Categories

Versions

What's New in RapidMiner Studio 10.1.1?

Released: January 31, 2023

Note: Version 10.1.0 has not been officially released and was skipped in favor of the 10.1.1 version.

The following sections describe what's new in RapidMiner Studio 10.1.1

Features

  • Enabled usage of RapidMiner Studio with Altair licenses
  • Introduced new operators which are powered by the new Belt data core. Existing processes will continue to use the previous operator versions, so existing processes will continue to work as before.
    • Generate Attributes
      • The expression parser can now access other rows via index, allowing for much more powerful expressions (e.g. Fibonacci, aggregations, etc.)
      • Added new lead/lag/cell_value/row_number functions
      • Added new time functions for time arithmetic
      • Improved date-time functions for a more consistent user experience across different time zones and locales.
    • Select Attributes
      • The new column types available in the new data core are now available here. This is a feature not yet used, however it will be in the future.
      • Greatly improved the user interface for the selection to make it much more user-friendly
      • Removed the very rarely used filter types 'block type' and 'numeric value filter'
    • Set Role
      • Roles can now be assigned more than once, e.g. selecting multiple label columns. This is a feature not yet used, however it will be in the future.
      • The new data core does not allow dynamic roles anymore, so only predefined roles are now available. The 'metadata' role can be used to mark special columns that should be ignored during operator calculations.
  • Advanced columns are now allowed outside operators
    • The new advanced column types text, text-list, text-set and real-array can now be used in Belt Data Tables between operators
    • They can be filtered out by type with Select Attributes to use the data table in operators that still operate on Example Sets
    • For now, this feature will not be visible to you unless you install future extensions that make use of these new column types. This is just the foundation so new extensions can start making use of these new capabilities.
  • Enabled future-proof JSON serialization for all IOObjects

Enhancements

  • Removed parameter create view from all preprocessing operators and Apply Model. This parameter was both virtually unused and broken - and is not being supported by our new internal data structure we introduced a while ago.
  • Improved automatic storage of custom result visualizations in cases where you have more than one file with the same name (but of different types) in the same repository folder
  • Activated Json serialization for all IOObjects (besides ExampleSets, Belt IOTables and IOObjectCollections, which all already have a new, special file format)
    • Old .ioo files can still be read
    • Non-converted IOObjects are still stored as .ioo
    • Works in local repositories and projects, but not in legacy repositories
  • Added admin settings with key rapidminer.disallow.decryption.storage. This impacts a user's interaction with connections
    • When set to true, any and all interactions with connections that need to decrypt values require a login (edit, create, move/copy, get metadata in a process)
    • If a project with disabled decryption storage is deleted from AI Hub, it cannot be recreated
  • Added support for OAuth for Salesforce connections
  • Support to manually configure the AI Hub frontend URL when connecting to a project (optional)
  • Added Setting to de-/activate automatic detection of remote changes for a project
  • Added Setting to de-/activate automatic detection of local changes for a project, to allow reducing filesystem access

Bugfixes

  • Fixed UI issue when opening an operator chain (e.g. Subprocess operator) with a breakpoint
  • Fixed issues with the data import wizard not closing on completion
  • Studio is now respecting the max memory setting again when started via the .exe on Windows
  • Fixed date-time formatting when starting via the .exe on Windows
  • Fixed some UI scaling issues when starting via the .exe on Windows
  • Error dialog improvements when using AI Hub
  • Fixed some json serialization that was behaving incorrectly
  • Fixed bug in Remember operator that could lead to unexpected modification of the stored IOObject
  • Fixed a bug in Weight by Relief which led to an incorrectly empty weight table when the sample ratio was smaller than one
  • Fixed a bug in SVM (Evolutionary) that occurred for nominal labels, when the hold out set ratio was non-zero.
  • Fixed a bug in SVM (Evolutionary) that occurred when the hold out ratio lead to empty training data.
  • Fixed a bug in the view for kernel models with infinite parameters.
  • Fixed possible NPE in process tool
  • Fixed issue in the Statistics tab of a data table failing when data contained infinite values in numerical columns Fix missing error message when trying to run a process execution on AI Hub from a project where the encryption key is not known anymore
  • Fixed Google Cloud Services connections using Service account for Google Drive access

Development

  • Server client now exposes loaded and failed extensions on AI Hub (ServerClient#getLoadedExtensions() and ServerClient#getFailedExtensions())
  • Renamed ServerClient methods listVersionedRepositories and deleteVersionedRepository to listProjects and deleteProject
  • Added ServerClient methods for getProject and createProject to match the naming convention
  • Deprecated and forwarded old ServerClient methods getVersionedRepository and createVersionedRepository to getProject and createProject
  • Exposed json serialization for IOObjects in API of JsonIOObjectEntry