Gluon provides prebuilt neural network components, as well as what AWS and Microsoft describe as a concise, user-friendly application programming interface (API), to make deep learning projects easier for developers unfamiliar with the technology.
To create a neural or deep learning network, developers need data, a neural network model and an algorithm that trains the model to recognize patterns in that data. Given the complexity of these algorithms and the large data sets often involved, this process can be complicated and time-consuming -- especially for developers new to deep learning technology.
Benefits of Gluon
Gluon attempts to reduce the time and complexity of this training process by tightly integrating the neural network model with the training algorithm. This structure can also help developers create and configure more sophisticated deep learning models, as well as streamline the debugging and updating process for their neural network.
Gluon aims to provide these benefits without a negative impact on performance, which AWS and Microsoft say differentiates it from other deep learning frameworks that typically struggle to provide a developer-friendly interface without slowing down training models. Gluon's support of programming loops and batch processes also help it execute jobs efficiently.
At launch, AWS provided 50 preconfigured examples for using Gluon via its Deep Learning AMI to help developers get started.
A developer can use Gluon via Apache MXNet, an open source deep learning framework. The cloud providers plan to add support for the Microsoft Cognitive Toolkit, as well as other frameworks in the future.