khoaguin/iris-partitions

Partitioned IRIS Datasets

This repository contains a script (dataset.py) to download the Iris dataset and split it into multiple partitions. Each partition is further divided into a public "mock" dataset and a "private" dataset.

IRIS Dataset Overview

The Iris dataset is a classic dataset in machine learning, consisting of 150 samples of iris flowers. Each sample has four features (sepal length, sepal width, petal length, and petal width) and belongs to one of three species: Iris Setosa, Iris Versicolor, or Iris Virginica.

Id	SepalLengthCm	SepalWidthCm	PetalLengthCm	PetalWidthCm	Species
88	6.3	2.3	4.4	1.3	Iris-versicolor
1	5.1	3.5	1.4	0.2	Iris-setosa
127	6.2	2.8	4.8	1.8	Iris-virginica
121	6.9	3.2	5.7	2.3	Iris-virginica
144	6.8	3.2	5.9	2.3	Iris-virginica

Generating the Partitioned Datasets

To generate the partitioned datasets, run the dataset.py script from the root of this repository (provided that you already have uv installed):

uv venv && source .venv/bin/activate
uv sync
uv run dataset.py

By default, the script will create 5 partitions. You can modify the num_partitions variable in the if __name__ == "__main__": block at the end of dataset.py to change the number of partitions generated.

Directory Structure

After running the script, you will find a directory for each partition, named iris-1, iris-2, and so on. Each of these partition directories will have the following structure:

iris-1/
├── mock/
│   ├── dataset.arrow
│   ├── dataset_info.json
│   └── state.json
├── private/
│   ├── dataset.arrow
│   ├── dataset_info.json
│   └── state.json
└── README.md

• mock/: This directory contains a small, public subset of the data for that partition (10% of the partition's data). This can be used for development and testing of models without accessing the private data.
• private/: This directory contains the larger, private subset of the data (90% of the partition's data). This data should be kept confidential.
• README.md: Each partition directory also contains its own README.md file, which provides a brief description of the Iris dataset and the mock/private split.

The data within the mock and private directories is saved in Apache Arrow format.

dataset_info:
  name: Iris Dataset
  description: A classic dataset in machine learning consisting of 150 samples of iris flowers.
  features:
    - SepalLengthCm: float
    - SepalWidthCm: float
    - PetalLengthCm: float
    - PetalWidthCm: float
  targets:
    - Species: string (Iris-setosa, Iris-versicolor, Iris-virginica)
  size: 150 samples
  source: https://archive.ics.uci.edu/ml/datasets/iris
  partitioned: true
  partitions:
    count: 5 (default, configurable in dataset.py)
    structure:
      mock:
        description: 10% of the partition's data, public.
        format: Apache Arrow
      private:
        description: 90% of the partition's data, confidential.
        format: Apache Arrow