Table of contents

What's new and changed in Watson Machine Learning

Watson Machine Learning updates can include new features, bug fixes, and security updates. Updates are listed in reverse chronological order so that the latest release is at the top.

You can see a list of the new features for the platform and all of the services at What's new in IBM Cloud Pak for Data?

Initial release of Watson Machine Learning Version 4.0.0

A new version of Watson Machine Learning was released as part of Cloud Pak for Data Version 4.0.0.

Assembly version: 4.0.0

This release includes the following changes:

New features
Version 4.0.0 of the Watson™ Machine Learning service includes the following features and updates:
Use groups to manage deployment space collaborators
You can now add a user group as a collaborator in an deployment space. All users in the group have the role that you assign to the group. For details, see Collaborator permissions for spaces.
New permission required to create deployment spaces
To create a deployment space, users must have the Manage deployment spaces or Create deployment spaces permission. Restricting who can create deployment spaces gives you more control over resource on the cluster.

By default, these permissions are only associated with the Administrator role. For details, see Predefined roles and permissions

If you are upgrading from Cloud Pak for Data Version 3.5, your existing users will not have permission to create deployment spaces unless they have the Administrator role.

To give users one of these permissions, you can edit an existing role or create a new role. For details, see Managing roles.

New permissions for managing deployment spaces
Cloud Pak for Data includes new permissions that give you more control over deployment spaces:
Manage deployment spaces permission
Users with the Manage deployment spaces permission can:
  • Create deployment spaces
  • View list of all deployment spaces
  • Join any deployment space as an Admin
  • View deployment activity across all spaces
Monitor deployment activity permission
Users with the Monitor deployment activity permission can:
  • View list of all deployment spaces
  • View deployment activity across all spaces

For details, see Collaboration permissions.

Use a connection asset to secure credentials
Previously, when you accessed connected data as an input to a deployment, you had to enter authentication credentials with every connection attempt. Now, you can use the connection_asset type to refer to the data source connection by its ID. This method improves security by eliminating the need to enter authentication details multiple times. For details, see Batch deployment details.
Support for additional frameworks
Watson Machine Learning includes support for an expanded set of popular frameworks and software specifications for building and deploying machine learning models.
AutoAI training enhancements
Now you can train AutoAI experiments with:
An expanded list of data sources
AutoAI experiments now support an expanded set of data sources.
Joined data sets
Tech preview You can optionally combine multiple data sources that share a common column (key) and use that data to train AutoAI experiments. For details, see Building an AutoAI experiment with joined data.
AutoAI time series experiments
Tech preview Use AutoAI time series experiments to predict future activity based on history, sequential data. For details, see Creating a time series experiment.
Save AutoAI experiment code to a notebook
You can save all of the AutoAI experiment code to a notebook, where you can review all of the transformations and feature engineering that was used to create the model pipelines. For details, see, Saving an AutoAI generated notebook.
Manage the lifecycle of models with the cpdctl CLI
You can use the cpdctl command-line interface (CLI) to manage lifecycle of a model in Cloud Pak for Data. The cpdctl CLI helps you automate the end-to-end flow, from creating projects, to training models, to creating deployment jobs. For details, see Managing AI lifecycle with cpdctl.
Support for additional data sources for SPSS models
Watson Machine Learning supports additional input data sources for deploying an SPSS model. For details, see Batch deployment details.
Centralized job management
Simplify your DevOps process by promoting notebooks to deployment spaces, which give you a centralized place to manage jobs. For details, see Deployment spaces.
Federated Learning
Tech preview Federated Learning provides new ways for you to tune your experiments, including support for:
  • A party threshold metric (quorum)
  • Terminating an experiment when experiment accuracy thresholds are not met

For details, see Federated learning.

New authentication method
Watson Machine Learning supports an additional authentication method for Cloud Pak for Data environments that use the IBM Cloud Pak® foundational services Identity and Access Management Service (IAM Service) to authenticate. For details, see Authentication.