NoLiES: The non-linear embeddings surveyor

This repository contains a Jupyter Notebook implementation of the method presented in "Attribute-based Explanation of Non-Linear Embeddings of High-Dimensional Data" by Jan-Tobias Sohns, Michaela Schmitt, Fabian Jirasek, Hans Hasse, and Heike Leitte, submitted to IEEE VIS 2021.

Requirements

• pandas
• bokeh
• pyviz::panel
• scikit-learn
• conda-forge::umap-learn
• shapely

Datasets used

• Iris (150x5)
• Wine (178x14)
• OECD Better Life (41x11)
• Penguins (344x7)
• Covertype (3500x11)

Working with your own data

Download the repository and create an environment with the dependencies:

git clone https://github.com/Kriti1106/Attribute-based-Explanations-of-non-linear-Embeddings-of-high-dimensional-data.git
conda env create -f nolies.yml
conda activate nolies

Make a copy of the temple jupyter notebook:

cp template.ipynb my_data.ipynb

Update the notebook to load your data. Open the notebook with jupyter lab or jupyter notebook and edit the section Load data. Important parameters are grouped in the section Parameters and Preprocessing:

jupyter lab

Acknowledgements

This is a clone code for the research paper- Attribute-based Explanation of Non-Linear Embeddings of High-Dimensional Data by Jan-Tobias Sohns, Michaela Schmitt, Fabian Jirasek, Hans Hasse, and Heike Leitte.