Implement entropy based feature selection

[mrohban]

Here is the entropy based feature selection, Marie investigated in her internship. It uses Rcpp and some other libraries to make it computationally more efficient.

[mrohban]

@shntnu I'm really confused with this. I've reverted all the commits to see if the base has caused the problem (in the branch named as tmp). And I noticed it is the case. Do you have any idea, what is wrong that causes segmentation fault at the end of tests? (this doesn't happen when I check the package locally).

[shntnu]

@mrohban Does devtools::check() pass on your computer (for the master and tmp branches`)?

[mrohban]

@shntnu Yes, it passes for both.

[shntnu]

Ok - seems to be a travis issue. I have cleared the cache and am restarting the build. Let's see what happens https://travis-ci.org/cytomining/cytominer/builds/317344745

On Sat, Dec 16, 2017 at 12:20 PM, mrohban notifications@github.com wrote:

@shntnu https://github.com/shntnu Yes, it passes for both.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/cytomining/cytominer/pull/118#issuecomment-352197108, or mute the thread https://github.com/notifications/unsubscribe-auth/ABJ4PFusNe_KI5M4O7lklhP2xkoDDct_ks5tA_vqgaJpZM4RCaRc .

[shntnu]

I've restarted the builds after clearing the travis cache. This should fix it.

[mrohban]

Great, thanks!

[shntnu]

@mrohban I profiled this code and I think we should stick with an R implementation. There are some performance gains (130s vs 170s for k = 1500 dimensions), but even that goes away if you just use foreach and 4 cores (60s)

http://rpubs.com/shantanu/svd_entropy

Let me know what you think

[shntnu]

This PR was supplanted by https://github.com/cytomining/cytominer/pull/123

[shntnu]

The files from this branch are archived at https://github.com/broadinstitute/imaging_metric_comparison/blob/master/cytominer-issues-50.zip