Amazon Web Services (AWS), Microsoft, and their long-running cloud rivalry were put on hold this week to announce a joint project: an open-source deep learning library called “Gluon.”
Gluon is an API that allows developers to use the Python programming language, along with some pre-included deeplearning templates, to create models that run neural networks faster and with less code complexity. AWS and Microsoft announced the project Tuesday in separate announcements. Gluon is an open-source deep learning interface that allows developers to quickly and easily build machine learning models without compromising their training performance.
The process of building a neural net can take up to a week, depending on how well network models are trained to parse large amounts of data. It’s time-consuming and labor-intensive. Developers often have to write long lines complex code that is difficult to modify, debug, and reuse.
The requirements for building a neural network can be difficult for even experienced developers. These barriers were removed by Microsoft and AWS.
“At AWS we’ve been exploring some ideas in MXNet around flexible, more accessible ways to create and train neural networks. Microsoft is also a contributor of the open-source MXNet project and was interested in these same ideas. We started talking about this and realized that we shared a similar vision: to use these methods to reduce the complexity in machine learning, making it more accessible to developers,” Matt Wood, AWS general manger of AI, wrote in a blog post.
According to the companies, Gluon is different from other deep-learning tools because of four attributes. Wood says that the Gluon interface allows developers to use “simple and clear” code to build networks. It allows them to create “dynamic” networks that are easier and more flexible to adapt and debug. Third, developers can create “more sophisticated algorithms” and models. These benefits are available without slowing down the training process.
AWS and Microsoft have collaborated on AI before with Gluon. In August, the companies agreed to let their respective voice-responsive digital assistants Alexa and Cortana interact in an effort to improve productivity for customers. Microsoft stated that the Cortana/Alexa integration was a result of Microsoft and Amazon’s belief that technology and people can work together.
This is the same philosophy that appears to have inspired the Gluon project. “We believe it’s important for the industry and the wider community to work together and pool their resources to create technology that benefits all. In an announcement, Eric Boyd, corporate vice-president of AI and Research at Microsoft, stated that Microsoft has partnered with AWS to create the Gluon Interface and allow developers freedom of choice.
Gluon currently supports Apache MXNet. According to the companies, Gluon will support the Microsoft Cognitive Toolkit and other frameworks in a future release.