AWS announces AutoGluon, an open-source library for writing AI models

Amazon Web Services Inc. today launched a new open-source library to help developers write, with just a few lines of code, machine learning-based applications that use image, text or tabular data sets.

Building machine learning apps that rely on such data isn’t an easy task. For example, developers need to know how to tune the “hyperparameters” that represent the choices made when constructing an AI model. They also need to grapple with issues such as neural architecture search, which enables them to find the best architecture design for their machine learning models.

AutoGluon automates many of these complicated tasks and can create a new machine earning model with as little as three lines of code by automatically tuning choices within default ranges that are known to perform well for a given task. All the developer has to do is specify how quickly they want their model to be trained, and AutoGluon will come up with the strongest model in the given timeframe.

Amazon said AutoGluon can identify models for tasks including image and text classification, object detection and tabular prediction. It also features an application programming interface for more experienced developers to fiddle with so they can improve a model’s predictive performance.

“We developed AutoGluon to truly democratize machine learning, and make the power of deep learning available to all developers,” AWS applied scientist Jonas Mueller said in a statement. “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.”

Constellation Research Inc. analyst Holger Mueller told SiliconANGLE that many enterprises lack the developer talent they need to build new machine learning models quickly, and so there’s a big demand for tools that can simplify the process.

“Amazon’s AutoGluon is a key step in that direction, empowering developers to use advanced AI technologies on AWS’s cloud infrastructure,” Mueller said. “The ability to limit searches for the best fit models by CPU is especially valuable, as real word limitations to budgets and time apply to AI solutions as well. Now it will be all about seeing the adoption of AutoGluon in the developer base. Needless to say, the battle for AI leadership will not be won by developers, but by enabling the mildly tech savvy business user to take advantage of AI.”

AutoGluon is just the latest new offering from AWS as it looks to democratize machine learning. The company recently updated its SageMaker tool for continuously training and deploying machine learning models to the cloud and edge environments. Updates include SageMaker Studio, a model training and workflow management tool that collects all of the code and notebooks models use and keeps them in one place, and SageMaker Autopilot, which automates model creation by automatically choosing the best algorithm and tuning it for a specific task.

Image: AWS