Motivation

Currently,plot_parallel_coordinate only supports visualizing hyperparameters stored intrial.params. Users often want to visualize custom metrics or attributes stored intrial.user_attrs (e.g., the number of estimators in a GBDT model, or other auxiliary metrics) alongside hyperparameters to analyze their relationships.

Description of Changes

• Modifiedoptuna/visualization/_parallel_coordinate.py to allow keys fromuser_attrs to be passed in theparams argument.
• Updated the value extraction logic to look up values intrial.user_attrs if the key is not found intrial.params.
• Implemented automatic type inference foruser_attrs: string and boolean values are treated as categorical, while others are treated as numerical.
• Updatedoptuna/visualization/_utils.py (_get_skipped_trial_numbers) to correctly check for the existence of keys inuser_attrs when filtering trials.
• Updated the docstring ofplot_parallel_coordinate to reflect this new capability.

Description of Tests

• Added a new test casetest_get_parallel_coordinate_info_user_attrs intests/visualization_tests/test_parallel_coordinate.py.
• Added comprehensive teststest_get_parallel_coordinate_info_user_attrs_boolean andtest_get_parallel_coordinate_info_user_attrs_missing to verify edge cases.
• Verified thatuser_attrs are correctly extracted and visualized.
• Verified that both numerical and categorical (string/boolean)user_attrs are handled correctly.
• Verified that trials missing requesteduser_attrs are correctly skipped.
• Ran existing tests to ensure no regressions.

Related Issues

#6329