1 // Ceres Solver - A fast non-linear least squares minimizer
2 // Copyright 2015 Google Inc. All rights reserved.
3 // http://ceres-solver.org/
5 // Redistribution and use in source and binary forms, with or without
6 // modification, are permitted provided that the following conditions are met:
8 // * Redistributions of source code must retain the above copyright notice,
9 // this list of conditions and the following disclaimer.
10 // * Redistributions in binary form must reproduce the above copyright notice,
11 // this list of conditions and the following disclaimer in the documentation
12 // and/or other materials provided with the distribution.
13 // * Neither the name of Google Inc. nor the names of its contributors may be
14 // used to endorse or promote products derived from this software without
15 // specific prior written permission.
17 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
18 // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 // ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
21 // LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
22 // CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
23 // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
24 // INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
25 // CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
26 // ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
27 // POSSIBILITY OF SUCH DAMAGE.
29 // Author: sameeragarwal@google.com (Sameer Agarwal)
31 #ifndef CERES_INTERNAL_COMPRESSED_ROW_SPARSE_MATRIX_H_
32 #define CERES_INTERNAL_COMPRESSED_ROW_SPARSE_MATRIX_H_
35 #include "ceres/internal/macros.h"
36 #include "ceres/internal/port.h"
37 #include "ceres/sparse_matrix.h"
38 #include "ceres/types.h"
39 #include "glog/logging.h"
47 class TripletSparseMatrix;
49 class CompressedRowSparseMatrix : public SparseMatrix {
53 // Matrix is assumed to be symmetric but only the lower triangular
54 // part of the matrix is stored.
56 // Matrix is assumed to be symmetric but only the upper triangular
57 // part of the matrix is stored.
61 // Create a matrix with the same content as the TripletSparseMatrix
62 // input. We assume that input does not have any repeated
65 // The storage type of the matrix is set to UNSYMMETRIC.
67 // Caller owns the result.
68 static CompressedRowSparseMatrix* FromTripletSparseMatrix(
69 const TripletSparseMatrix& input);
71 // Create a matrix with the same content as the TripletSparseMatrix
72 // input transposed. We assume that input does not have any repeated
75 // The storage type of the matrix is set to UNSYMMETRIC.
77 // Caller owns the result.
78 static CompressedRowSparseMatrix* FromTripletSparseMatrixTransposed(
79 const TripletSparseMatrix& input);
81 // Use this constructor only if you know what you are doing. This
82 // creates a "blank" matrix with the appropriate amount of memory
83 // allocated. However, the object itself is in an inconsistent state
84 // as the rows and cols matrices do not match the values of
85 // num_rows, num_cols and max_num_nonzeros.
87 // The use case for this constructor is that when the user knows the
88 // size of the matrix to begin with and wants to update the layout
89 // manually, instead of going via the indirect route of first
90 // constructing a TripletSparseMatrix, which leads to more than
91 // double the peak memory usage.
93 // The storage type is set to UNSYMMETRIC.
94 CompressedRowSparseMatrix(int num_rows,
96 int max_num_nonzeros);
98 // Build a square sparse diagonal matrix with num_rows rows and
99 // columns. The diagonal m(i,i) = diagonal(i);
101 // The storage type is set to UNSYMMETRIC
102 CompressedRowSparseMatrix(const double* diagonal, int num_rows);
104 // SparseMatrix interface.
105 virtual ~CompressedRowSparseMatrix();
106 virtual void SetZero();
107 virtual void RightMultiply(const double* x, double* y) const;
108 virtual void LeftMultiply(const double* x, double* y) const;
109 virtual void SquaredColumnNorm(double* x) const;
110 virtual void ScaleColumns(const double* scale);
112 virtual void ToDenseMatrix(Matrix* dense_matrix) const;
113 virtual void ToTextFile(FILE* file) const;
114 virtual int num_rows() const { return num_rows_; }
115 virtual int num_cols() const { return num_cols_; }
116 virtual int num_nonzeros() const { return rows_[num_rows_]; }
117 virtual const double* values() const { return &values_[0]; }
118 virtual double* mutable_values() { return &values_[0]; }
120 // Delete the bottom delta_rows.
121 // num_rows -= delta_rows
122 void DeleteRows(int delta_rows);
124 // Append the contents of m to the bottom of this matrix. m must
125 // have the same number of columns as this matrix.
126 void AppendRows(const CompressedRowSparseMatrix& m);
128 void ToCRSMatrix(CRSMatrix* matrix) const;
130 CompressedRowSparseMatrix* Transpose() const;
132 // Destructive array resizing method.
133 void SetMaxNumNonZeros(int num_nonzeros);
135 // Non-destructive array resizing method.
136 void set_num_rows(const int num_rows) { num_rows_ = num_rows; }
137 void set_num_cols(const int num_cols) { num_cols_ = num_cols; }
139 // Low level access methods that expose the structure of the matrix.
140 const int* cols() const { return &cols_[0]; }
141 int* mutable_cols() { return &cols_[0]; }
143 const int* rows() const { return &rows_[0]; }
144 int* mutable_rows() { return &rows_[0]; }
146 const StorageType storage_type() const { return storage_type_; }
147 void set_storage_type(const StorageType storage_type) {
148 storage_type_ = storage_type;
151 const std::vector<int>& row_blocks() const { return row_blocks_; }
152 std::vector<int>* mutable_row_blocks() { return &row_blocks_; }
154 const std::vector<int>& col_blocks() const { return col_blocks_; }
155 std::vector<int>* mutable_col_blocks() { return &col_blocks_; }
157 // Create a block diagonal CompressedRowSparseMatrix with the given
158 // block structure. The individual blocks are assumed to be laid out
159 // contiguously in the diagonal array, one block at a time.
161 // Caller owns the result.
162 static CompressedRowSparseMatrix* CreateBlockDiagonalMatrix(
163 const double* diagonal,
164 const std::vector<int>& blocks);
166 // Options struct to control the generation of random block sparse
167 // matrices in compressed row sparse format.
169 // The random matrix generation proceeds as follows.
171 // First the row and column block structure is determined by
172 // generating random row and column block sizes that lie within the
175 // Then we walk the block structure of the resulting matrix, and with
176 // probability block_density detemine whether they are structurally
177 // zero or not. If the answer is no, then we generate entries for the
178 // block which are distributed normally.
179 struct RandomMatrixOptions {
180 RandomMatrixOptions()
182 min_row_block_size(0),
183 max_row_block_size(0),
185 min_col_block_size(0),
186 max_col_block_size(0),
191 int min_row_block_size;
192 int max_row_block_size;
194 int min_col_block_size;
195 int max_col_block_size;
197 // 0 < block_density <= 1 is the probability of a block being
198 // present in the matrix. A given random matrix will not have
199 // precisely this density.
200 double block_density;
203 // Create a random CompressedRowSparseMatrix whose entries are
204 // normally distributed and whose structure is determined by
205 // RandomMatrixOptions.
207 // Caller owns the result.
208 static CompressedRowSparseMatrix* CreateRandomMatrix(
209 const RandomMatrixOptions& options);
212 static CompressedRowSparseMatrix* FromTripletSparseMatrix(
213 const TripletSparseMatrix& input, bool transpose);
217 std::vector<int> rows_;
218 std::vector<int> cols_;
219 std::vector<double> values_;
220 StorageType storage_type_;
222 // If the matrix has an underlying block structure, then it can also
223 // carry with it row and column block sizes. This is auxilliary and
224 // optional information for use by algorithms operating on the
225 // matrix. The class itself does not make use of this information in
227 std::vector<int> row_blocks_;
228 std::vector<int> col_blocks_;
231 } // namespace internal
234 #endif // CERES_INTERNAL_COMPRESSED_ROW_SPARSE_MATRIX_H_