1 // Copyright 2013 The Rust Project Developers. See the COPYRIGHT
2 // file at the top-level directory of this distribution and at
3 // http://rust-lang.org/COPYRIGHT.
5 // Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
6 // http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
7 // <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
8 // option. This file may not be copied, modified, or distributed
9 // except according to those terms.
11 //! The normal and derived distributions.
13 use core::kinds::Copy;
16 use {Rng, Rand, Open01};
17 use distributions::{ziggurat, ziggurat_tables, Sample, IndependentSample};
19 /// A wrapper around an `f64` to generate N(0, 1) random numbers
20 /// (a.k.a. a standard normal, or Gaussian).
22 /// See `Normal` for the general normal distribution. That this has to
23 /// be unwrapped before use as an `f64` (using either `*` or
24 /// `mem::transmute` is safe).
26 /// Implemented via the ZIGNOR variant[1] of the Ziggurat method.
28 /// [1]: Jurgen A. Doornik (2005). [*An Improved Ziggurat Method to
29 /// Generate Normal Random
30 /// Samples*](http://www.doornik.com/research/ziggurat.pdf). Nuffield
32 pub struct StandardNormal(pub f64);
34 impl Copy for StandardNormal {}
36 impl Rand for StandardNormal {
37 fn rand<R:Rng>(rng: &mut R) -> StandardNormal {
39 fn pdf(x: f64) -> f64 {
43 fn zero_case<R:Rng>(rng: &mut R, u: f64) -> f64 {
44 // compute a random number in the tail by hand
46 // strange initial conditions, because the loop is not
47 // do-while, so the condition should be true on the first
48 // run, they get overwritten anyway (0 < 1, so these are
53 while -2.0 * y < x * x {
54 let Open01(x_) = rng.gen::<Open01<f64>>();
55 let Open01(y_) = rng.gen::<Open01<f64>>();
57 x = x_.ln() / ziggurat_tables::ZIG_NORM_R;
61 if u < 0.0 { x - ziggurat_tables::ZIG_NORM_R } else { ziggurat_tables::ZIG_NORM_R - x }
64 StandardNormal(ziggurat(
66 true, // this is symmetric
67 &ziggurat_tables::ZIG_NORM_X,
68 &ziggurat_tables::ZIG_NORM_F,
73 /// The normal distribution `N(mean, std_dev**2)`.
75 /// This uses the ZIGNOR variant of the Ziggurat method, see
76 /// `StandardNormal` for more details.
82 /// use std::rand::distributions::{Normal, IndependentSample};
84 /// // mean 2, standard deviation 3
85 /// let normal = Normal::new(2.0, 3.0);
86 /// let v = normal.ind_sample(&mut rand::task_rng());
87 /// println!("{} is from a N(2, 9) distribution", v)
94 impl Copy for Normal {}
97 /// Construct a new `Normal` distribution with the given mean and
98 /// standard deviation.
102 /// Panics if `std_dev < 0`.
103 pub fn new(mean: f64, std_dev: f64) -> Normal {
104 assert!(std_dev >= 0.0, "Normal::new called with `std_dev` < 0");
111 impl Sample<f64> for Normal {
112 fn sample<R: Rng>(&mut self, rng: &mut R) -> f64 { self.ind_sample(rng) }
114 impl IndependentSample<f64> for Normal {
115 fn ind_sample<R: Rng>(&self, rng: &mut R) -> f64 {
116 let StandardNormal(n) = rng.gen::<StandardNormal>();
117 self.mean + self.std_dev * n
122 /// The log-normal distribution `ln N(mean, std_dev**2)`.
124 /// If `X` is log-normal distributed, then `ln(X)` is `N(mean,
125 /// std_dev**2)` distributed.
131 /// use std::rand::distributions::{LogNormal, IndependentSample};
133 /// // mean 2, standard deviation 3
134 /// let log_normal = LogNormal::new(2.0, 3.0);
135 /// let v = log_normal.ind_sample(&mut rand::task_rng());
136 /// println!("{} is from an ln N(2, 9) distribution", v)
138 pub struct LogNormal {
142 impl Copy for LogNormal {}
145 /// Construct a new `LogNormal` distribution with the given mean
146 /// and standard deviation.
150 /// Panics if `std_dev < 0`.
151 pub fn new(mean: f64, std_dev: f64) -> LogNormal {
152 assert!(std_dev >= 0.0, "LogNormal::new called with `std_dev` < 0");
153 LogNormal { norm: Normal::new(mean, std_dev) }
156 impl Sample<f64> for LogNormal {
157 fn sample<R: Rng>(&mut self, rng: &mut R) -> f64 { self.ind_sample(rng) }
159 impl IndependentSample<f64> for LogNormal {
160 fn ind_sample<R: Rng>(&self, rng: &mut R) -> f64 {
161 self.norm.ind_sample(rng).exp()
169 use distributions::{Sample, IndependentSample};
170 use super::{Normal, LogNormal};
174 let mut norm = Normal::new(10.0, 10.0);
175 let mut rng = ::test::rng();
176 for _ in range(0u, 1000) {
177 norm.sample(&mut rng);
178 norm.ind_sample(&mut rng);
183 fn test_normal_invalid_sd() {
184 Normal::new(10.0, -1.0);
189 fn test_log_normal() {
190 let mut lnorm = LogNormal::new(10.0, 10.0);
191 let mut rng = ::test::rng();
192 for _ in range(0u, 1000) {
193 lnorm.sample(&mut rng);
194 lnorm.ind_sample(&mut rng);
199 fn test_log_normal_invalid_sd() {
200 LogNormal::new(10.0, -1.0);
208 use self::test::Bencher;
209 use std::mem::size_of;
210 use distributions::{Sample};
214 fn rand_normal(b: &mut Bencher) {
215 let mut rng = ::test::weak_rng();
216 let mut normal = Normal::new(-2.71828, 3.14159);
219 for _ in range(0, ::RAND_BENCH_N) {
220 normal.sample(&mut rng);
223 b.bytes = size_of::<f64>() as u64 * ::RAND_BENCH_N;