sparse axis aligned kernel for GP modeling

[roussel-ryan]

This pull request introduces support for constructing MAP SAAS (Sparse Axis-Aligned Subspace) Gaussian Process models in the Bayesian modeling workflow. It adds a utility method for building and training MAP SAAS GPs, enables users to specify which outputs should use SAAS priors, and integrates this functionality into the standard model construction pipeline. Additionally, a new example notebook demonstrates building GP models from scratch, and relevant tests are updated.

MAP SAAS GP Model Support

• Added a static method build_map_saas_gp to base_model.py for creating and training MAP SAAS SingleTaskGP models, with error handling and documentation.
• Integrated the option to apply SAAS priors to specific outputs via a new saas_outputs field in StandardModelConstructor, and updated the model construction logic to use MAP SAAS GPs when appropriate. [1] [2]

Documentation and Examples

• Added a new notebook saas.ipynb demonstrating how to build GP models (including with SAAS priors) from scratch for data visualization and sensitivity analysis.
• Updated the existing model_creation.ipynb notebook to use the xopt-dev kernel and reflect the latest Python version. [1] [2]

Testing

• Added a test to ensure that the SAAS model constructor works as expected when specified in StandardModelConstructor.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[roussel-ryan]

@copilot add tests to meet coverage criteria

[Copilot]

@copilot add tests to meet coverage criteria

Added targeted SAAS coverage tests in test_model_constructor.py to cover MAP SAAS training/error paths and SAAS-output behavior (single and multi-output cases). Addressed in commit 7b6b43d.