1 //===----------------------------------------------------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is dual licensed under the MIT and the University of Illinois Open
6 // Source Licenses. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 
10 // <random>
11 
12 // class bernoulli_distribution
13 
14 // template<class _URNG> result_type operator()(_URNG& g);
15 
16 #include <random>
17 #include <numeric>
18 #include <vector>
19 #include <cassert>
20 
21 template <class T>
22 inline
23 T
sqr(T x)24 sqr(T x)
25 {
26     return x * x;
27 }
28 
main()29 int main()
30 {
31     {
32         typedef std::bernoulli_distribution D;
33         typedef std::minstd_rand G;
34         G g;
35         D d(.75);
36         const int N = 100000;
37         std::vector<D::result_type> u;
38         for (int i = 0; i < N; ++i)
39             u.push_back(d(g));
40         double mean = std::accumulate(u.begin(), u.end(),
41                                               double(0)) / u.size();
42         double var = 0;
43         double skew = 0;
44         double kurtosis = 0;
45         for (int i = 0; i < u.size(); ++i)
46         {
47             double d = (u[i] - mean);
48             double d2 = sqr(d);
49             var += d2;
50             skew += d * d2;
51             kurtosis += d2 * d2;
52         }
53         var /= u.size();
54         double dev = std::sqrt(var);
55         skew /= u.size() * dev * var;
56         kurtosis /= u.size() * var * var;
57         kurtosis -= 3;
58         double x_mean = d.p();
59         double x_var = d.p()*(1-d.p());
60         double x_skew = (1 - 2 * d.p())/std::sqrt(x_var);
61         double x_kurtosis = (6 * sqr(d.p()) - 6 * d.p() + 1)/x_var;
62         assert(std::abs((mean - x_mean) / x_mean) < 0.01);
63         assert(std::abs((var - x_var) / x_var) < 0.01);
64         assert(std::abs((skew - x_skew) / x_skew) < 0.01);
65         assert(std::abs((kurtosis - x_kurtosis) / x_kurtosis) < 0.02);
66     }
67     {
68         typedef std::bernoulli_distribution D;
69         typedef std::minstd_rand G;
70         G g;
71         D d(.25);
72         const int N = 100000;
73         std::vector<D::result_type> u;
74         for (int i = 0; i < N; ++i)
75             u.push_back(d(g));
76         double mean = std::accumulate(u.begin(), u.end(),
77                                               double(0)) / u.size();
78         double var = 0;
79         double skew = 0;
80         double kurtosis = 0;
81         for (int i = 0; i < u.size(); ++i)
82         {
83             double d = (u[i] - mean);
84             double d2 = sqr(d);
85             var += d2;
86             skew += d * d2;
87             kurtosis += d2 * d2;
88         }
89         var /= u.size();
90         double dev = std::sqrt(var);
91         skew /= u.size() * dev * var;
92         kurtosis /= u.size() * var * var;
93         kurtosis -= 3;
94         double x_mean = d.p();
95         double x_var = d.p()*(1-d.p());
96         double x_skew = (1 - 2 * d.p())/std::sqrt(x_var);
97         double x_kurtosis = (6 * sqr(d.p()) - 6 * d.p() + 1)/x_var;
98         assert(std::abs((mean - x_mean) / x_mean) < 0.01);
99         assert(std::abs((var - x_var) / x_var) < 0.01);
100         assert(std::abs((skew - x_skew) / x_skew) < 0.01);
101         assert(std::abs((kurtosis - x_kurtosis) / x_kurtosis) < 0.02);
102     }
103 }
104