| /external/ceres-solver/internal/ceres/ |
| local_parameterization.cc | 53 double* jacobian) const {
54 MatrixRef(jacobian, size_, size_) = Matrix::Identity(size_, size_);
100 double* jacobian) const {
101 MatrixRef m(jacobian, constancy_mask_.size(), local_size_);
133 double* jacobian) const {
134 jacobian[0] = -x[1]; jacobian[1] = -x[2]; jacobian[2] = -x[3]; // NOLINT
135 jacobian[3] = x[0]; jacobian[4] = x[3]; jacobian[5] = -x[2]; // NOLIN
[all...] |
| block_evaluate_preparer.cc | 49 // Point the jacobian blocks directly into the block sparse matrix.
52 SparseMatrix* jacobian,
54 // If the overall jacobian is not available, use the scratch space.
55 if (jacobian == NULL) {
58 jacobian,
64 down_cast<BlockSparseMatrix*>(jacobian)->mutable_values();
|
| block_evaluate_preparer.h | 31 // A evaluate preparer which puts jacobian the evaluated jacobian blocks
33 // The evaluator takes care to avoid evaluating the jacobian for fixed
57 // Point the jacobian blocks directly into the block sparse matrix, if
58 // jacobian is non-null. Otherwise, uses an internal per-thread buffer to
62 SparseMatrix* jacobian,
68 // For the case that the overall jacobian is not available, but the
|
| corrector.h | 42 // to the residual and jacobian of a least squares problem based on a
47 // corresponding corrections to the residual and jacobian. For the
64 // CorrectResidual, because the jacobian correction depends on the
71 // jacobian = sqrt(rho[1]) * jacobian -
72 // sqrt(rho[1]) * alpha / sq_norm * residuals residuals' * jacobian.
74 // The method assumes that the jacobian has row-major storage. It is
76 // jacobian is not null.
78 double* residuals, double* jacobian);
|
| trust_region_minimizer.h | 55 void EstimateScale(const SparseMatrix& jacobian, double* scale) const;
58 const SparseMatrix* jacobian,
|
| corrector_test.cc | 60 double jacobian = 10.0; local
74 // The jacobian in this case will be
75 // sqrt(kRho[1]) * (1 - kAlpha) * jacobian.
76 const double kExpectedJacobian = sqrt(kRho[1]) * (1 - kAlpha) * jacobian;
79 c.CorrectJacobian(1.0, 1.0, &residuals, &jacobian);
83 ASSERT_NEAR(kExpectedJacobian, jacobian, 1e-6);
88 double jacobian = 10.0; local
102 // The jacobian in this case will be
103 // sqrt(kRho[1]) * jacobian.
104 const double kExpectedJacobian = sqrt(kRho[1]) * jacobian;
116 double jacobian = 10.0; local
147 double jacobian[2 * 3]; local
215 double jacobian[2 * 3]; local
[all...] |
| dogleg_strategy.h | 48 // because the Jacobian is often rank-deficient and in such cases
62 SparseMatrix* jacobian,
82 LinearSolver::Summary ComputeGaussNewtonStep(SparseMatrix* jacobian,
84 void ComputeCauchyPoint(SparseMatrix* jacobian);
85 void ComputeGradient(SparseMatrix* jacobian, const double* residuals);
87 bool ComputeSubspaceModel(SparseMatrix* jacobian);
132 // user has recomputed the Jacobian matrix and new Gauss-Newton
|
| corrector.cc | 47 // and the jacobian are scaled by the squareroot of the derivative
113 double* residuals, double* jacobian) {
115 DCHECK(jacobian != NULL);
117 MatrixRef j_ref(jacobian, nrow, ncol);
|
| evaluator.h | 51 // squares objective. This insulates the optimizer from issues like Jacobian
73 // This is used for computing the cost, residual and Jacobian for
78 // The residual, gradients and jacobian pointers can be NULL, in
96 CRSMatrix* jacobian);
98 // Build and return a sparse matrix for storing and working with the Jacobian
99 // of the objective function. The jacobian has dimensions
101 // sparse. Since the sparsity pattern of the Jacobian remains constant over
111 // the jacobian for use with CHOLMOD, where as BlockOptimizationProblem
112 // creates a BlockSparseMatrix representation of the jacobian for use in the
117 // residuals, and jacobian in the corresponding arguments. Both residuals an
[all...] |
| scratch_evaluate_preparer.h | 33 // care to avoid evaluating the jacobian for fixed parameters.
57 SparseMatrix* jacobian,
61 // Scratch space for the jacobians; each jacobian is packed one after another.
|
| levenberg_marquardt_strategy_test.cc | 114 Matrix jacobian(2,3);
115 jacobian.setZero();
116 jacobian(0,0) = 0.0;
117 jacobian(0,1) = 1.0;
118 jacobian(1,1) = 1.0;
119 jacobian(0,2) = 100.0;
123 DenseSparseMatrix dsm(jacobian);
|
| compressed_row_jacobian_writer.cc | 50 // Count the number of jacobian nonzeros.
64 // Allocate storage for the jacobian with some extra space at the end.
65 // Allocate more space than needed to store the jacobian so that when the LM
68 CompressedRowSparseMatrix* jacobian = local
76 int* rows = jacobian->mutable_rows();
77 int* cols = jacobian->mutable_cols();
112 // parameter vector. This code mirrors that in Write(), where jacobian
121 // This is the position in the values array of the jacobian where this
122 // row of the jacobian block should go.
139 vector<int>& col_blocks = *(jacobian->mutable_col_blocks())
158 CompressedRowSparseMatrix* jacobian = local
[all...] |
| dogleg_strategy.cc | 78 SparseMatrix* jacobian,
81 CHECK_NOTNULL(jacobian);
85 const int n = jacobian->num_cols();
120 jacobian->SquaredColumnNorm(diagonal_.data());
126 ComputeGradient(jacobian, residuals);
127 ComputeCauchyPoint(jacobian);
130 ComputeGaussNewtonStep(jacobian, residuals);
147 if (!ComputeSubspaceModel(jacobian)) {
164 // and all calculations involving the Jacobian have to
167 SparseMatrix* jacobian,
[all...] |
| evaluator.cc | 111 scoped_ptr<CompressedRowSparseMatrix> jacobian; local
113 jacobian.reset(
135 jacobian.get())) {
140 jacobian->ToCRSMatrix(output_jacobian);
|
| levenberg_marquardt_strategy.h | 53 SparseMatrix* jacobian,
|
| parameter_block_test.cc | 71 // Ensure the local parameterization jacobian result is correctly computed.
102 double* jacobian) const {
103 jacobian[0] = *x * 2;
134 // Stops computing the jacobian after the first time.
149 virtual bool ComputeJacobian(const double* x, double* jacobian) const {
151 jacobian[0] = 0;
|
| dense_jacobian_writer.h | 31 // A jacobian writer that writes to dense Eigen matrices.
58 // them over to the larger jacobian later.
72 SparseMatrix* jacobian) {
74 if (jacobian != NULL) {
75 dense_jacobian = down_cast<DenseSparseMatrix*>(jacobian);
82 // Now copy the jacobians for each parameter into the dense jacobian matrix.
|
| local_parameterization_test.cc | 55 double jacobian[9]; local
56 parameterization.ComputeJacobian(x, jacobian);
60 EXPECT_EQ(jacobian[k], (i == j) ? 1.0 : 0.0);
106 double jacobian[4 * 3]; local
107 parameterization.ComputeJacobian(x, jacobian);
113 EXPECT_EQ(jacobian[jacobian_cursor], delta_cursor == k ? 1.0 : 0.0);
118 EXPECT_EQ(jacobian[jacobian_cursor], 0.0);
184 // Autodiff jacobian at delta_x = 0.
188 double jacobian[12]; local
189 param.ComputeJacobian(x, jacobian);
[all...] |
| evaluator_test.cc | 68 // evaluator into the "local" jacobian. In the tests, the "subset
70 // from these jacobians. Put values in the jacobian that make this
80 MatrixRef jacobian(jacobians[k],
84 jacobian.col(j).setConstant(kFactor * (j + 1));
99 const double jacobian[200]; member in struct:ceres::internal::ExpectedEvaluation
188 scoped_ptr<SparseMatrix> jacobian(evaluator->CreateJacobian());
192 ASSERT_EQ(expected_num_rows, jacobian->num_rows());
193 ASSERT_EQ(expected_num_cols, jacobian->num_cols());
202 expected_jacobian != NULL ? jacobian.get() : NULL));
206 jacobian->ToDenseMatrix(&actual_jacobian)
609 double* jacobian = jacobians[0]; local
786 CRSMatrix jacobian; local
[all...] |
| levenberg_marquardt_strategy.cc | 66 SparseMatrix* jacobian,
69 CHECK_NOTNULL(jacobian);
73 const int num_parameters = jacobian->num_cols();
79 jacobian->SquaredColumnNorm(diagonal_.data());
103 // Then x can be found as x = -y, but the inputs jacobian and residuals
106 linear_solver_->Solve(jacobian, residuals, solve_options, step);
|
| normal_prior_test.cc | 70 double * jacobian = new double[num_rows * num_cols]; local
74 prior.Evaluate(&x, residuals.data(), &jacobian);
82 MatrixRef J(jacobian, num_rows, num_cols);
87 delete []jacobian;
|
| trust_region_minimizer.cc | 77 // of the Jacobian.
78 void TrustRegionMinimizer::EstimateScale(const SparseMatrix& jacobian,
80 jacobian.SquaredColumnNorm(scale);
81 for (int i = 0; i < jacobian.num_cols(); ++i) {
94 const SparseMatrix* jacobian,
121 SparseMatrix* jacobian = CHECK_NOTNULL(options_.jacobian); local
155 // Do initial cost and Jacobian evaluation.
157 if (!evaluator->Evaluate(x.data(), &cost, residuals.data(), NULL, jacobian)) {
158 LOG(WARNING) << "Terminating: Residual and Jacobian evaluation failed."
[all...] |
| minimizer.h | 83 jacobian = NULL;
119 // Jacobian matrix. The Options struct does not own this pointer.
127 // Object holding the Jacobian matrix. It is assumed that the
131 SparseMatrix* jacobian; member in struct:ceres::internal::Minimizer::Options
|
| /external/ceres-solver/include/ceres/ |
| local_parameterization.h | 109 // Jacobian which is needed to compute the Jacobian of f w.r.t delta.
123 // The jacobian of Plus(x, delta) w.r.t delta at delta = 0.
124 virtual bool ComputeJacobian(const double* x, double* jacobian) const = 0;
144 double* jacobian) const;
162 double* jacobian) const;
182 double* jacobian) const;
|
| /external/eigen/unsupported/Eigen/src/AutoDiff/ |
| AutoDiffVector.h | 73 : m_values(other.values()), m_jacobian(other.jacobian())
77 : m_values(other.values()), m_jacobian(other.jacobian())
84 m_jacobian = other.jacobian();
91 m_jacobian = other.jacobian();
98 inline const JacobianType& jacobian() const { return m_jacobian; } function in class:Eigen::AutoDiffVector
99 inline JacobianType& jacobian() { return m_jacobian; } function in class:Eigen::AutoDiffVector
111 m_jacobian + other.jacobian());
119 m_jacobian += other.jacobian();
133 m_jacobian - other.jacobian());
141 m_jacobian -= other.jacobian();
[all...] |
