JuliaML / LIBSVM.jl

LIBSVM bindings for Julia

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Read datasets from libsvm

matthiasbe opened this issue · comments

Is it possible to read datasets from libsvm webpage in Julia ?

Here is an example of data (text file)

751.0 5:1 8:0.0 10:0.652913 11:0.701456 12:0.682039 13:0.677184 14:0.599515 15:0.589806 16:0.631068
703.0 6:1 8:0.0 10:0.696601941748 11:0.652913 12:0.701456 13:0.682039 14:0.677184 15:0.599515 16:0.589806
677.0 8:0.0 10:0.580097087379 11:0.696601941748 12:0.652913 13:0.701456 14:0.682039 15:0.677184 16:0.599515
718.0 1:1 8:0.0 10:0.516990291262 11:0.580097087379 12:0.696601941748 13:0.652913 14:0.701456 15:0.682039 16:0.677184

I couldn't find a direct way using CSV.jl or LIBSVM.jl, did I miss something ?

Thank you
Matthias

I don't know. I'm pretty sure LIBSVM.jl does not provide that functionality.

Some Julia grabbers for ML/stats data sets include RDatasets.jl, OpenML.jl. And there is MLDatasets, which is focused on deep learning applications.

This functionality is not provided.

You could write a parser yourself but it might be easier to check out the packages @ablaom mentioned (I would particularly recommend RDatasets.jl) or download datasets from their original source as listed on the libsvm page (e.g. from the UCI ML Repository) where they tend to use more common formats.