GEMSEO 6.3.0 is out!

Announcements Releases
1

GEMSEO 6.3.0 has been released on October 14, 2025, introducing new features, configuration improvements, and several significant fixes and enhancements to usability and functionality.

Highlights

GEMSEO 6.3.0 adds flexible configuration management, a more informative and customizable progress bar for optimization, new design of experiments (DOE) options, robust database listener management, and major updates to machine learning capabilities via integration with scikit-learn linear model fitting and functional chaos expansion tools.

New Features

  • Progress Bar Customization: Drivers now support a progress_bar_data_name setting, enabling users to choose which data to display at each optimization iteration. The progress bars can now also indicate feasibility status during execution.
  • Configuration Overhaul: Global settings can now be managed through the gemseo.configuration variable, environment variables, or .env files, providing more flexible and centralized control over the library and its logging.
  • DOE Preprocessors: New preprocessors option available for DOE algorithms allows preprocessing before sample generation.
  • Listener Output Management: Database listeners now offer fine-grained control over which outputs are saved, and new helper functions identify these outputs.
  • Enhanced Machine Learning Tools: Introduction of FCERegressor, multiple linear model fitting classes (including scikit-learn algorithms such as LinearRegression, ElasticNet, Lasso, Ridge, and their cross-validated counterparts), and a dedicated subpackage for linear model fitters.
  • SPGL1 Algorithm: Added support for the Spectral Projected Gradient for L1 minimization algorithm.
  • Example Gallery Expansions: A new batch sampling example has been added to DOE documentation.
  • Plot Management: New features for managing figure display and closure with matplotlib post-processors.

Fixes

  • Improved default value validation in settings.
  • Correct storage of output values in the database prior to termination checks.
  • Reliable progress bar logging, regardless of parallelization or database usage.
  • Ensured correct use of MDA settings models and improved multi-disciplinary optimization scenario support.
  • Fixed improper handling of environment variables in .env files.
  • Resolved crashes in specific optimizer uses and matplotlib post-processing edge cases.
  • Corrected parameter passing to SciPy optimizers.

Changes

  • Logging is now enabled by default; it can be disabled by setting configuration.logging.enable = False.
  • The default logger for LoggingContext is now the GEMSEO logger.
  • The logging tools module has been renamed (gemseo.utils.logging_tools is now gemseo.utils.logging).
  • BaseFormulation now distinguishes minimization and maximization problems more clearly.
  • DOE example gallery has been reorganized.
  • CSV read/write functions in DesignSpace now offer a delimiter selection; some arguments have been deprecated to streamline the API.

Be sure to check the complete Changelog for a detailed description of all the elements listed above.

3 Likes