2 min Analytics

AWS announces cloud machine learning updates

AWS announces cloud machine learning updates

Amazon Web Services wants to make machine learning more accessible for developers, so the company is launching a number of updates for machine learning in the cloud. This should make it easier to add artificial intelligence (AI) predictions to applications and services.

To make machine learning more accessible to more developers, several updates have been announced. This makes it possible to run machine learning predictions on unstructured or relational data in Amazon S3, writes Silicon Angle.

In addition, the predictions can run on data in Amazon Aurora. Aurora is a cloud hosted relational database service that supports MySQL and PostgreSQL.

The two new options allow users to train machine learning models in SQL using Aurora or AWS Athena. AWS Athena is an interactive query service to analyse data in S3.

Visualisation with AI

It will also be possible to use AWS QuickSight for machine learning. QuickSight is a data visualisation tool to create and publish dashboards that highlight AI insights. The new features also allow QuickSight to visualise and report model predictions from machine learning services such as AWS SageMaker.

All these new features together reduce the need to write and maintain your own code. Where developers used to have to copy and transform data from stores into a supported format, this is no longer necessary.

AWS has ensured that calls no longer need to be made from an application, which makes it easier to add predictions to applications without having to set up integrations for them. Also, no other tools need to be learned for this.

Rekognition Custom Labels

Amazon has issued several updates over the past few days. In the field of AI, an entirely new tool appeared yesterday: Amazon Rekognition Custom Labels. This tool makes it possible to apply machine learning to AI models, to let them recognise certain objects, even when only limited information is available.

The tool ensures that large datasets no longer need to be used to train models, but that models can also be trained with small datasets for very specific use cases.