Topologically Regularized Data Embeddings

The code in this repository is accompanying the manuscript "Topologically Regularized Data Embeddings".

Installation

Python

We provide a conda_env.yml file listing the required packages. You can install by creating a new conda environment

conda env create -f topembedding/conda_env.yml -p ./topo_env
conda activate topo_env/

(1) Install TopologyLayer

pip install git+https://github.com/bruel-gabrielsson/TopologyLayer.git

(2) Install Aleph by following the instructions on their GitHub.

git clone https://github.com/Pseudomanifold/Aleph.git
cd Aleph && mkdir build && cd build && cmake ../ && make && make test

cd bindings/python/aleph
python setup.py install

(3) Optional: Install DioDe to exerun the pseudotime analysis in the CellCyle notebook.

R

To execute the R scripts in the Scripts folder you need:

TDA
ggplot2
latex2exp
gridExtra

Repository Content

(Back to top)

Code

Try out the example config files in /Code/config by providing one of them to main.py.

python main.py Code/config/synthetic_random_optimize.yaml

Note that 'cell_cycle.yaml' takes about 5 Minutes to run. Upon completion the final embeddings will be shown.

Data

Synthetic data are generated using Data/datasets.py
Cell trajectory (source): included as Data/CellCycle.rds
Cell bifurcation (source): included as Data/CellBifurcation.rds
Karate[1]: partially (graph) loaded from networkx and partially (weights) from Data/Karate.txt
Harry Potter (source): included in Data/HarryPotter

Experiments

In this folder you find one notebook for every dataset to reproduce the results in the experiments section. You can open the .hml version to see the code with output.

Scripts

Folder "Scripts": contains two Jupyter notebooks and two R scripts to reproduce the visualizations of Section (2) in the manuscript.

Acknowledgements