RelationalDatasets.jl

A small library for loading and downloading relational datasets.

or: The relational-datasets Python package—but written in Julia.

add RelationalDatasets

Usage

Loading and Downloading Benchmark Datasets

RelationalDatasets.jl exports a single function: load to return train and test folds.

using RelationalDatasets

train, test = load("webkb", "v0.0.4", fold=2)
length(train.facts)
# 1344

Converting Vector-based Datasets

Binary classification when y is a vector of 0/1:

using RelationalDatasets

data, modes = RelationalDatasets.from_vector(
  [0 1 1; 1 0 2; 2 2 0],
  [0, 0, 1],
)

Regression when y is a vector of floats:

using RelationalDatasets

data, modes = RelationalDatasets.from_vector(
  [0 1 1; 1 0 2; 2 2 0],
  [1.1, 0.9, 2.5],
)

Install

From Registry

add RelationalDatasets

From GitHub

using Pkg
Pkg.add(url="https://github.com/srlearn/RelationalDatasets.jl")