1 | // $Id: Local.h 495 2006-01-11 17:27:22Z peter $ |
---|
2 | |
---|
3 | #ifndef _theplu_statistics_regression_local_ |
---|
4 | #define _theplu_statistics_regression_local_ |
---|
5 | |
---|
6 | #include <c++_tools/statistics/Kernel.h> |
---|
7 | #include <c++_tools/statistics/OneDimensionalWeighted.h> |
---|
8 | #include <c++_tools/gslapi/vector.h> |
---|
9 | |
---|
10 | #include <iostream> |
---|
11 | |
---|
12 | namespace theplu { |
---|
13 | namespace statistics { |
---|
14 | namespace regression { |
---|
15 | |
---|
16 | /// |
---|
17 | /// Class for Locally weighted regression. |
---|
18 | /// |
---|
19 | /// Locally weighted regression is an algorithm for learning |
---|
20 | /// continuous non-linear mappings in a non-parametric manner. In |
---|
21 | /// locally weighted regression, points are weighted by proximity to |
---|
22 | /// the current x in question using a Kernel. A weighted regression |
---|
23 | /// is then computed using the weighted points and a specific |
---|
24 | /// Regression method. This procedure is repeated, which results in |
---|
25 | /// a pointwise approximation of the underlying (unknown) function. |
---|
26 | /// |
---|
27 | class Local |
---|
28 | { |
---|
29 | |
---|
30 | public: |
---|
31 | /// |
---|
32 | /// Constructor taking type of \a regressor, |
---|
33 | /// type of \a kernel. |
---|
34 | /// |
---|
35 | inline Local(OneDimensionalWeighted& r, Kernel& k) |
---|
36 | : kernel_(&k), regressor_(&r) {} |
---|
37 | |
---|
38 | /// |
---|
39 | /// Destructor |
---|
40 | /// |
---|
41 | virtual ~Local(void) {}; |
---|
42 | |
---|
43 | |
---|
44 | /// |
---|
45 | /// adding a data point |
---|
46 | /// |
---|
47 | inline void add(const double x, const double y) |
---|
48 | { data_.push_back(std::make_pair(x,y)); } |
---|
49 | |
---|
50 | /// |
---|
51 | /// Function returning predicted values |
---|
52 | /// |
---|
53 | inline const gslapi::vector& y_predicted(void) const |
---|
54 | { return y_predicted_; } |
---|
55 | |
---|
56 | /// |
---|
57 | /// Function returning error of predictions |
---|
58 | /// |
---|
59 | inline const gslapi::vector& y_err(void) const { return y_err_; } |
---|
60 | |
---|
61 | /// |
---|
62 | /// @todo doc |
---|
63 | /// @param nof_points Number of points used in each fit |
---|
64 | /// @param step_size Size of step between each fit |
---|
65 | /// |
---|
66 | void fit(const size_t step_size, const size_t nof_points); |
---|
67 | |
---|
68 | /// |
---|
69 | /// @return x-values where fitting was performed. |
---|
70 | /// |
---|
71 | inline const gslapi::vector& x(void) const { return x_; } |
---|
72 | |
---|
73 | private: |
---|
74 | /// |
---|
75 | /// Copy Constructor. (Not implemented) |
---|
76 | /// |
---|
77 | Local(const Local&); |
---|
78 | |
---|
79 | std::vector<std::pair<double, double> > data_; |
---|
80 | Kernel* kernel_; |
---|
81 | OneDimensionalWeighted* regressor_; |
---|
82 | gslapi::vector x_; |
---|
83 | gslapi::vector y_predicted_; |
---|
84 | gslapi::vector y_err_; |
---|
85 | |
---|
86 | }; |
---|
87 | |
---|
88 | /// |
---|
89 | /// The output operator for the RegressionLocal class. |
---|
90 | /// |
---|
91 | std::ostream& operator<<(std::ostream&, const Local& ); |
---|
92 | |
---|
93 | |
---|
94 | }}} // of namespaces regression, statistics and thep |
---|
95 | |
---|
96 | #endif |
---|