porco

Composable probability distributions.

Examples

Create simple probability distributions.

enum Coin {
    Heads,
    Tails,
}
 
impl Coin {
    fn flip() -> Distribution<Coin> {
        Distribution::uniform([Coin::Heads, Coin::Tails])
    }
}
 
let coin = Coin::flip();
assert_eq!(coin.pmf(&Coin::Heads), Probability(0.5));

Compose operations over distributions using combinators.

fn reflip_if_tails(coin: Coin) -> Distribution<Coin> {
    match coin {
        Coin::Heads => Distribution::always(Coin::Heads),
        Coin::Tails => Coin::flip(),
    }
}
 
let coin = Coin::flip().and_then(reflip_if_tails);
assert_eq!(coin.pmf(&Coin::Heads), Probability(0.75));

Compute summary statistics of random variables.

let die = Distribution::uniform([1, 2, 3, 4, 5, 6]);
let ev = die.given(|&v| v <= 4).expectation();
assert_eq!(ev, 2.5);

lib.rs:

Porco is a library for working with and composing probability distributions.

The API is inspired by the contents of Probabilistic Functional Programming in Haskell but with naming conventions that match those of Option and Result (such as Option::and_then).

enum Coin {
    Heads,
    Tails,
}

impl Coin {
    fn flip() -> Distribution<Coin> {
        Distribution::uniform(vec![Coin::Heads, Coin::Tails])
    }
}

let coin = Coin::flip();
assert_eq!(coin.pmf(&Coin::Heads), Probability(0.5));

You can compose various operations over Distributions using combinators like Distribution::map, Distribution::and_then, and Distribution::given.

fn reflip_if_tails(coin: Coin) -> Distribution<Coin> {
    match coin {
        Coin::Heads => Distribution::always(Coin::Heads),
        Coin::Tails => Coin::flip(),
    }
}

let coin = Coin::flip().and_then(reflip_if_tails);
assert_eq!(coin.pmf(&Coin::Heads), Probability(0.75));

You can also manipulate random variables and compute summary statistics.

let die = Distribution::uniform(vec![1, 2, 3, 4, 5, 6]);
let ev = die.given(|&v| v <= 4).expectation();
assert_eq!(ev, 2.5);

fn two_sided_die() -> Distribution<u8> {
    Distribution::uniform(vec![1, 2])
}

let x = two_sided_die();
let y = two_sided_die();
let sum = x.convolve(y);
assert_eq!(sum.pmf(&2), Probability(0.25));
assert_eq!(sum.pmf(&3), Probability(0.5));