]> git.proxmox.com Git - rustc.git/blobdiff - vendor/rand-0.7.3/src/distributions/dirichlet.rs
Merge tag 'debian/1.52.1+dfsg1-1_exp2' into proxmox/buster
[rustc.git] / vendor / rand-0.7.3 / src / distributions / dirichlet.rs
diff --git a/vendor/rand-0.7.3/src/distributions/dirichlet.rs b/vendor/rand-0.7.3/src/distributions/dirichlet.rs
new file mode 100644 (file)
index 0000000..a75678a
--- /dev/null
@@ -0,0 +1,126 @@
+// Copyright 2018 Developers of the Rand project.
+// Copyright 2013 The Rust Project Developers.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// https://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+
+//! The dirichlet distribution.
+#![allow(deprecated)]
+#![allow(clippy::all)]
+
+use crate::distributions::gamma::Gamma;
+use crate::distributions::Distribution;
+use crate::Rng;
+
+/// The dirichelet distribution `Dirichlet(alpha)`.
+///
+/// The Dirichlet distribution is a family of continuous multivariate
+/// probability distributions parameterized by a vector alpha of positive reals.
+/// It is a multivariate generalization of the beta distribution.
+#[deprecated(since = "0.7.0", note = "moved to rand_distr crate")]
+#[derive(Clone, Debug)]
+pub struct Dirichlet {
+    /// Concentration parameters (alpha)
+    alpha: Vec<f64>,
+}
+
+impl Dirichlet {
+    /// Construct a new `Dirichlet` with the given alpha parameter `alpha`.
+    ///
+    /// # Panics
+    /// - if `alpha.len() < 2`
+    #[inline]
+    pub fn new<V: Into<Vec<f64>>>(alpha: V) -> Dirichlet {
+        let a = alpha.into();
+        assert!(a.len() > 1);
+        for i in 0..a.len() {
+            assert!(a[i] > 0.0);
+        }
+
+        Dirichlet { alpha: a }
+    }
+
+    /// Construct a new `Dirichlet` with the given shape parameter `alpha` and `size`.
+    ///
+    /// # Panics
+    /// - if `alpha <= 0.0`
+    /// - if `size < 2`
+    #[inline]
+    pub fn new_with_param(alpha: f64, size: usize) -> Dirichlet {
+        assert!(alpha > 0.0);
+        assert!(size > 1);
+        Dirichlet {
+            alpha: vec![alpha; size],
+        }
+    }
+}
+
+impl Distribution<Vec<f64>> for Dirichlet {
+    fn sample<R: Rng + ?Sized>(&self, rng: &mut R) -> Vec<f64> {
+        let n = self.alpha.len();
+        let mut samples = vec![0.0f64; n];
+        let mut sum = 0.0f64;
+
+        for i in 0..n {
+            let g = Gamma::new(self.alpha[i], 1.0);
+            samples[i] = g.sample(rng);
+            sum += samples[i];
+        }
+        let invacc = 1.0 / sum;
+        for i in 0..n {
+            samples[i] *= invacc;
+        }
+        samples
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use super::Dirichlet;
+    use crate::distributions::Distribution;
+
+    #[test]
+    fn test_dirichlet() {
+        let d = Dirichlet::new(vec![1.0, 2.0, 3.0]);
+        let mut rng = crate::test::rng(221);
+        let samples = d.sample(&mut rng);
+        let _: Vec<f64> = samples
+            .into_iter()
+            .map(|x| {
+                assert!(x > 0.0);
+                x
+            })
+            .collect();
+    }
+
+    #[test]
+    fn test_dirichlet_with_param() {
+        let alpha = 0.5f64;
+        let size = 2;
+        let d = Dirichlet::new_with_param(alpha, size);
+        let mut rng = crate::test::rng(221);
+        let samples = d.sample(&mut rng);
+        let _: Vec<f64> = samples
+            .into_iter()
+            .map(|x| {
+                assert!(x > 0.0);
+                x
+            })
+            .collect();
+    }
+
+    #[test]
+    #[should_panic]
+    fn test_dirichlet_invalid_length() {
+        Dirichlet::new_with_param(0.5f64, 1);
+    }
+
+    #[test]
+    #[should_panic]
+    fn test_dirichlet_invalid_alpha() {
+        Dirichlet::new_with_param(0.0f64, 2);
+    }
+}