Template Function proxsuite::proxqp::dense::compute_backward
Defined in File compute_ECJ.hpp
Function Documentation
-
template<typename T>
void proxsuite::proxqp::dense::compute_backward(dense::QP<T> &solved_qp, VecRef<T> loss_derivative, T eps = 1.E-4, T rho_new = 1.E-6, T mu_new = 1.E-6) Computes the backward pass for the QP solver.
- Parameters:
solved_qp – The solved quadratic programming (QP) problem.
loss_derivative – The derivative of the loss function wrt the solution of the QP.
eps – The epsilon value for the iterative refinement, with a default value of 1.E-4.
rho_new – The new primal proximal parameter, with a default value of 1.E-6.
mu_new – The new dual proximal parameter used for both equality and inequality, with a default value of 1.E-6.