Amazon SageMaker now supports incremental learning for its built-in visual recognition algorithms – Image Classification and Object Detection. With incremental learning, you can initialize your model with knowledge learned from prior training. This enables the model to preserve the knowledge gained previously and extend it by training the model on new data. It becomes useful when your training data comes in batches over time. With this capability, you do not need to retrain your model from scratch every time you get new data, saving you time and compute resources.

AWS Fargate is now available in US West (Northern California) and Asia Pacific (Seoul) Regions.

AWS Fargate is a compute engine for Amazon ECS that lets you run containers in production without deploying or managing servers. Fargate lets you focus on designing and building your applications instead of managing the infrastructure that runs them.

For a full list of AWS Regions where Fargate is available, please visit our Region table .

 

Amazon SageMaker now supports Apache MXNet 1.3 and TensorFlow 1.11 in its built-in containers for MXNet and TensorFlow respectively. This makes it easier to run MXNet and TensorFlow scripts, while taking advantage of the capabilities Amazon SageMaker offers, including a library of high-performance algorithms, managed and distributed training with automatic model tuning, one-click deployment, and managed hosting.

Amazon SageMaker now supports AWS Key Management Service (KMS) to encrypt data used for Batch Transform. Using AWS KMS, customers have the ability to control the encryption keys to protect the storage volumes used in their Batch Transform jobs. 

You can now clone existing hyperparameter tuning jobs to create new jobs through the Amazon SageMaker console. It is common to run multiple hyperparameter tuning jobs with the same parameters such as datasets, hyperparameter ranges, and compute resources. Setting up many jobs with the same details can be tedious and time-consuming. Cloned hyperparameter tuning jobs use the same configuration of the parent tuning job. This makes it easier and faster to create new hyperparameter tuning jobs.

AWS Key Management Service (KMS) now offers a standalone console that simplifies the process of managing encryption keys. The new console separates AWS managed keys from customer managed keys within your account and provides enhanced search and filtering capabilities to more easily find the keys you want to manage, even if you have hundreds or thousands of keys. 

Amazon Aurora with PostgreSQL compatibility now supports AWS Identity and Access Management (IAM) to manage database access. Database administrators can associate database users with IAM users and roles. This way, you can manage user access to all AWS resources from a single location, avoiding issues caused by permissions being out of sync on different AWS resources.

Amazon Elasticsearch Service Reserved Instances (RIs) offer significant discounts compared to standard On-Demand Instances. Starting today, you can use AWS Budgets to set custom utilization and coverage budgets based on your Amazon Elasticsearch RIs that alert you when you fall below the threshold you define.

Amazon Simple Queue Service (SQS) First-in, First-out (FIFO) queues are now available in Asia Pacific (Tokyo) and Asia Pacific (Sydney) regions, in addition to existing availability in US East (N. Virginia), US East (Ohio), US West (Oregon), and EU (Ireland) regions.

Resource tags allow you to add metadata and apply access policies to your Amazon Elastic Block Store (Amazon EBS) resources. Starting today, you can allow resource tags set on your EBS volumes to be automatically copied to any EBS snapshots that are created using Amazon Data Lifecycle Manager (DLM). This allows you to easily set snapshot metadata, such as access policies, to match the parent volume. You may enable this functionality on new or existing lifecycle policies. You may also choose to disable it at a future date.