Python realization for Vectorized Persistence Block vectorization

The goal is to implement TDAvec library in python using R as a source (see GitHub repository https://github.com/alexey-luchinsky/TDAvec)

How to run:

1. create python environment and install all packages from requirements.txt
2. run python setup.py build_ext --inplace from the python/ directory to compile pyx file
3. launch R/TDA_vec.Rmd to get R results
4. run through test_TDAvectorizer.ipynb notebook to see python results and save figures
5. launch docker
6. Run source make_pdf.sh from paper directory to create pdf

File structure:

• python/: python files live here
  • TDAvec.pyx: Cython version of the package. Note that you need to recompile it after each change
  • TDAvectorizer.py: TDAvectorizer class
  • test_TDAvectorizer.ipynb: Jupyter notebook, testing TDAvectorizer
  • setup.py: Cython make file. See TDAvec_pyx_test.ipynb for instructions
  • unit_tests.py: unit tests. You should run python -m unittest unit_test.py from the python/ directory to run them.
• R/: R files live here
  • TDA_vec.Rmd: extracting test results for different functions
• paper/: paper files live here

TODO

• reread/rewrite paper
• unit tests for TDAvectorizer
• VPB: PD -> D
• VPB, PI: more simple tests, default xSeq, ySeq

Done tasks

• DONE: Aug 10, 2024, 09:55: Create remote repository:
• DONE: Aug 10, 2024, 10:05: Start the JOSS paper template
• DONE: Aug 10, 2024, 16:39: Find other python libraries for vectorization
• DONE: Aug 11, 2024, 16:39: Incorporate my into ML workflow
• DONE: Aug 15, 2024, 07:38: Redo simulations:
• DONE: Aug 15, 2024, 07:38: Correct table:
• DONE: Aug 15, 2024, 08:50: Add dim0+dim1 set of predictors:
• DONE: Aug 15, 2024, 11:54: Rewrite paper:
• DONE: Aug 15, 2024, 13:00: Remove outliers:
• DONE: Aug 15, 2024, 16:53: Send to Umar:
• DONE: Aug 16, 2024, 07:56: FDA in R
• DONE: Aug 16, 2024, 11:40: FDA in Python
• DONE: Aug 16, 2024, 11:40: FDA: compare to R
• DONE: Aug 16, 2024, 11:40: add FDA tables and plots

MISC

Other python libraries:

• skit-tda: https://docs.scikit-tda.org/en/latest/libraries.html
• Giotto-TDA: https://giotto-ai.github.io/gtda-docs/0.5.1/index.html
• Geometry Understanding in Higher Dimensions: https://github.com/GUDHI, https://gudhi.inria.fr/introduction/
• Dionysus 2: https://www.mrzv.org/software/dionysus2/

The following functions were translated from R and checked against R results:

• computeVPB: A Vector Summary of the Persistence Block
• computePL: A Vector Summary of the Persistence Landscape Function
• computePS: A Vector Summary of the Persistence Silhouette
• computeNL: A Vector Summary of the Normalized Life Curve
• computeVAB: A Vector Summary of the Betti Curve
• computeECC: A Vector Summary of the Euler Characteristic Curve
• computePES: A Vector Summary of the Persistent Entropy Summary Function
• computePI: A Vector Summary of the Persistence Surface