bayes 0.0.1

Composable abstractions to build probabilistic models and inference algorithms
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162

use nalgebra::*;
use bayes::gsl::randist::*;
use bayes::gsl::vector_double::*;
use bayes::gsl::matrix_double::*;
use bayes::gsl::utils::*;
use bayes::distr::*;

const EPS : f64 = 10E-8;

fn unit_interval_seq(n : usize) -> DMatrix<f64> {
    DMatrix::from_fn(n, 1, |i,j| (i+1) as f64 * (1. / n as f64) )
}

fn count_seq(n : usize) -> DMatrix<f64> {
    DMatrix::from_fn(n, 1, |i,j| (i+1) as f64)
}

#[test]
fn bernoulli() {
    for y in [0.0, 1.0].iter() {
        for p in (1..100).map(|p| 0.01 * p as f64) {
            let gsl_p = unsafe{ gsl_ran_bernoulli_pdf(*y as u32, p) };
            let mut b = Bernoulli::new(1, Some(0.5));
            let pv = DVector::from_element(1,p);
            b.set_parameter(pv.rows(0,1), false);
            let ym = DMatrix::from_element(1,1,*y);
            assert!( (gsl_p - b.prob(ym.rows(0,1))).abs() < EPS);
        }
    }
}

#[test]
fn poisson() {
    let lambda = unit_interval_seq(100);
    let count = count_seq(100);
    unsafe {
        for i in 0..100 {
            for l in lambda.iter() {
                let gsl_prob = gsl_ran_poisson_pdf(count[(i,0)] as u32, *l);
                let poiss = Poisson::new(1, Some(*l));
                let bayes_prob = poiss.prob(count.slice((i,0), (1,1)));
                assert!( (gsl_prob -  bayes_prob).abs() < EPS);
            }
        }
    }
}

#[test]
fn beta() {
    let theta = unit_interval_seq(100);
    let count = count_seq(10);
    unsafe {
        for i in 0..100 {
            for (a, b) in count.iter().zip(count.iter()) {
                let gsl_prob = gsl_ran_beta_pdf(theta[(i,0)], *a, *b);
                let beta = Beta::new(*a as usize, *b as usize);
                let bayes_prob = beta.prob(theta.slice((i,0), (1,1)));
                //println!("Gsl prob: {}", gsl_prob);
                //println!("Bayes prob: {}", bayes_prob);
                //println!("a: {}; b: {}; theta: {}", a, b, theta[(i,0)]);
                assert!( (gsl_prob -  bayes_prob).abs() < EPS);
            }
        }
    }
}

#[test]
fn gamma() {
    let theta = unit_interval_seq(100);
    let count = count_seq(10);
    unsafe {
        for i in 0..100 {
            for (a, b) in count.iter().zip(count.iter()) {
                // GSL follows the shape/scale parametrization;
                // bayes follows the shape/inv-scale parametrization
                let gsl_prob = gsl_ran_gamma_pdf(theta[(i,0)], *a, 1. / *b);
                let gamma = Gamma::new(*a, *b);
                let bayes_prob = gamma.prob(theta.slice((i,0), (1,1)));
                assert!( (gsl_prob -  bayes_prob).abs() < EPS);
            }
        }
    }
}

#[test]
fn multinormal() {
    let mu = DVector::from_element(5, 0.0);
    let mut sigma = DMatrix::from_element(5, 5, 0.0);
    sigma.set_diagonal(&DVector::from_element(5, 1.));
    let mn : MultiNormal = MultiNormal::new(mu.clone(), sigma.clone());
    let lu = Cholesky::new(sigma).unwrap();
    let lower = lu.l();
    println!("{}", lower);
    unsafe {
        let lower_gsl : gsl_matrix = lower.into();
        //let mu : [f64; 5] = [0., 0., 0., 0., 0.];
        let ws_orig = DVector::<f64>::zeros(5);
        let mut ws : gsl_vector = ws_orig.into();
        let mut gsl_prob : f64 = 0.0;

        let mu_vec : gsl_vector = mu.clone().into();
        let x_vec : gsl_vector = mu.clone().into();
        let ans = gsl_ran_multivariate_gaussian_pdf(
            &x_vec as *const _,
            &mu_vec as *const _,
            &lower_gsl as *const _,
            &mut gsl_prob as *mut _,
            &mut ws as *mut _
        );
        if ans != 0 {
            panic!("Error calculating multivariate density");
        }
        let x = DMatrix::<f64>::from_iterator(1, 5, mu.iter().map(|x| *x));
        // println!("Prob: {}", gsl_prob);
        assert!((gsl_prob - mn.prob(x.slice((0,0), (x.nrows(), x.ncols())))).abs() < EPS);
    }
}

#[test]
fn normal() {
    let mu = unit_interval_seq(100);
    let values = mu.clone();
    unsafe {
        //for m in mu.iter() {
        for (i, y) in values.iter().enumerate() {
            // GSL parametrizes gaussians by the standard deviation;
            // bayes by the variance.
            let gsl_prob = gsl_ran_gaussian_pdf(*y, (1.).sqrt());
            let norm = Normal::new(1, Some(0.), Some(1.));
            let bayes_prob = norm.prob(values.slice((i,0), (1,1)));
            // println!("Gsl prob: {} (evaluated at {})", gsl_prob, *y);
            // println!("Bayes prob: {} (evaluated at {})", bayes_prob, *y);
            assert!( (gsl_prob -  bayes_prob).abs() < EPS);
        }
        //}
    }
    /*let n = Normal::new(1, Some(0.), Some(1.));
    let ym = DMatrix::from_element(1, 1, 0.0);
    unsafe {
        assert!( (gsl_ran_gaussian_pdf(0., 1.) - n.prob(ym.rows(0,1))).abs()  < EPS );
    }*/
}

#[test]
fn categorical() {
    // Obs: Categorical = Multinomial when n=1. Also note that bayes parametrizes
    // categorical with k-1 classes.
    let c : Categorical = Categorical::new(4, Some(&[0.2, 0.2, 0.2, 0.2]));
    let probs : [f64; 5] = [0.2,0.2,0.2,0.2,0.2];
    let outcome : [u32; 5] = [1, 0, 0, 0, 0];
    unsafe {
        let gsl_cat_p = gsl_ran_multinomial_pdf(5, &probs[0] as *const _, &outcome[0] as *const _);
        let vprobs = DMatrix::from_iterator(1, 4, outcome[0..4].iter().map(|o| *o as f64));
        let cat_p = c.prob(vprobs.slice((0, 0), (1, 4)));
        // println!("Categorical: {} {}", gsl_cat_p, cat_p);
        // println!("{}", gsl_cat_p - cat_p);
        assert!((gsl_cat_p - cat_p).abs() < EPS);
    }
}