gallantlab
Totally outlandish multi-colormap idea
An oft-requested feature is to create the fancy and colorful RGB flatmaps automatically. Generally, this is only an option for 3 gaussian-distributed variables, like PCA maps, IE Alex's paper.
What if we created some sort of manifold mapping algorithm that lets you project an arbitrary number of dimensions down onto 3 color dimensions? We'll have a big fat warning saying that this is not reproducible. This could be nicely implemented in dataviews by allowing higher-than-2-dimensional views.
Thoughts? Comments?
Meh, I think this likely falls outside the realm of things pycortex. I think we should definitely make it easier to create RGB-colormapped things though, so that you can just toss some variables into an “RGBVolumeData” or whatever and it’ll nicely scale and crop them so that it looks good. All we would provide would be a wrapper around sklearn or tsne or w/e anyway, and people can do that on their own, no?
On Apr 29, 2014, at 9:40 PM, James Gao [email protected] wrote:
An oft-requested feature is to create the fancy and colorful RGB flatmaps automatically. Generally, this is only an option for 3 gaussian-distributed variables, like PCA maps, IE Alex's paper.
What if we created some sort of manifold mapping algorithm that lets you project an arbitrary number of dimensions down onto 3 color dimensions? We'll have a big fat warning saying that this is not reproducible. This could be nicely implemented in dataviews by allowing higher-than-2-dimensional views.
Thoughts? Comments?
— Reply to this email directly or view it on GitHub.
