News

New AWS SageMaker Autopilot Automates Machine Learning Modeling

• By Becky Nagel
• 12/03/2019

On Tuesday Amazon announced SageMaker Autopilot, a new tool for its SageMaker machine learning platform that the company says will automate machine learning tasks like data preprocessing, training parameters and classification, including correcting for training issues like overfitting.

SageMaker is the company's managed service for machine learning. Within SageMaker, companies can work within the SageMaker Studio IDE environment to "stitch together" various tools, including SageMaker Neo, a training tool; SageMaker Augmented AI, for "human review" of model predictions; SageMaker Model Tuning for automated optimization; and, now, SageMaker Autopilot, for automating the process of building and training machine learning models.

As for how it works, Amazon shared the following:

SageMaker Autopilot first inspects your data set, and runs a number of candidates to figure out the optimal combination of data preprocessing steps, machine learning algorithms and hyperparameters. Then, it uses this combination to train an Inference Pipeline, which you can easily deploy either on a real-time endpoint or for batch processing. As usual with Amazon SageMaker, all of this takes place on fully-managed infrastructure.

Last but not least, SageMaker Autopilot also generate Python code showing you exactly how data was preprocessed: not only can you understand what SageMaker Autopilot did, you can also reuse that code for further manual tuning if you're so inclined.

Amazon is emphasizing that SageMaker Autopilot allows inspection of what's happening underneath, unlike with a "black box" set up of other tools.

A detailed tutorial on how to get started with the Autopilot tool can be found further down the page here.

SageMaker Autopilot is now live and available via an Amazon SageMaker subscription, which can be found here.