Source code for efficiently computing Mapper at multiple scales
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore
README.md
WV6_Data_R_v_2016_01_01.rds
diab_bench_ecc.rds
diab_bench_kde.rds
multiscale_benchmarks.Rmd
multiscale_benchmarks.nb.html
torus_bench_le.rds
torus_bench_umap.rds
wvs_bench_ica.rds
wvs_bench_pca.rds

README.md

Indexed Mapper

Source code for the method that is described here. The method and results are ongoing research efforts! The actual source code is part of the Mapper R package. Specifically, see the files:

  • src/MultiScale.xx
  • src/GridIndex.xx
  • R/MapperRef.R
  • R/multiscale.R

To reproduce

To reproduce the experiments reported in the paper, do the following:

  1. Install the Mapper package
  2. Install the dependencies
    • R packages needed: data.table, ks, fastICA, RANN, reticulate, parallelDist
    • Python package needed: numpy, scikit-learn, umap-learn
  3. Open the notebook multiscale_benchmarks.Rmd in Rstudio and click 'Run all'
    • The figures require ggplot2 and gridExtra to be installed

I use reticulate+miniconda to manage and load the python dependencies, and the path needs to be set to an appropriate conda environment with the above dependencies to run.