There are many ways to do one hot encoding in Julia. I want to list down some ways Method Comment Roll your own It’s not too hard Flux.onehot & Flux.onehotbatch pretty heavy to depend on Flux so normally avoided unless u r using Flux anyway DataConvenience.onehot can be applied directly…

julia> using StatsBase julia> x = ["a", "b", "c", "a", "b", "d"] 6-element Vector{String}: "a" "b" "c" "a" "b" "d" julia> indicatormat(x) 4×6 Matrix{Bool}: 1 0 0 1 0 0 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 1 but normally I do your option 1 :smiley: : julia> df = DataFra…

Yeah. Nice. I think the issue with DataConvenience.jl is that it’s not designed for production use. So missing categories at production time is harder to handle.

StatsModels.jl DummyCoding https://juliastats.org/StatsModels.jl/stable/contrasts/#StatsModels.DummyCoding

[image] Mattriks: unique(x) .== permutedims(x) Mind Blown. julia> x = [1, 2, 1, 3, 2]; julia> unique(x) .== permutedims(x) 3×5 BitMatrix: 1 0 1 0 0 0 1 0 0 1 0 0 0 1 0 That is so elegant.

Brilliant! Not sure what permutedims does in case x is multidimensional, so maybe the following is slightly more general: unique(x) .== reshape(x, (1, size(x)…))

how do you one-hot encode a multidimensional x? It’s so nice our broadcast would do outer automatically :wink:

Not in a million years would I have thought of that. :joy: Awesome :+1:t2:

Flux’s OneHotArray has gotten a lot of work recently and is pretty robust. It’s also self-contained , so we could theoretically pull it into a separate package if there’s enough interest.

https://github.com/cossio/OneHot.jl

All the ways to do one-hot encoding

General Usage

Mattriks July 17, 2021, 11:38am 4

unique(x) .== permutedims(x)

Generating dummy variables from a vector of strings (one-hot encoding)

Topic		Replies	Views
Learning Julia: Writing a onehot encoder Tooling	5	1569	October 23, 2019
Flux: 1D convolutions (on genomic data) Machine Learning flux	20	2832	May 28, 2023
Coming from python this took a while to figure out New to Julia	3	356	March 17, 2024
Generating dummy variables from a vector of strings (one-hot encoding) New to Julia	9	3491	July 31, 2021
Encoding categorical variables within a matrix Machine Learning machine-learning	3	2799	December 28, 2019

All the ways to do one-hot encoding

Related topics