Julia package to simplify discretization operations that are routine in statistical tasks. The key idea of the package is that a lot of statistical operations can be simplified by giving first class support to Intervals from the IntervalSets.jl package.
Let ℐ the set of all intervals on ℝ. The key idea is that a discretizer acts as a function from ( subset of ) ℝ ∪ ℐ ↦ ℐ with the following properties:
- Callable as
discretizer(z)and broadcastablediscretizer.(zs). - Returns intervals:
typeof(discretizer(z)) <: AbstractInterval. - Idempotent:
discretizer(discretizer(z)) == discretizer(z).
The discretizers that are currently implemented are the following
This is a discretizer that partitions the whole real line into intervals with breakpoints at grid. For example:
julia> using StatsDiscretizations
julia> discr = RealLineDiscretizer{:open,:closed}(-2:0.1:2)
julia> discr(-5)
-Inf..-2.0 (open–closed)
julia> discr(0.05)
0.0..0.1 (open–closed)
julia> discr(0.00)
-0.1..0.0 (open–closed)