themis-ml

Seamlessly integrates with existing scikit-learn workflows, as it's built on top of sklearn, allowing easy adoption in standard ML pipelines.

Implements techniques across preprocessing (e.g., relabelling), model estimation (e.g., additive counterfactually fair estimator), and postprocessing (e.g., reject option classification) for comprehensive bias mitigation.

Includes utility functions to load commonly used fairness datasets like German Credit and Census Income, facilitating quick experimentation and benchmarking.

Provides references to academic papers and comprehensive documentation on Read the Docs, ensuring users can understand the underlying fairness concepts.