Proving the Business Case for the Internet of Things

Amazon unveils machine learning capabilities at Re:Invent

Steve Rogerson
December 6, 2018
At its Re:Invent conference in Seattle last month, Amazon Web Services announced 13 machine learning capabilities and services, across all layers in the machine learning stack, to help put the technology in the hands of even more developers. It also announced four services and capabilities that make it easier to ingest data from edge devices and build IoT applications.
AWS introduced SageMaker features making it easier for developers to build, train and deploy machine learning models, including low cost, automatic data labelling and reinforcement learning (RL). The company revealed services, framework enhancements and a custom chip to speed up machine learning training and inference, while reducing cost.
It announced artificial intelligence (AI) services that can extract text from virtually any document, read medical information, and provide customised personalisation, recommendations and forecasts using the same technology used by And AWS will help developers get rolling with machine learning with AWS DeepRacer, an 1/18th scale autonomous model race car for developers, driven by RL.
“We want to help all our customers embrace machine learning, no matter their size, budget, experience or skill level,” said Swami Sivasubramanian, vice president of Amazon Machine Learning. “Today’s announcements remove significant barriers to the successful adoption of machine learning, by reducing the cost of machine learning training and inference, introducing new SageMaker capabilities that make it easier for developers to build, train and deploy machine learning models in the cloud and at the edge, and delivering new AI services based on our years of experience at Amazon.”
Most machine learning models are trained by an algorithm that finds patterns in large amounts of data. The model can then make predictions on new data in a process called inference. Developers use machine learning frameworks to define these algorithms, train models and infer predictions. Frameworks such as TensorFlow, Apache MXNet and PyTorch allow developers to design and train sophisticated models, often using multiple GPUs to reduce training times. Most developers use more than one of these frameworks in their day-to-day work. AWS announced improvements for developers building with all these popular frameworks, by improving performance and reducing cost for both training and inference.
SageMaker is a fully managed service that removes the heavy lifting and guesswork from each step of the machine learning process. It makes it easier for developers to build, train, tune and deploy machine learning models.
The journey to build machine learning models requires developers to prepare their datasets for training their ML models. Before developers can select their algorithms, build their models and deploy them to make predictions, human annotators manually review thousands of examples and add the labels required to train machine learning models. This process is time consuming and expensive.
SageMaker Ground Truth makes it much easier for developers to label their data using human annotators through Mechanical Turk, third party vendors or their own employees. It learns from these annotations in real time and can automatically apply labels to much of the remaining dataset, reducing the need for human review. This creates highly accurate training data sets, saves time and complexity, and reduces costs by up to up to 70 per cent when compared with human annotation.
Machine learning is moving quickly, with new models and algorithms from academia and industry appearing virtually every week. SageMaker includes some of the most popular models and algorithms built-in, but to make sure developers continue to have access to the broadest set of capabilities, the AWS Marketplace for Machine Learning includes over 150 algorithms and models with more coming every day. These can be deployed directly to SageMaker. Developers can start using these immediately from SageMaker.
At the conference,, an open source AI company, launched H2O Driverless AI, an automatic machine learning platform on the AWS Marketplace for Machine Learning.
In machine learning circles, there is a lot of buzz about reinforcement learning because it’s an exciting technology with a ton of potential. RL trains models, without large amounts of training data, and it’s broadly useful when the reward function of a desired outcome is known but the path to achieving it is not and requires a lot of iteration to discover. Healthcare treatments, optimising manufacturing supply chains and solving gaming problems are a few of the areas that RL can help address. However, RL has a steep learning curve and many moving parts, which effectively puts it out of the reach of all but the most well-funded and technical organisations.
SageMaker RL, said to be the cloud’s first managed RL service, allows any developer to build, train and deploy with RL through managed 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 robotics service, which provides a simulation platform that integrates well with SageMaker RL.
SageMaker Neo, a deep learning model compiler, lets users train models once and run them anywhere with up to double an improvement in performance. Applications running on connected devices at the edge are particularly sensitive to performance of machine learning models. They require low latency decisions, and are often deployed across a broad number of different hardware platforms. SageMaker Neo compiles models for specific hardware platforms, optimising their performance automatically, allowing them to run at up to twice the performance, without any loss in accuracy. As a result, developers no longer need to spend time hand tuning their trained models for each and every hardware platform, saving time and expense. Neo supports hardware platforms from Nvidia, Intel, Xilinx, Cadence and Arm, and popular frameworks such as TensorFlow, Apache MXNet and PyTorch. AWS will also make Neo available as an open source project.
AWS also announced four services and capabilities that make it easier to ingest data from edge devices and build rich IoT applications.
AWS IoT SiteWise is a managed service that collects, structures and searches IoT data from industrial facility devices and uses them to analyse equipment and process performance data. IoT Events is a managed IoT service that makes it easy to detect and respond to changes indicated by IoT sensors and applications, such as malfunctioning equipment or a stuck conveyor belt, and automatically trigger actions or alerts.
IoT Things Graph is a service that makes it easy to build IoT applications with little or no code by connecting different devices and cloud services, such as linking humidity sensors to sprinklers to weather data services to create an agricultural application, through a visual drag-and-drop interface. And IoT Greengrass Connectors gives developers the ability to connect third-party applications such as ServiceNow for service management, on-premises software such as Splunk for log analytics, and AWS services such as Amazon Kinesis for data ingest via common cloud APIs. With this ability, developers can easily add more features such as location-based services, replenishment, industrial data processing, alarm and messaging, repair and maintenance, and logistics without writing code.
"Customers tell us they want to spend less time on the undifferentiated heavy lifting of getting different devices and services to work together and more time innovating on full-featured IoT applications,” said Dirk Didascalou, vice president for IoT at AWS. “We are giving customers tools that remove the cost and complexity of building applications at the edge with rich data sources to drive better business decision-making. This frees them up to spend time innovating in their core business, instead of writing code to connect devices and applications and to ingest actionable sensor data.”