Datasets

linfa-datasets provides a collection of commonly used datasets ready to be used in tests and examples.

The Big Picture

linfa-datasets is a crate in the linfa ecosystem, an effort to create a toolkit for classical Machine Learning implemented in pure Rust, akin to Python's scikit-learn.

Current State

Currently the following datasets are provided:

The purpose of this crate is to faciliate dataset loading and make it as simple as possible. Loaded datasets are returned as a linfa::Dataset structure with named features. The crate also includes helper functions for reading arrays from CSV files.

Additionally, this crate provides utility functions to randomly generate test datasets.

Using a dataset

To use one of the provided datasets in your project add the linfa-datasets crate to your Cargo.toml and enable the corresponding feature:

linfa-datasets = { version = "0.x", features = ["winequality"] }

You can then use the dataset in your working code:

let (train, valid) = linfa_datasets::winequality()
    .split_with_ratio(0.8);

Reading from a file

linfa-datasets is also capable of reading 2D arrays from CSV files:

use std::fs::File;
use std::io::Read;

let file = File::open("data.csv.gz").unwrap();
// Read the array from a GZipped CSV file with a header and separated by commas
let array = linfa_datasets::array_from_gz_csv(file, true, b',').unwrap();

Data generation

To generate datasets randomly, enable the generate feature on linfa-datasets. The API is in the generate module of the crate.