AWS Details SageMaker Improvements
Amazon Web Services (AWS) launched several enhancements to SageMaker, its machine learning platform, at this week's re:Invent conference.
The company announced the general availability of SageMaker Ground Truth, a service that promises to reduce the manual labor and costs associated with labeling data in preparation for machine learning training.
SageMaker Ground Truth lets users tap "human annotators" (whether they're from a third-party provider, the Amazon Mechanical Turk crowdsource service or their own workforce) to do part of the labeling tasks. The service then uses that human-labeled data as a guide to do the rest of the work, greatly reducing the time it takes to finish the dataset and potentially slashing costs by up to 70 percent, according to AWS' announcement.
"Amazon SageMaker Ground Truth can optionally use active learning to automate the labeling of your input data. Active learning is a machine learning technique that identifies data that needs to be labeled by humans and data that can be labeled by machine," explained Julien Simon, an AI and machine learning evangelist for AWS, in a blog post.
Also generally available is a "reinforcement learning" service called SageMaker RL. Reinforcement learning is a "reward"-based method for training a machine learning model. Simon explained the concept this way in another blog post:
Here, a computer program (aka an agent) interacts with its environment: most of the time, this takes place in a simulator. The agent receives a positive or negative reward for actions that it takes: rewards are computed by a user-defined function which outputs a numeric representation of the actions that should be incentivized. By trying to maximize positive rewards, the agent learns an optimal strategy for decision making.
SageMaker RL is a managed service that lets developers integrate reinforcement learning into their processes. It "allows any developer to build, train, and deploy with reinforcement learning through managed reinforcement learning algorithms, support for multiple frameworks (including Intel Coach and Ray RL), multiple simulation environments (including SimuLink and MatLab), and integration with AWS RoboMaker, AWS's new robotics service, which provides a simulation platform that integrates well with SageMaker RL," AWS said in its announcement.
A third new SageMaker capability, SageMaker Neo, is also generally available as of this week. AWS touts SageMaker Neo as a "deep learning model compiler [that] lets customers train models once, and run them anywhere with up to 2X improvement in performance."
SageMaker Neo supports a variety of hardware platforms, including those from Arm, Intel and Nvidia, as well as top machine learning frameworks. AWS also plans to open source SageMaker Neo "soon" under the Apache license.
