Megaman benchmarks about megaman HOT 6 CLOSED

We did some rough benchmarking in an earlier version of megaman here: https://github.com/mmp2/megaman/tree/1c8e9010d5e03e8e0c257885209e866ad30771c5/benchmarks

Ultimately, however, when the data set becomes larger than say 100,000 points Spectral Embedding vastly out-scales the other methods. Since our data sets were all significantly larger than 100K we dedicated most of the time to benchmarking and evaluating Spectral Embedding as in the JMLR paper.

from megaman.

from megaman.

Sorry--my explanation wasn't clear.

I'm using megaman for dimensionality reduction, prior to clustering.

I'm looking at the tutorial notebook, which provides an ordered list of how well the algorithms scale from best to worst, but I was hoping for more detailed benchmarks on the embedding algorithms. Where can I find those?

Thank you,

Andrew

from megaman.

Somehow I managed to completely miss the folder clearly marked "Benchmarks".

Very helpful, thank you.

from megaman.

You didn't miss the folder, that's from an old commit -- it doesn't exist in live anymore :)

If you're interested, it would be fantastic to do a more formal benchmarking of the methods!

from megaman.

I'll put this on my to-do list, but unfortunately it's cluttered with so many other things already.

from megaman.