Bias Mitigation

Built-in debiasing methods

Type	Algorithm	Regression	Links
Pre-	BD CB JB BTEO	Natural Discretization Discretization Discretization	Paper, Github, Code Paper, Github, Code Paper, Github, Code Paper, Github, Code
At-	Adv EAdv DAdv AADv & ADAdv Gate FairBatch FairSCL EO_CLA & EO_GLB	Natural Natural Natural Discretization Discretization Discretization Discretization Discretization	Paper, Github, Code Paper, Github, Code Paper, Github, Code Paper, Github, Code Paper, Github, Code Paper, Github, Code Paper, Github, Code Paper, Github, Code
Post-	INLP Soft Gated	Discretization Discretization	Paper, Github, Code Paper, Github, Code

• A list of built-in debiasing methods, check out our paper and cheat sheet for more details.

• All debiasing methods are designed for discrete outputs, but may not naturally support for continuous outputs. Natural refers to methods that naturally support continuous outputs, and discretization refers to discrete proxy label which is an built-in method to adopt classification