search

biocore / mds-approximations

Multidimensional scaling algorithms for microbiology-ecology datasets.
6 stars 7 forks source link

Adding code to call all methods #1

Closed antgonza closed 6 years ago

antgonza commented 9 years ago

The methods are:

rob-knight commented 9 years ago

While we’re looking at these, it might be worth seeing if there are parallel libraries for svd that could be called direct from ipython as this is a hard limit on what we can do currently? I cc:ed Fernando in case he knows of any that are drop-in and highly parallel.

On Oct 19, 2015, at 12:38 PM, Antonio Gonzalez notifications@github.com wrote:

The methods are:

eigh scmds nystrom svd ssvd fsvd eigsh — Reply to this email directly or view it on GitHub https://github.com/biocore/mds-approximations/issues/1.

mortonjt commented 9 years ago

Note that eigh, eigsh and svd are already available in scipy.

ssvd is also implemented in scikit-learn

wasade commented 9 years ago

Highly recommend adding in scikit-learn to the mix. Decomposition methods here

wasade commented 9 years ago

btw, there was work done within scikit-bio on this. I _think_the present branch is this one

wasade commented 9 years ago

@rob-knight, it is likely that numpy and scipy are already highly parallel here. We typically do not link against OpenBLAS but it is natively parallel and numpy/scipy will take advantage of it if linked. That was one of subtle but potentially important things we learned at DS4DS

ElDeveloper commented 9 years ago

Looks like @mortonjt, has already tackled my share of this, issue, what method should I focus on then?