Lines Matching refs:jacobian
79 SparseMatrix* jacobian,
82 CHECK_NOTNULL(jacobian);
86 const int n = jacobian->num_cols();
121 jacobian->SquaredColumnNorm(diagonal_.data());
127 ComputeGradient(jacobian, residuals);
128 ComputeCauchyPoint(jacobian);
131 ComputeGaussNewtonStep(per_solve_options, jacobian, residuals);
148 if (!ComputeSubspaceModel(jacobian)) {
165 // and all calculations involving the Jacobian have to
168 SparseMatrix* jacobian,
171 jacobian->LeftMultiply(residuals, gradient_.data());
177 void DoglegStrategy::ComputeCauchyPoint(SparseMatrix* jacobian) {
179 Vector Jg(jacobian->num_rows());
181 // The Jacobian is scaled implicitly by computing J * (D^-1 * (D^-1 * g))
185 jacobian->RightMultiply(scaled_gradient.data(), Jg.data());
512 SparseMatrix* jacobian,
514 const int n = jacobian->num_cols();
518 // The Jacobian matrix is often quite poorly conditioned. Thus it is
559 // either jacobian or residuals.
561 linear_solver_summary = linear_solver_->Solve(jacobian,
571 jacobian,
637 bool DoglegStrategy::ComputeSubspaceModel(SparseMatrix* jacobian) {
639 Matrix basis_vectors(jacobian->num_cols(), 2);
689 basis_qr.householderQ() * Matrix::Identity(jacobian->num_cols(), 2);
694 Jb(2, jacobian->num_rows());
699 jacobian->RightMultiply(tmp.data(), Jb.row(0).data());
701 jacobian->RightMultiply(tmp.data(), Jb.row(1).data());
