Comparing R packages for calculation of persistent homology

Eashwar V. Somasundaram, Shael E. Brown, Adam Litzler, Jacob G. Scott, Raoul R. Wadhwa

Directory Structure

• Functions.R: R script that collects benchmarking data. The workhorse function is TDA_bench.
• Cluster_data: contains benchmarking data (in CSV format) from the high-performance computing node at the Cleveland Clinic's Lerner Research Institute.
• Mac_data: contains benchmarking data (in CSV format) from a local machine.
• Cluster_Scripts: R and Slurm scripts to collect benchmark data (stored under the Cluster_data directory).
• Figures: contains R scripts and the PNG files they generate (see subdirectories as well). A subset of the PNG files were included the final report.
• Supplemental_Regression: Supplement section code and data, most of which was used to determine if functions grow on the order of a power function or an exponential function.
• tdabench-paper: contains the final report (in R markdown).

Dependencies

The following R packages (see report for version details) were used in this report.

• readr: read rectangular data.
• ggplot2: visualize data.
• scatterplot3d: visualize data.
• recexcavAAR: visualize data.
• magick: visualize data.
• bench: measuring R code execution time.
• pryr: measure R object size.
• TDA: calculating persistent homology.
• TDAstats: calculating persistent homology.

Instructions to Reproduce Data

Video explanations can be found here. Videos 1-4 describe the background and theory for persistent homology. Video 5 explains how to reproduce the figures and analysis using our data. Video 6 explains how to generate the data by performing the benchmarking on your machine.

Summarized Steps:

1. Clone whole repository to your local machine.
2. To reproduce figures, open the Final_Figures_Reproducible.R script. Running the whole script reproduces the figures. Detailed comments are provided in the script to explain each step.
3. To reproduce data for execution time, open the "Benchmark_Runtime.R" script. Detailed comments are provided in the script that explain how to select which point clouds you would like to benchmark. Running the whole script will generate a csv file that the script will append benchmark new data to as the script finishes each benchmark.
4. To reproduce data for object size, perform the same steps except in the "Benchmark_Object_Size.R" script.