1
// Ceres Solver - A fast non-linear least squares minimizer
2
// Copyright 2010, 2011, 2012 Google Inc. All rights reserved.
3
// http://code.google.com/p/ceres-solver/
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
// Helpers for making CostFunctions as needed by the least squares framework,
32
// with Jacobians computed via automatic differentiation. For more information
33
// on automatic differentation, see the wikipedia article at
34
// http://en.wikipedia.org/wiki/Automatic_differentiation
36
// To get an auto differentiated cost function, you must define a class with a
37
// templated operator() (a functor) that computes the cost function in terms of
38
// the template parameter T. The autodiff framework substitutes appropriate
39
// "jet" objects for T in order to compute the derivative when necessary, but
40
// this is hidden, and you should write the function as if T were a scalar type
41
// (e.g. a double-precision floating point number).
43
// The function must write the computed value in the last argument (the only
44
// non-const one) and return true to indicate success.
46
// For example, consider a scalar error e = k - x'y, where both x and y are
47
// two-dimensional column vector parameters, the prime sign indicates
48
// transposition, and k is a constant. The form of this error, which is the
49
// difference between a constant and an expression, is a common pattern in least
50
// squares problems. For example, the value x'y might be the model expectation
51
// for a series of measurements, where there is an instance of the cost function
52
// for each measurement k.
54
// The actual cost added to the total problem is e^2, or (k - x'k)^2; however,
55
// the squaring is implicitly done by the optimization framework.
57
// To write an auto-differentiable cost function for the above model, first
60
// class MyScalarCostFunction {
61
// MyScalarCostFunction(double k): k_(k) {}
63
// template <typename T>
64
// bool operator()(const T* const x , const T* const y, T* e) const {
65
// e[0] = T(k_) - x[0] * y[0] + x[1] * y[1];
73
// Note that in the declaration of operator() the input parameters x and y come
74
// first, and are passed as const pointers to arrays of T. If there were three
75
// input parameters, then the third input parameter would come after y. The
76
// output is always the last parameter, and is also a pointer to an array. In
77
// the example above, e is a scalar, so only e[0] is set.
79
// Then given this class definition, the auto differentiated cost function for
80
// it can be constructed as follows.
82
// CostFunction* cost_function
83
// = new AutoDiffCostFunction<MyScalarCostFunction, 1, 2, 2>(
84
// new MyScalarCostFunction(1.0)); ^ ^ ^
86
// Dimension of residual ------+ | |
87
// Dimension of x ----------------+ |
88
// Dimension of y -------------------+
90
// In this example, there is usually an instance for each measumerent of k.
92
// In the instantiation above, the template parameters following
93
// "MyScalarCostFunction", "1, 2, 2", describe the functor as computing a
94
// 1-dimensional output from two arguments, both 2-dimensional.
96
// The autodiff cost function also supports cost functions with a
97
// runtime-determined number of residuals. For example:
99
// CostFunction* cost_function
100
// = new AutoDiffCostFunction<MyScalarCostFunction, DYNAMIC, 2, 2>(
101
// new CostFunctionWithDynamicNumResiduals(1.0), ^ ^ ^
102
// runtime_number_of_residuals); <----+ | | |
105
// Actual number of residuals ------+ | | |
106
// Indicate dynamic number of residuals ---------+ | |
107
// Dimension of x -------------------------------------+ |
108
// Dimension of y ----------------------------------------+
110
// The framework can currently accommodate cost functions of up to 6 independent
111
// variables, and there is no limit on the dimensionality of each of them.
113
// WARNING #1: Since the functor will get instantiated with different types for
114
// T, you must to convert from other numeric types to T before mixing
115
// computations with other variables of type T. In the example above, this is
116
// seen where instead of using k_ directly, k_ is wrapped with T(k_).
118
// WARNING #2: A common beginner's error when first using autodiff cost
119
// functions is to get the sizing wrong. In particular, there is a tendency to
120
// set the template parameters to (dimension of residual, number of parameters)
121
// instead of passing a dimension parameter for *every parameter*. In the
122
// example above, that would be <MyScalarCostFunction, 1, 2>, which is missing
123
// the last '2' argument. Please be careful when setting the size parameters.
125
#ifndef CERES_PUBLIC_AUTODIFF_COST_FUNCTION_H_
126
#define CERES_PUBLIC_AUTODIFF_COST_FUNCTION_H_
128
#include <glog/logging.h>
129
#include "ceres/internal/autodiff.h"
130
#include "ceres/internal/scoped_ptr.h"
131
#include "ceres/sized_cost_function.h"
132
#include "ceres/types.h"
136
// A cost function which computes the derivative of the cost with respect to
137
// the parameters (a.k.a. the jacobian) using an autodifferentiation framework.
138
// The first template argument is the functor object, described in the header
139
// comment. The second argument is the dimension of the residual (or
140
// ceres::DYNAMIC to indicate it will be set at runtime), and subsequent
141
// arguments describe the size of the Nth parameter, one per parameter.
143
// The constructors take ownership of the cost functor.
145
// If the number of residuals (argument "M" below) is ceres::DYNAMIC, then the
146
// two-argument constructor must be used. The second constructor takes a number
147
// of residuals (in addition to the templated number of residuals). This allows
148
// for varying the number of residuals for a single autodiff cost function at
150
template <typename CostFunctor,
151
int M, // Number of residuals, or ceres::DYNAMIC.
152
int N0, // Number of parameters in block 0.
153
int N1 = 0, // Number of parameters in block 1.
154
int N2 = 0, // Number of parameters in block 2.
155
int N3 = 0, // Number of parameters in block 3.
156
int N4 = 0, // Number of parameters in block 4.
157
int N5 = 0> // Number of parameters in block 5.
158
class AutoDiffCostFunction :
159
public SizedCostFunction<M, N0, N1, N2, N3, N4, N5> {
161
// Takes ownership of functor. Uses the template-provided value for the
162
// number of residuals ("M").
163
explicit AutoDiffCostFunction(CostFunctor* functor)
164
: functor_(functor) {
165
CHECK_NE(M, DYNAMIC) << "Can't run the fixed-size constructor if the "
166
<< "number of residuals is set to ceres::DYNAMIC.";
169
// Takes ownership of functor. Ignores the template-provided number of
170
// residuals ("M") in favor of the "num_residuals" argument provided.
172
// This allows for having autodiff cost functions which return varying
173
// numbers of residuals at runtime.
174
AutoDiffCostFunction(CostFunctor* functor, int num_residuals)
175
: functor_(functor) {
176
CHECK_EQ(M, DYNAMIC) << "Can't run the dynamic-size constructor if the "
177
<< "number of residuals is not ceres::DYNAMIC.";
178
SizedCostFunction<M, N0, N1, N2, N3, N4, N5>::set_num_residuals(num_residuals);
181
virtual ~AutoDiffCostFunction() {}
183
// Implementation details follow; clients of the autodiff cost function should
184
// not have to examine below here.
186
// To handle varardic cost functions, some template magic is needed. It's
187
// mostly hidden inside autodiff.h.
188
virtual bool Evaluate(double const* const* parameters,
190
double** jacobians) const {
192
return internal::VariadicEvaluate<
193
CostFunctor, double, N0, N1, N2, N3, N4, N5>
194
::Call(*functor_, parameters, residuals);
196
return internal::AutoDiff<CostFunctor, double,
197
N0, N1, N2, N3, N4, N5>::Differentiate(
200
SizedCostFunction<M, N0, N1, N2, N3, N4, N5>::num_residuals(),
206
internal::scoped_ptr<CostFunctor> functor_;
211
#endif // CERES_PUBLIC_AUTODIFF_COST_FUNCTION_H_