Google Proposes Efficient and Modular Implicit Differentiation for Optimization Problems
A research team from Google Research combines the benefits of implicit differentiation and autodiff and proposes a unified, efficient and modular approach for implicit differentiation of optimization problems.