1 | //===----------------------------------------------------------------------===// |
---|---|
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | |
9 | // <random> |
10 | |
11 | // class bernoulli_distribution |
12 | |
13 | // template<class _URNG> result_type operator()(_URNG& g); |
14 | |
15 | #include <random> |
16 | #include <numeric> |
17 | #include <vector> |
18 | #include <cassert> |
19 | #include <cstddef> |
20 | |
21 | #include "test_macros.h" |
22 | |
23 | template <class T> |
24 | inline |
25 | T |
26 | sqr(T x) |
27 | { |
28 | return x * x; |
29 | } |
30 | |
31 | int main(int, char**) |
32 | { |
33 | { |
34 | typedef std::bernoulli_distribution D; |
35 | typedef std::minstd_rand G; |
36 | G g; |
37 | D d(.75); |
38 | const int N = 100000; |
39 | std::vector<D::result_type> u; |
40 | for (int i = 0; i < N; ++i) |
41 | u.push_back(d(g)); |
42 | double mean = std::accumulate(u.begin(), u.end(), |
43 | double(0)) / u.size(); |
44 | double var = 0; |
45 | double skew = 0; |
46 | double kurtosis = 0; |
47 | for (std::size_t i = 0; i < u.size(); ++i) |
48 | { |
49 | double dbl = (u[i] - mean); |
50 | double d2 = sqr(dbl); |
51 | var += d2; |
52 | skew += dbl * d2; |
53 | kurtosis += d2 * d2; |
54 | } |
55 | var /= u.size(); |
56 | double dev = std::sqrt(x: var); |
57 | skew /= u.size() * dev * var; |
58 | kurtosis /= u.size() * var * var; |
59 | kurtosis -= 3; |
60 | double x_mean = d.p(); |
61 | double x_var = d.p()*(1-d.p()); |
62 | double x_skew = (1 - 2 * d.p())/std::sqrt(x: x_var); |
63 | double x_kurtosis = (6 * sqr(d.p()) - 6 * d.p() + 1)/x_var; |
64 | assert(std::abs((mean - x_mean) / x_mean) < 0.01); |
65 | assert(std::abs((var - x_var) / x_var) < 0.01); |
66 | assert(std::abs((skew - x_skew) / x_skew) < 0.01); |
67 | assert(std::abs((kurtosis - x_kurtosis) / x_kurtosis) < 0.02); |
68 | } |
69 | { |
70 | typedef std::bernoulli_distribution D; |
71 | typedef std::minstd_rand G; |
72 | G g; |
73 | D d(.25); |
74 | const int N = 100000; |
75 | std::vector<D::result_type> u; |
76 | for (int i = 0; i < N; ++i) |
77 | u.push_back(d(g)); |
78 | double mean = std::accumulate(u.begin(), u.end(), |
79 | double(0)) / u.size(); |
80 | double var = 0; |
81 | double skew = 0; |
82 | double kurtosis = 0; |
83 | for (std::size_t i = 0; i < u.size(); ++i) |
84 | { |
85 | double dbl = (u[i] - mean); |
86 | double d2 = sqr(dbl); |
87 | var += d2; |
88 | skew += dbl * d2; |
89 | kurtosis += d2 * d2; |
90 | } |
91 | var /= u.size(); |
92 | double dev = std::sqrt(x: var); |
93 | skew /= u.size() * dev * var; |
94 | kurtosis /= u.size() * var * var; |
95 | kurtosis -= 3; |
96 | double x_mean = d.p(); |
97 | double x_var = d.p()*(1-d.p()); |
98 | double x_skew = (1 - 2 * d.p())/std::sqrt(x: x_var); |
99 | double x_kurtosis = (6 * sqr(d.p()) - 6 * d.p() + 1)/x_var; |
100 | assert(std::abs((mean - x_mean) / x_mean) < 0.01); |
101 | assert(std::abs((var - x_var) / x_var) < 0.01); |
102 | assert(std::abs((skew - x_skew) / x_skew) < 0.01); |
103 | assert(std::abs((kurtosis - x_kurtosis) / x_kurtosis) < 0.02); |
104 | } |
105 | |
106 | return 0; |
107 | } |
108 |