Module diff_qp 

Differentiable Quadratic Programming (OptNet-style).

Solves the QP:

min ½ x’Qx + c’x s.t. Gx ≤ h Ax = b

and computes gradients of the optimal solution x* w.r.t. all problem parameters (Q, c, G, h, A, b) via implicit differentiation of the KKT conditions.

References

• Amos & Kolter (2017). “OptNet: Differentiable Optimization as a Layer in Neural Networks.” ICML.

Structs

DifferentiableQP

A differentiable QP layer.