Use giotto-ph instead of GUDHI to compute flag persistence generators

aywagner / DIPOLE

DIstributed Persistence-Optimized Local Embeddings

MIT License

12 stars 8 forks source link

Use giotto-ph instead of GUDHI to compute flag persistence generators #5

Closed ulupo closed 2 years ago

ulupo commented 2 years ago

Being based on ripser (with additional performance improvements and parallelisation capabilities), giotto-ph's persistent homology backend is faster than GUDHI's.

ulupo commented 2 years ago

@aywagner I made this to facilitate your experimentation with this solution, in isolation from the parallelism introduced in #4. I'm not asking to merge until we have some nice benchmarks!

ulupo commented 2 years ago

Note: Since ripser_parallel can parallelize individual PH computations, we could in principle think of exploiting this when the higher-level parallelism does not saturate the available physical cores.