package ceres.internal

Get desktop application:
View/edit binary Protocol Buffers messages

Used in: BlockStructureProto, CompressedRowProto

• optional int32 size = 1

  The span of the block.

• optional int32 position = 2

  Position along the row or column (depending on storage orientation).

A block sparse matrix, either in column major or row major format.

Used in: SparseMatrixProto

• optional int64 num_rows = 2

• optional int64 num_cols = 3

• optional int64 num_nonzeros = 4

• repeated double values = 1

• optional BlockStructureProto block_structure = 5

Used in: BlockSparseMatrixProto

• repeated BlockProto cols = 1

• repeated CompressedRowProto rows = 2

Used in: CompressedRowProto

• optional int32 block_id = 1

  Column or row block id as appropriate.

• optional int32 position = 2

  Position in the values array the cell is located. Each cell is stored as a row-major chunk inside the values array.

A single row or column, depending on the matrix type.

Used in: BlockStructureProto

• optional BlockProto block = 2

• repeated CellProto cells = 1

Used in: SparseMatrixProto

• optional int64 num_rows = 4

• optional int64 num_cols = 5

• repeated int64 rows = 1

• repeated int64 cols = 2

• repeated double values = 3

Used in: SparseMatrixProto

• optional int64 num_rows = 1

• optional int64 num_cols = 2

• repeated double values = 3

  Entries are stored in row-major order.

A linear least squares problem. Given a matrix A, an optional diagonal matrix D as a vector, and a vector b, the proto represents the following linear least squares problem. | A | x = | b | | D | | 0 | If D is empty, then the problem is considered to be A x = b The desired solution for the problem is the vector x that solves the following optimization problem: arg min_x ||Ax - b||^2 + ||Dx||^2 If x is present, then it is the expected solution to the problem. The dimensions of A, b, x, and D should be consistent.

• optional SparseMatrixProto a = 1

• repeated double b = 2

• repeated double d = 3

• repeated double x = 4

• optional int32 num_eliminate_blocks = 5

  If the problem is of SfM type, i.e it has a generalized bi-partite structure, then num_eliminate_blocks is the number of column blocks that are to eliminated in the formation of the Schur complement. For more details see explicit_schur_complement_solver.h.

A sparse matrix. It is a union; only one field is permitted. If new sparse implementations are added, update this proto accordingly.

Used in: LinearLeastSquaresProblemProto

• optional TripletSparseMatrixProto triplet_matrix = 1

• optional BlockSparseMatrixProto block_matrix = 2

• optional CompressedRowSparseMatrixProto compressed_row_matrix = 3

• optional DenseSparseMatrixProto dense_matrix = 4

Used in: SparseMatrixProto

• optional int64 num_rows = 4

• optional int64 num_cols = 5

• optional int64 num_nonzeros = 6

• repeated int64 rows = 1

  The data is stored as three arrays. For each i, values(i) is stored at the location (rows(i), cols(i)). If the there are multiple entries with the same (rows(i), cols(i)), the values entries corresponding to them are summed up.

• repeated int64 cols = 2

• repeated double values = 3