AWS Evolves SageMaker Machine Learning Platform

Not to be left out of the exploding artificial intelligence/machine learning market, Amazon Web Services (AWS) has been busy expanding its SageMaker platform for creating and working with ML models.

Leveraging the scaling functionality of the AWS cloud, Amazon SageMaker is described by the company as an end-to-end service to help data scientists, developers and machine learning experts quickly build, train, and host machine learning models for production applications.

Demonstrating the importance that AWS places on the ML space, the company recently beefed up the SageMaker platform in a number of ways, including:

  • Working with AWS Glue. The company on Friday announced better integration of SageMaker with AWS Glue, the AWS cloud's fully managed extract, transform, and load (ETL) service to help customers prepare and load data for analytics.

    "You can now create an Amazon SageMaker notebook from the AWS Glue Console and connect it to an AWS Glue development endpoint," AWS said. These endpoints are serverless Apache Spark environments that customers can use to interactively develop, debug and test AWS Glue ETL scripts.

    "With this integration, you can now use Amazon SageMaker’s fully managed notebooks instead of provisioning and managing your own notebook servers, making it easier and faster to start developing your AWS Glue ETL scripts," AWS said.

  • Image classification update. AWS last week announced an enhancement to SageMaker's built-in image classification algorithm, targeting one of the most common use cases for ML modeling. Specifically, the new functionality involves support for multi-label inputs and mixed-precision mode for faster training.

    "The Image Classification algorithm can now take an image as an input and add multiple labels to that image," AWS said. "This is helpful because most images contain multiple elements and the ability to add multiple labels can produce better and more usable results. As an example, this can be used to detect and classify different elements during medical image processing for better detection and treatment."

    Additionally, faster model training and lower costs are said to be provided with the mixed-precision mode, as accuracy is preserved even while less memory is consumed during modeling.

  • AWS GovCloud availability. The company recently announced that SageMaker is available in the GovCloud (US) Region. AWS GovCloud is specifically designed to host sensitive data, regulated workloads and support stringent U.S. government requirements for security and compliance. Along with U.S. government customers, it also targets contractors, educational institutions, and other U.S. customers who need to run sensitive workloads in the cloud.

Earlier we reported on SageMaker's debut at the company's re:Invent conference, and noted a refresh of SageMaker algorithms and frameworks in July.

Stay tuned for more AWS updates to SageMaker as the cloud service seeks to stay apace in the burgeoning AI/ML arena.

About the Author

David Ramel is an editor and writer for Converge360.


Subscribe on YouTube