Librec-auto

This release contains data and configuration to demonstrate the capabilities of librec-auto 0.2.6.

Sorry for the confusion!

New functionality in librec-auto:

- Support for multiple protected features in re-ranking and evaluation
- Support for Python-side evaluation metrics
- Support for Python-side recommendation algorithms and MS Recommenders integration
- Support of optuna-based TPE black-box optimization
- Additional re-ranker implementations including FA*IR and maximum-flow.

This release extends (and makes mandatory) the error-checking capabilities of librec-auto. Users will get more information about problems with their configuration files and fewer incomprehensible messages. Also, error messages are now consolidated in the output.xml file.

Minor bug fix for black-box optimization function.

This release has additional extensibility features for librec-auto including Python implementations of evaluation metrics, and easier creation of user-defined re-ranking algorithms. The re-ranking algorithms also have been substantially optimized. Additional fairness-aware metrics are available including miscalibration. Study output is now better integrated and there is a simple wizard for setting up new study configurations.