How can I write an arbitrary discrete distribution in Julia?

Question

For example, a distribution that returns 1.0 with probability 0.3 and returns 1.1 with probability 0.7. Thank you.

Answer 1

Maybe you do not need a full blown distribution type but just sampling from such a distribution is enough for you?

If this is the case, then the simplest way to do it is:

using StatsBase  # corrected a typo here

values = [1.0, 1.1]
probabilities = [0.3, 0.7]
w = Weights(probabilities)
sample(values, w) # sampling

If you actually want to use a distribution the closest thing you can get now is:

using Distributions

values = [1.0, 1.1]
probabilities = [0.3, 0.7]

d = Categorical(probabilities)
values[rand(d)] # sampling

but it will be a bit slower.

If you want to define your own distribution following Distributions package type system, the simplest way is to take this code https://github.com/JuliaStats/Distributions.jl/blob/master/src/univariate/discrete/categorical.jl and modify it according to your needs (but this would be a significant effort I would say).

Answer 2

There isn't yet a built-in way to do it, but you could look at https://github.com/JuliaStats/Distributions.jl/pull/634).