slope 6.6.0
Loading...
Searching...
No Matches
folds.h
Go to the documentation of this file.
1
9#pragma once
10
11#include "eigen_compat.h"
12#include "utils.h"
13#include <Eigen/Core>
14#include <cstdint>
15#include <vector>
16
17namespace slope {
18
19using slope::all;
20
28class Folds
29{
30public:
37 : n_folds(0)
38 , n_repeats(0)
39 {
40 }
41
55 Folds(int n_samples, int n_folds, int n_repeats = 1, uint64_t seed = 42)
58 {
59 folds.resize(n_repeats);
60 for (int rep = 0; rep < n_repeats; ++rep) {
61 // Use a different seed for each repetition
62 uint64_t rep_seed = seed + rep;
63 folds[rep] = createFolds(n_samples, n_folds, rep_seed);
64 }
65 }
66
74 explicit Folds(const std::vector<std::vector<int>>& user_folds)
75 : n_folds(user_folds.size())
76 , n_repeats(1)
77 {
78 folds.emplace_back(user_folds);
79 }
80
86 explicit Folds(const std::vector<std::vector<std::vector<int>>>& user_folds)
87 : folds(user_folds)
88 , n_folds(user_folds[0].size())
89 , n_repeats(user_folds.size())
90 {
91 }
92
100 const std::vector<int>& getTestIndices(size_t fold_idx,
101 size_t rep_idx = 0) const;
102
111 std::vector<int> getTrainingIndices(size_t fold_idx,
112 size_t rep_idx = 0) const;
113
130 template<typename T>
131 auto split(Eigen::EigenBase<T>& x,
132 const Eigen::MatrixXd& y,
133 size_t fold_idx,
134 size_t rep_idx = 0) const
135 {
136 auto test_idx = getTestIndices(fold_idx, rep_idx);
137 auto train_idx = getTrainingIndices(fold_idx, rep_idx);
138
139 auto x_test = subset(x.derived(), test_idx);
140 Eigen::MatrixXd y_test = y(test_idx, all);
141
142 auto x_train = subset(x.derived(), train_idx);
143 Eigen::MatrixXd y_train = y(train_idx, all);
144
145 return std::make_tuple(x_train, y_train, x_test, y_test);
146 }
147
153 size_t numFolds() const { return n_folds; }
154
160 size_t numRepetitions() const { return n_repeats; }
161
167 size_t numEvals() const { return n_repeats * n_folds; }
168
169 std::vector<std::vector<std::vector<int>>>
171 std::size_t n_folds;
172 std::size_t n_repeats;
173
174private:
189 static std::vector<std::vector<int>> createFolds(int n_samples,
190 int n_folds,
191 uint64_t seed);
192};
193
194} // namespace slope
std::size_t n_folds
Number of folds.
Definition folds.h:171
Folds(const std::vector< std::vector< int > > &user_folds)
Constructor for user-provided folds.
Definition folds.h:74
Folds(int n_samples, int n_folds, int n_repeats=1, uint64_t seed=42)
Constructor for generating random folds with optional repetitions.
Definition folds.h:55
size_t numEvals() const
Get the total number of folds (repetitions * folds).
Definition folds.h:167
size_t numFolds() const
Get the number of folds.
Definition folds.h:153
std::vector< int > getTrainingIndices(size_t fold_idx, size_t rep_idx=0) const
Get training indices for a specific fold and repetition.
const std::vector< int > & getTestIndices(size_t fold_idx, size_t rep_idx=0) const
Get test indices for a specific fold and repetition.
size_t numRepetitions() const
Get the number of repetitions.
Definition folds.h:160
auto split(Eigen::EigenBase< T > &x, const Eigen::MatrixXd &y, size_t fold_idx, size_t rep_idx=0) const
Split data into training and test sets for a specific fold and repetition.
Definition folds.h:131
Folds(const std::vector< std::vector< std::vector< int > > > &user_folds)
Constructor for user-provided repeated folds.
Definition folds.h:86
Folds()
Default constructor.
Definition folds.h:36
std::vector< std::vector< std::vector< int > > > folds
Indices for each fold in each repetition.
Definition folds.h:170
std::size_t n_repeats
Number of repetitions.
Definition folds.h:172
Eigen compatibility layer for version differences.
Namespace containing SLOPE regression implementation.
Definition clusters.h:11
T subset(const Eigen::EigenBase< T > &x, const std::vector< int > &indices)
Extract a subset of rows from an Eigen matrix.
Definition utils.h:258
Various utility functions.