AWS debuts AutoGluon to help devs get stuck into machine learning

AWS debuts AutoGluon to help devs get stuck into machine learning

AWS has quietly open sourced an open source tool kit it claims will “democratise” machine learning by removing much of the hand tooling data scientists currently fill their time with.

The cloud to everything giant’s AutoGluon slipped onto GitHub last year, under the Apache 2 license, but AWS is only now cranking up the publicity machine, with the promise that even neophytes will be able to “quickly prototype deep learning solutions for your data with few lines of code”.

According to the GitHub page, “AutoGluon automates machine learning tasks enabling you to easily achieve strong predictive performance in your applications. With just a few lines of code, you can train and deploy high-accuracy deep learning models on image, text, and tabular data.”

The tool set relies on Python 3.6 or 3.7, and right now it’s only available for Linux, with MacOS and Windows versions promised “soon”.

Either way, it will offer a “a focus on deep learning and real-world applications spanning image, text, or tabular data.” The AutoGluon website offers quick starts for tabular prediction, image classification, object detection and text classification.

AWS’s pitch is that the framework will take much of the grunt work out of developing and deploying deep learning models. 

In a blog announcing AutoGluon, it said that the likes of Theano had made calculating gradients simpler, and Keras had removed “much of the boilerplate code that was necessary in the existing libraries at the time.”

However, it continued, “even with these advancements, deep learning experts and developers today must still grapple with many cumbersome issues, including hyperparameter tuning, data pre-processing, neural architecture search, and decisions related to leveraging transfer learning.”

Much of this can be automated with AutoGluon, AWS claims, meaning devs “can produce a high-performance neural network model with as few as three lines of code.”

“There’s no need for developers to manually experiment with the hundreds of individual choices that must be made while designing a deep learning model,” it continues. “Rather, they can simply specify when they would like to have their trained model ready. In response, AutoGluon leverages the available compute resources to find the strongest model within its allotted run-time.”

The toolkit is being driven by AWS applied scientist Jonas Muller, who added “Due to the inherently opaque nature of deep learning, many of the choices made by deep learning experts are based on ad hoc intuition, rather than a rigorous scientific understanding of how individual choices affect desired outcomes. AutoGluon solves this problem as all choices are automatically tuned within default ranges that are known to perform well for the particular task and model.”

In November, AWS revved its Sagemaker machine learning platform, adding Sagemaker Worklows and additional algorithms and frameworks. This was followed by the unveiling of a Quantum Computing platform, Braket in December, which promises to allow developers to design quantum algorithms and run them on some of the incredibly esoteric hardware known to man.