WeightedSampling.jl

Sum-heap based sampler for efficiently sampling without replacement.

• WeightedSampler(w): Construct a sampler from weight vector w.
• sample(::WeightedSampler, k; ordered=true): Sample k times without replacement. If ordered, sort the output.
• weight(::WeightedSampler, j): weight of the j-th index.
• adjust_weight!(::WeightedSampler, j, w): Set weight of j-th index to w.

Usage

julia> using WeightedSampling

julia> v = randn(10).^2;

julia> ws = WeightedSampler(v);

julia> sample(ws, 5, ordered=true)
5-element Vector{Int64}:
 2
 4
 5
 6
 8

julia> weight(ws, 3) == v[3]
true

julia> adjust_weight!(ws, 3, 0.0)