute workloads on several central and graphic processors and also on programmable logic arrays of FPGA.
All calculations will be carried out in Azure cloud that will help to cut down expenses of developers and to simplify their work. If desired users will be able independently to select type of devices for training of models.
Developers and specialists in the field of data will be able to configure the environment for training of the neural network models that as notes the TechCrunch edition, should become a decisive factor for use of this service after public start. By May, 2017 Azure Batch AI Training is in a test stage a certain circle of developers.
More detailed information on Azure Batch AI Training remains not opened. In particular, terms of start of the final version of a product and cost of the solution are unknown.
At the Build 2017 conference within which the tool kit for training of neural networks Azure Batch AI Training, Microsoft was announced showed how developers can independently train artificial intelligence for work with photos. On the example of flora it was shown how it is possible to distinguish different plants on their appearance.[1]