auto merge of #10223 : huonw/rust/gamma, r=cmr

Implements the [Gamma distribution](https://en.wikipedia.org/wiki/Gamma_distribution), using the algorithm described by Marsaglia & Tsang 2000[1]. I added tests checking that the mean and variance of this implementation is as expected for a range of values of the parameters in 5d87c00a0f (they pass locally, but obviously won't even build on Travis until this is merged).

Also, moves `std::rand::distributions` to a subfolder, and performs a minor clean-up of the benchmarking (makes the number of iterations shared by the whole `std::rand` subtree).

[1]: George Marsaglia and Wai Wan Tsang. 2000. "A Simple Method for Generating Gamma Variables" *ACM Trans. Math. Softw.* 26, 3 (September 2000), 363-372. DOI:[10.1145/358407.358414](http://doi.acm.org/10.1145/358407.358414).
This commit is contained in:
bors 2013-11-01 18:36:42 -07:00
commit c15038db08
5 changed files with 234 additions and 18 deletions

View File

@ -0,0 +1,216 @@
// Copyright 2013 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//! The Gamma distribution.
use rand::Rng;
use super::{IndependentSample, Sample, StandardNormal, Exp};
use num;
/// The Gamma distribution `Gamma(shape, scale)` distribution.
///
/// The density function of this distribution is
///
/// ```
/// f(x) = x^(k - 1) * exp(-x / θ) / (Γ(k) * θ^k)
/// ```
///
/// where `Γ` is the Gamma function, `k` is the shape and `θ` is the
/// scale and both `k` and `θ` are strictly positive.
///
/// The algorithm used is that described by Marsaglia & Tsang 2000[1],
/// falling back to directly sampling from an Exponential for `shape
/// == 1`, and using the boosting technique described in [1] for
/// `shape < 1`.
///
/// # Example
///
/// ```rust
/// use std::rand;
/// use std::rand::distributions::{IndependentSample, Gamma};
///
/// fn main() {
/// let gamma = Gamma::new(2.0, 5.0);
/// let v = gamma.ind_sample(rand::task_rng());
/// println!("{} is from a Gamma(2, 5) distribution", v);
/// }
/// ```
///
/// [1]: George Marsaglia and Wai Wan Tsang. 2000. "A Simple Method
/// for Generating Gamma Variables" *ACM Trans. Math. Softw.* 26, 3
/// (September 2000),
/// 363-372. DOI:[10.1145/358407.358414](http://doi.acm.org/10.1145/358407.358414)
pub enum Gamma {
priv Large(GammaLargeShape),
priv One(Exp),
priv Small(GammaSmallShape)
}
// These two helpers could be made public, but saving the
// match-on-Gamma-enum branch from using them directly (e.g. if one
// knows that the shape is always > 1) doesn't appear to be much
// faster.
/// Gamma distribution where the shape parameter is less than 1.
///
/// Note, samples from this require a compulsory floating-point `pow`
/// call, which makes it significantly slower than sampling from a
/// gamma distribution where the shape parameter is greater than or
/// equal to 1.
///
/// See `Gamma` for sampling from a Gamma distribution with general
/// shape parameters.
struct GammaSmallShape {
inv_shape: f64,
large_shape: GammaLargeShape
}
/// Gamma distribution where the shape parameter is larger than 1.
///
/// See `Gamma` for sampling from a Gamma distribution with general
/// shape parameters.
struct GammaLargeShape {
shape: f64,
scale: f64,
c: f64,
d: f64
}
impl Gamma {
/// Construct an object representing the `Gamma(shape, scale)`
/// distribution.
///
/// Fails if `shape <= 0` or `scale <= 0`.
pub fn new(shape: f64, scale: f64) -> Gamma {
assert!(shape > 0.0, "Gamma::new called with shape <= 0");
assert!(scale > 0.0, "Gamma::new called with scale <= 0");
match shape {
1.0 => One(Exp::new(1.0 / scale)),
0.0 .. 1.0 => Small(GammaSmallShape::new_raw(shape, scale)),
_ => Large(GammaLargeShape::new_raw(shape, scale))
}
}
}
impl GammaSmallShape {
fn new_raw(shape: f64, scale: f64) -> GammaSmallShape {
GammaSmallShape {
inv_shape: 1. / shape,
large_shape: GammaLargeShape::new_raw(shape + 1.0, scale)
}
}
}
impl GammaLargeShape {
fn new_raw(shape: f64, scale: f64) -> GammaLargeShape {
let d = shape - 1. / 3.;
GammaLargeShape {
shape: shape,
scale: scale,
c: 1. / num::sqrt(9. * d),
d: d
}
}
}
impl Sample<f64> for Gamma {
fn sample<R: Rng>(&mut self, rng: &mut R) -> f64 { self.ind_sample(rng) }
}
impl Sample<f64> for GammaSmallShape {
fn sample<R: Rng>(&mut self, rng: &mut R) -> f64 { self.ind_sample(rng) }
}
impl Sample<f64> for GammaLargeShape {
fn sample<R: Rng>(&mut self, rng: &mut R) -> f64 { self.ind_sample(rng) }
}
impl IndependentSample<f64> for Gamma {
fn ind_sample<R: Rng>(&self, rng: &mut R) -> f64 {
match *self {
Small(ref g) => g.ind_sample(rng),
One(ref g) => g.ind_sample(rng),
Large(ref g) => g.ind_sample(rng),
}
}
}
impl IndependentSample<f64> for GammaSmallShape {
fn ind_sample<R: Rng>(&self, rng: &mut R) -> f64 {
// Need (0, 1) here.
let mut u = rng.gen::<f64>();
while u == 0. {
u = rng.gen();
}
self.large_shape.ind_sample(rng) * num::pow(u, self.inv_shape)
}
}
impl IndependentSample<f64> for GammaLargeShape {
fn ind_sample<R: Rng>(&self, rng: &mut R) -> f64 {
loop {
let x = *rng.gen::<StandardNormal>();
let v_cbrt = 1.0 + self.c * x;
if v_cbrt <= 0.0 { // a^3 <= 0 iff a <= 0
continue
}
let v = v_cbrt * v_cbrt * v_cbrt;
// Need (0, 1) here, not [0, 1). This would be faster if
// we were generating an f64 in (0, 1) directly.
let mut u = rng.gen::<f64>();
while u == 0.0 {
u = rng.gen();
}
let x_sqr = x * x;
if u < 1.0 - 0.0331 * x_sqr * x_sqr ||
num::ln(u) < 0.5 * x_sqr + self.d * (1.0 - v + num::ln(v)) {
return self.d * v * self.scale
}
}
}
}
#[cfg(test)]
mod bench {
use super::*;
use mem::size_of;
use rand::distributions::IndependentSample;
use rand::{StdRng, RAND_BENCH_N};
use extra::test::BenchHarness;
use iter::range;
use option::{Some, None};
#[bench]
fn bench_gamma_large_shape(bh: &mut BenchHarness) {
let gamma = Gamma::new(10., 1.0);
let mut rng = StdRng::new();
do bh.iter {
for _ in range(0, RAND_BENCH_N) {
gamma.ind_sample(&mut rng);
}
}
bh.bytes = size_of::<f64>() as u64 * RAND_BENCH_N;
}
#[bench]
fn bench_gamma_small_shape(bh: &mut BenchHarness) {
let gamma = Gamma::new(0.1, 1.0);
let mut rng = StdRng::new();
do bh.iter {
for _ in range(0, RAND_BENCH_N) {
gamma.ind_sample(&mut rng);
}
}
bh.bytes = size_of::<f64>() as u64 * RAND_BENCH_N;
}
}

View File

@ -27,8 +27,10 @@ use rand::{Rng,Rand};
use clone::Clone;
pub use self::range::Range;
pub use self::gamma::Gamma;
pub mod range;
pub mod gamma;
/// Types that can be used to create a random instance of `Support`.
pub trait Sample<Support> {
@ -554,25 +556,23 @@ mod tests {
#[cfg(test)]
mod bench {
use extra::test::BenchHarness;
use rand::*;
use rand::{XorShiftRng, RAND_BENCH_N};
use super::*;
use iter::range;
use option::{Some, None};
use mem::size_of;
static N: u64 = 100;
#[bench]
fn rand_normal(bh: &mut BenchHarness) {
let mut rng = XorShiftRng::new();
let mut normal = Normal::new(-2.71828, 3.14159);
do bh.iter {
for _ in range(0, N) {
for _ in range(0, RAND_BENCH_N) {
normal.sample(&mut rng);
}
}
bh.bytes = size_of::<f64>() as u64 * N;
bh.bytes = size_of::<f64>() as u64 * RAND_BENCH_N;
}
#[bench]
fn rand_exp(bh: &mut BenchHarness) {
@ -580,10 +580,10 @@ mod bench {
let mut exp = Exp::new(2.71828 * 3.14159);
do bh.iter {
for _ in range(0, N) {
for _ in range(0, RAND_BENCH_N) {
exp.sample(&mut rng);
}
}
bh.bytes = size_of::<f64>() as u64 * N;
bh.bytes = size_of::<f64>() as u64 * RAND_BENCH_N;
}
}

View File

@ -833,58 +833,58 @@ mod test {
}
}
static RAND_BENCH_N: u64 = 100;
#[cfg(test)]
mod bench {
use extra::test::BenchHarness;
use rand::*;
use rand::{XorShiftRng, StdRng, IsaacRng, Isaac64Rng, Rng, RAND_BENCH_N};
use mem::size_of;
use iter::range;
use option::{Some, None};
static N: u64 = 100;
#[bench]
fn rand_xorshift(bh: &mut BenchHarness) {
let mut rng = XorShiftRng::new();
do bh.iter {
for _ in range(0, N) {
for _ in range(0, RAND_BENCH_N) {
rng.gen::<uint>();
}
}
bh.bytes = size_of::<uint>() as u64 * N;
bh.bytes = size_of::<uint>() as u64 * RAND_BENCH_N;
}
#[bench]
fn rand_isaac(bh: &mut BenchHarness) {
let mut rng = IsaacRng::new();
do bh.iter {
for _ in range(0, N) {
for _ in range(0, RAND_BENCH_N) {
rng.gen::<uint>();
}
}
bh.bytes = size_of::<uint>() as u64 * N;
bh.bytes = size_of::<uint>() as u64 * RAND_BENCH_N;
}
#[bench]
fn rand_isaac64(bh: &mut BenchHarness) {
let mut rng = Isaac64Rng::new();
do bh.iter {
for _ in range(0, N) {
for _ in range(0, RAND_BENCH_N) {
rng.gen::<uint>();
}
}
bh.bytes = size_of::<uint>() as u64 * N;
bh.bytes = size_of::<uint>() as u64 * RAND_BENCH_N;
}
#[bench]
fn rand_std(bh: &mut BenchHarness) {
let mut rng = StdRng::new();
do bh.iter {
for _ in range(0, N) {
for _ in range(0, RAND_BENCH_N) {
rng.gen::<uint>();
}
}
bh.bytes = size_of::<uint>() as u64 * N;
bh.bytes = size_of::<uint>() as u64 * RAND_BENCH_N;
}
#[bench]