<ul><li>Algorithmic Recourse aims to provide recommendations for individuals impacted by automated decisions to achieve a favorable outcome.</li><li>Existing methods train for proximity, plausibility, and validity separately, leading to subpar recourse recommendations.</li><li>GenRe is a new generative recourse model that trains these objectives jointly, leading to improved performance in cost, plausibility, and validity trade-offs.</li><li>GenRe simplifies recourse recommendation by performing forward sampling over the generative model, providing better outcomes compared to other methods.</li></ul>

From Search To Sampling: Generative Models For Robust Algorithmic Recourse

Discover more