compute_correlation() intent
perone opened this issue · comments
Hi @tfjgeorge, thanks for the library, a very useful framework. I have a quick question regarding the compute_correlation
you're calling in the notebook, why is that being done before visualization of the FIM ? Not sure if I understood what is the intention. Thank you !
Hi, thank you for your interest in NNGeometry.
compute_correlation
takes a covariance matrix, and computes the corresponding correlation matrix (i.e. by normalizing by the standard deviation). It is just more visually appealing to plot the correlation rather than looking at a covariance matrix, because otherwise we would just be looking at a matrix with very few high components (yellow pixels in the notebook), so basically a dark blue square with very few yellow pixels.
I suggest you just try removing compute_correlation
to just see the difference.
Best, Thomas
Thanks @tfjgeorge, make total sense, I tried also to visualize it but was struggling w/ the colormaps, so I suspected it would be because of viz. Thanks again for the open-sourcing it!