This repository contains code to recreate the microenvironment classification results from the paper Flagifying the Dowker Complex.

---

Requirements

Required dependencies are specified in pyproject.toml. The dependencies dowker-complex and dowker-rips-complex can be installed either from source from here and here, respectively, or via pip by running e.g. pip install -U dowker-complex and pip install -U dowker-rips-complex, respectively.

---

Reproducing results

To reproduce the results, run the command python main.py <complex> <n_repeats> <overwrite> <verbose>, where

• <complex> must be one of dowker_rips and dowker, and indicates whether to use the Dowker or the Dowker-Rips complex;
• <n_repeats> must be a positive integer, and indicates the number of times training of the SVM is repeated;
• <overwrite> must be one of True and False, and indicates whether results existing on disk should be overwritten or not; and
• <verbose> must be a non-negative integer, and indicates the level of verbosity during execution of the script.

For example, to reproduce the results of the paper using the Dowker-Rips complex, run python main.py dowker_rips 10 False 1

Executing main.py as above will create a directory named outfiles that contains the processed point cloud files, the persistence data, the persistence images as well as an array containing the accuracies of each of the <n_repeats> many SVMs trained and evaluated in the process.

---

For users of uv

If uv is installed, required dependencies can be installed by running uv pip install -r pyproject.toml. The environment specified in uv.lock can be recreated by running uv sync. The dependencies dowker-complex and dowker-rips-complex can be installed by running e.g. uv add dowker-complex and uv add dowker-rips-complex, respectively.

To reproduce the results from the paper, run uv run main.py <complex> <n_repeats> <overwrite> <verbose>, with parameters specified as above.